Compare commits

...

421 Commits

Author SHA1 Message Date
Enrico Ros d936629ead 2.0.3: update readme 2026-02-02 15:48:02 -08:00
Enrico Ros 9bd1a66208 2.0.3: update news 2026-02-02 15:43:30 -08:00
Enrico Ros 1a0c029ee8 2.0.3: update package 2026-02-02 15:26:53 -08:00
Enrico Ros e7be228703 Roll AIX 2026-02-02 15:16:57 -08:00
Enrico Ros 0ab4dc972f Dockerfile: suppress CopyIgnoredFile warning for whitelist-style .dockerignore 2026-02-02 15:16:15 -08:00
Enrico Ros 5f1ca8954f Force touch to doubleClick support (disabled) 2026-02-02 15:09:11 -08:00
Enrico Ros 3ec1b033ce BlockEdit_TextFragment: support 'xs' editing. #961 2026-02-02 14:47:56 -08:00
Enrico Ros 0caf27af9b LLMs: skip prod warning for connection errors, they're still shown in the router warnings in dev 2026-02-02 12:55:38 -08:00
Enrico Ros bd67e14fa4 Debug: Wire: off 2026-02-02 12:37:13 -08:00
Enrico Ros 494c3b542c AIX/LLMs: LMStudio: generic conversion 2026-02-02 12:26:36 -08:00
Enrico Ros 8e0884eb64 AIX/LLMs: LMStudio: convert WebP -> JPG instead 2026-02-02 12:21:54 -08:00
Enrico Ros 73c4dc4ac8 AIX/LLMs: support for WebP -> PNG conversion at the hotfix (pre-CGR) stage 2026-02-02 12:18:31 -08:00
Enrico Ros d77274058d LLMs: LMStudio: use native API for detailed model information 2026-02-02 12:07:45 -08:00
Enrico Ros 0c8460419b AIX: ImageContentPart: allow not detail 2026-02-02 11:34:10 -08:00
Enrico Ros eabb589390 AIX: relax error parsing to .error: { message: '..' } 2026-02-02 11:29:52 -08:00
Enrico Ros 62f860ae93 Debug: Wire: clip curl to 4096 inner 2026-02-02 11:26:57 -08:00
Enrico Ros 605aae873c Roll packages with net removal 2026-01-31 16:40:23 -08:00
Enrico Ros 62e9ee5b05 Roll react hook form which shall be even lighter 2026-01-31 16:32:02 -08:00
Enrico Ros d686f5d143 Roll verified changes 2026-01-31 16:26:44 -08:00
Enrico Ros 3922f232ae Roll some types and prettier 2026-01-31 16:22:31 -08:00
Enrico Ros 6735b438d3 Roll Next to fix CVE 2026-01-31 16:15:57 -08:00
Enrico Ros fb1e30ab32 Roll PostHog-node 2026-01-31 16:11:32 -08:00
Enrico Ros 0ec06edb57 Roll PostHog-js 2026-01-31 16:10:21 -08:00
Enrico Ros 2a52673c56 Merge pull request #959 from enricoros/dependabot/github_actions/docker/login-action-3.7.0
chore(deps): bump docker/login-action from 3.6.0 to 3.7.0
2026-01-31 16:09:49 -08:00
Enrico Ros cc20d00d8a Drive picker: improve token handling with expiration 2026-01-31 15:52:24 -08:00
Enrico Ros 3d9201f7dc Drive picker: add a button to close and reset 2026-01-31 15:12:26 -08:00
dependabot[bot] 176732a6c0 chore(deps): bump docker/login-action from 3.6.0 to 3.7.0
Bumps [docker/login-action](https://github.com/docker/login-action) from 3.6.0 to 3.7.0.
- [Release notes](https://github.com/docker/login-action/releases)
- [Commits](https://github.com/docker/login-action/compare/5e57cd118135c172c3672efd75eb46360885c0ef...c94ce9fb468520275223c153574b00df6fe4bcc9)

---
updated-dependencies:
- dependency-name: docker/login-action
  dependency-version: 3.7.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
2026-01-31 11:32:23 +00:00
Enrico Ros 39815b3af3 CustomMarkdownRenderer: reduce log 2026-01-31 03:13:46 -08:00
Enrico Ros bcce517089 Drive: token provider interface 2026-01-31 03:01:36 -08:00
Enrico Ros a4b50d0d97 OpenAI: Raise instancelimit to 10 2026-01-31 03:01:36 -08:00
Enrico Ros 2a124e7588 Drive: text bits 2026-01-30 20:04:24 -08:00
Enrico Ros a85556ab5b Attach content (docs, images, pdf, etc.) from Google Drive. Fixes #943 2026-01-30 19:18:58 -08:00
Enrico Ros cef93d6084 Attachments: cloud types 2026-01-30 17:00:57 -08:00
Enrico Ros 207e257778 AIX: Demuxers: add recovery of incomplete buffers and server-side logging 2026-01-30 14:23:29 -08:00
Enrico Ros 12203daa22 LLMs: New Heuristic model enumeration for the OpenAI protocol 2026-01-30 12:59:14 -08:00
Enrico Ros 27f8e9248d ModelsList: hidden LLMs get faded chips 2026-01-30 12:42:37 -08:00
Enrico Ros 51384dc984 AIX: remove unused OAI Realtime IF 2026-01-30 12:36:44 -08:00
Enrico Ros bc76cbb5ad Server-side fetchers: opt-in optional headers map 2026-01-30 11:55:18 -08:00
Enrico Ros 5a1ca83f6d Server-side fetchers: debug print headers 2026-01-30 11:50:27 -08:00
Enrico Ros c9f585f808 Server/Client Wire improvement, including listModels 2026-01-30 11:39:35 -08:00
Enrico Ros 9f559e1dbf LLMs: Groq: re-add SDAIA 2026-01-30 11:10:52 -08:00
Enrico Ros e458bca1a7 Revert "LLMs: OpenAI Responses API by default for new models. Fixes #937"
This caused all unknown models to default to the responses API.
We need heuristics for determining OpenAI vs OpenAI-compatible.

This reverts commit a6862d8c58.
2026-01-30 03:09:16 -08:00
Enrico Ros 43d2226019 AIX: Fix LiteLLM breakage 2026-01-30 02:01:21 -08:00
Enrico Ros 122bc34701 AIX: don't retry insufficient balance errors 2026-01-29 23:46:19 -08:00
Enrico Ros e01358e268 LRS: add back mmmu 2026-01-29 23:00:59 -08:00
Enrico Ros 847c84c3e6 LLMs: remove cbaMmmu 2026-01-29 22:39:47 -08:00
Enrico Ros b11cac4328 LLMs: sync CB 2026-01-29 22:31:09 -08:00
Enrico Ros f617b06109 LLMs: copy over initialtemperature if/when available on a per-model basis for 'fromManualMapping' 2026-01-29 22:31:08 -08:00
Enrico Ros 345ccf3369 DLLM/ModelDescription: remove trainigDataCutoff: not used anymore in this day and age 2026-01-29 21:43:03 -08:00
Enrico Ros d111b8af62 LLMs: Update XAI verified 2026-01-29 21:23:43 -08:00
Enrico Ros 8f964c5c49 LLMs: Update OpenPipe 2026-01-29 21:22:34 -08:00
Enrico Ros b6f3f4538f LLMs: Update OpenAI Instant models 2026-01-29 21:22:02 -08:00
Enrico Ros f6dd30d5d8 LLMs: Update Mistral 2026-01-29 21:21:12 -08:00
Enrico Ros af8b79f849 LLMs: Update Groq 2026-01-29 21:20:24 -08:00
Enrico Ros 0cfccc423b LLMs: Update Gemini (deprecations) 2026-01-29 21:17:40 -08:00
Enrico Ros f9a5d582d4 LLMs: Update Ollama 2026-01-29 21:14:25 -08:00
Enrico Ros 684e00d594 Model Services: bits2 2026-01-29 21:08:20 -08:00
Enrico Ros 3cd2df0b50 Model Services: Duplicates removal 2026-01-29 21:01:44 -08:00
Enrico Ros 02197f4ee6 OpenAI: fix sequence number validation - broken by LiteLLM. Fixes #957 2026-01-29 13:49:46 -08:00
Enrico Ros f9049a3fea ModelsWizard: fix build 2026-01-29 13:36:05 -08:00
Enrico Ros 462bddc271 Model Service: Service menu 2026-01-29 13:16:25 -08:00
Enrico Ros f79000cf39 Model Service: 3 dots button 2026-01-29 13:16:25 -08:00
Enrico Ros 1d95273f4d Models Refetch: update icon 2026-01-29 13:16:25 -08:00
Enrico Ros 6c4579f434 LLM Duplication copy 2026-01-29 13:16:25 -08:00
Enrico Ros 4ef56ade21 LLMs: OpenAI: default 5.2 to medium, since we have the no-thinking model too 2026-01-29 13:05:16 -08:00
Enrico Ros 7c1369d6e9 LLM Store: allow for removal/reset service-wide 2026-01-29 13:05:16 -08:00
Enrico Ros 533d54b106 LLM Options: reflect status elsewhere 2026-01-29 12:36:33 -08:00
Enrico Ros cce0ca6560 LLM Options: Clone advanced. #941 2026-01-29 12:36:01 -08:00
Enrico Ros e87ce2593c LLM Options: bits 2026-01-29 12:32:32 -08:00
Enrico Ros 431dc8b667 LLM Temperature: allow for Vendor-set no temperature (future) 2026-01-29 12:32:32 -08:00
Enrico Ros 5caf614bf7 LLM Clones: correctly group services when updating, inserting custom models 2026-01-29 12:32:32 -08:00
Enrico Ros ecf9703570 LLM Clones: User can create clones. #941 2026-01-29 12:32:32 -08:00
Enrico Ros e7641393a0 LLMParametersEditor: fix controlled component 2026-01-29 12:32:31 -08:00
Enrico Ros 2201f6ff5a ModelsList: memo feature chips 2026-01-29 12:31:59 -08:00
Enrico Ros 557e1ce293 Clipboard Content Transform: fix copying of code 2026-01-29 11:10:56 -08:00
Enrico Ros cbe9a6b9a5 LLMOptions: style a bit 2026-01-29 00:36:16 -08:00
Enrico Ros 9bbcb038d4 LLMOptions: launch purpose (not on) 2026-01-29 00:28:28 -08:00
Enrico Ros 3602204420 LLMOptions: add commands 2026-01-28 23:52:10 -08:00
Enrico Ros 6f485e5589 GoodModal: shrink on xs 2026-01-28 23:49:29 -08:00
Enrico Ros 2f46a3dfaf LLM Options: Details/Override 2026-01-28 17:45:39 -08:00
Enrico Ros 267845bba3 LLM Options: LLM actions 2026-01-28 17:21:52 -08:00
Enrico Ros 6f33a8eebf LLMs: improve variants handling across the board. #941 2026-01-28 17:03:59 -08:00
Enrico Ros b0d2b09a2e AIX: Fix show injector 2026-01-28 17:03:58 -08:00
Enrico Ros c699b6b16b OpenAI: add a no-thinking variant for GPT-5.2, which allows to change temperature. #941 2026-01-28 17:03:58 -08:00
Enrico Ros 1789bac28d Roll AIX 2026-01-28 03:21:30 -08:00
Enrico Ros 60c05f615f AIX Debugger: unfiler 2026-01-28 03:02:12 -08:00
Enrico Ros bd84523671 LLMs: Gemini: removed models 2026-01-28 02:48:58 -08:00
Enrico Ros eb21b9c770 Speex: open config if closed 2026-01-28 02:42:24 -08:00
Enrico Ros ff3ac11afb LLMParametersEditor: survive undefined temperature 2026-01-28 02:23:39 -08:00
Enrico Ros 1ef8c3d02b LLMs: Gemini: temperature defaults to 1.0 and can be changed 2026-01-28 02:01:55 -08:00
Enrico Ros 2ebaf6279b AIX: Injector: works well 2026-01-28 01:42:50 -08:00
Enrico Ros a5ee40e184 Speex: fix build 2026-01-28 01:15:56 -08:00
Enrico Ros b17a97eac7 AIX: request body injection. #953 2026-01-28 01:09:32 -08:00
Enrico Ros 63908bfaf6 Speex: update README 2026-01-27 23:53:27 -08:00
Enrico Ros 3f9a419a19 Speex: +Inworld Config 2026-01-27 23:50:33 -08:00
Enrico Ros bae691e33e Speex: +Inworld 2026-01-27 23:50:33 -08:00
Enrico Ros 91539346ee Speex: Fix #624 2026-01-27 22:08:39 -08:00
Enrico Ros 4842ca81b3 Speex: Cancellable preview 2026-01-27 22:08:38 -08:00
Enrico Ros 9c77a1a4ab Speex: Chunk test: remove button 2026-01-27 22:08:36 -08:00
Enrico Ros 4af284be42 Speex: prevent voice/engine mismatch 2026-01-27 22:08:15 -08:00
Enrico Ros 6aec68bb3c Speex: fix unlimited chunking 2026-01-27 22:08:15 -08:00
Enrico Ros d4e2b0834f Speex: allow inner calls, for bytes access 2026-01-27 20:23:10 -08:00
Enrico Ros 24c2702f96 Speex: ghost key to 'tts' for remembering the pass 2026-01-27 20:18:05 -08:00
Enrico Ros 4691fc9bad Speex: pre-wrap errors 2026-01-27 20:17:58 -08:00
Enrico Ros 8c6c60b6f1 Speex: fix webspeech voice selection 2026-01-27 20:17:02 -08:00
Enrico Ros bc482407fe Speex: overhaul for chunking and future synchroniciy / controllability 2026-01-27 20:11:49 -08:00
Enrico Ros ff05593db8 AudioAutoPlayer: handler for any streaming or full-file play, with stop and await ending 2026-01-27 20:11:49 -08:00
Enrico Ros 3d304d9374 AudioLivePlayer: extra safety 2026-01-27 20:11:49 -08:00
Enrico Ros 1734f0c2f1 AudioLivePlayer: extra safety, if stop was called already, waitForPlaybackEnd would return immediately 2026-01-27 19:38:48 -08:00
Enrico Ros 1b25e5df85 AudioLivePlayer: anti-leak the objectUrl and resolve wait on stop 2026-01-27 19:38:48 -08:00
Enrico Ros ea8eb32b0b AudioLivePlayer: await until done 2026-01-27 19:38:48 -08:00
Enrico Ros 614a1f95de AudioPlayer: improve straight play function 2026-01-27 19:38:48 -08:00
Enrico Ros d36bc28914 blobUtils: combine arrayBuffers 2026-01-27 19:38:48 -08:00
Enrico Ros deec48d7c1 CC: enable gh issues list 2026-01-27 19:38:48 -08:00
Enrico Ros b318ec8d39 Merge pull request #951 from jayrepo/patch-3
Add middleware.ts to docker image
2026-01-26 23:11:08 -08:00
Jay Chen b4b0e2befc Update .dockerignore 2026-01-27 14:22:22 +08:00
Enrico Ros 51d3fe13da Roll AIX 2026-01-26 19:51:41 -08:00
Enrico Ros 58220216d3 LLMs/AIX: support for Kimi Thinking On/Off 2026-01-26 19:43:06 -08:00
Enrico Ros cac75cca42 LLMs: Detail Kimi-K2.5 2026-01-26 19:24:09 -08:00
Enrico Ros 47f247907f LLMs: Add Kimi-K2.5 2026-01-26 18:57:50 -08:00
Enrico Ros 81e04b7322 ChatDrawer: buckets: fewer splits, more stability 2026-01-25 19:39:05 -08:00
Enrico Ros 56a964b700 AIX debugger: highlight non-conversation frames 2026-01-25 19:37:44 -08:00
Enrico Ros 458341d79f AIX debugger: don't auto-advance frame for support operations 2026-01-25 19:37:43 -08:00
Enrico Ros d1d212b075 Copy: intercept to others too 2026-01-24 20:24:52 -08:00
Enrico Ros 59c9996489 LLM Params update: TS improvements 2026-01-24 19:33:25 -08:00
Enrico Ros bf8221a2f1 LLM Params update: MDS with int (not float) 2026-01-24 19:33:25 -08:00
Enrico Ros 787a11a040 LLM Params system: Improve types definition 2026-01-24 19:33:25 -08:00
Enrico Ros 05d114be2f Copy: redo visual copy and copy interception (Ctrl+c, etc) for Plain text and HTML 2026-01-24 19:33:25 -08:00
Enrico Ros 3c04a7dbac Copy: also disable copy of collapsed Expanders 2026-01-24 19:33:25 -08:00
Enrico Ros 1673e1148d Copy: annotate what to Not copy 2026-01-24 19:33:25 -08:00
Enrico Ros de416b035d Copy: remove overlay copy button on the message 2026-01-24 19:33:25 -08:00
Enrico Ros 08aaf2989d Beam: always show model selector for Custom fusion 2026-01-24 15:49:54 -08:00
Enrico Ros a50964060c Stop ResizeObserver issue notifications 2026-01-24 15:41:41 -08:00
Enrico Ros 54b6108719 Beam: hide Merge Model selector for Custom fusion 2026-01-24 15:32:58 -08:00
Enrico Ros 585e5c254a Roll AIX 2026-01-23 11:42:32 -08:00
Enrico Ros 477808c9bb AIX: OAI Responses: allow 'failed' on web_search_call 2026-01-23 11:42:22 -08:00
Enrico Ros 6c58a2b688 AIX: OAI Responses: allow 'failed' on web_search_call 2026-01-23 11:42:10 -08:00
Enrico Ros c9854bf30f LLMs: OpenAI: partial Reasoning Effort updates. #944 2026-01-23 11:28:03 -08:00
Enrico Ros cfed4bbd41 LLMs: OpenAI: remove restore markdown on GPT-5 models 2026-01-23 11:20:53 -08:00
Enrico Ros 2dd6485b0e LLMParametersEditor: align to XAI X search default off 2026-01-23 11:19:32 -08:00
Enrico Ros bf1dd5b860 LLMs: Toggle code execution in Params Editor only (not quick) 2026-01-23 10:52:32 -08:00
Enrico Ros 765c373f7d LLMs: OAI: allow code execution 2026-01-23 10:52:32 -08:00
Enrico Ros 32d752e82b LLMs: OAI Responses: (unused) parameter for OpenAI code execution 2026-01-23 10:35:58 -08:00
Enrico Ros 4623e438fa AIX: OAI Responses: code interpreter Hosted tool def 2026-01-23 10:35:08 -08:00
Enrico Ros 8a44ff396f AIX: XAI: relax annotation title presence 2026-01-23 10:16:59 -08:00
Enrico Ros 086d7ecae4 Speex: TTS character limit settings. Fixes #942 2026-01-23 10:05:35 -08:00
Enrico Ros d6adebb711 Attachment buttons: full name in tooltip. Fixes #946 2026-01-23 09:55:29 -08:00
Enrico Ros 8325fe7b3c Roll AIX 2026-01-23 09:29:41 -08:00
Enrico Ros 7cf83f878b AIX: XAI: Response API Request + wiretypes. Fixes #938 2026-01-23 04:29:13 -08:00
Enrico Ros 597ba26424 AIX: Code Executor wires 2026-01-23 04:29:06 -08:00
Enrico Ros 7bccea47f5 AIX: OpenAI Responses: parse Code Execution and Custom Tools 2026-01-23 04:29:06 -08:00
Enrico Ros 5770116779 DMessage: Code Executor 2026-01-23 04:29:06 -08:00
Enrico Ros 0679144f69 LLMs/AIX: XAI new parametrization 2026-01-23 04:29:05 -08:00
Enrico Ros c9fd288b52 AIX: OpenAI chatCompletions: remove obsolete X search params 2026-01-23 04:16:52 -08:00
Enrico Ros 9ae449fcfd LLMs: type check server params against the client params def 2026-01-23 00:46:07 -08:00
Enrico Ros 249f67f796 AIX: improve dispatch messaging validation for 4 protocols 2026-01-23 00:46:07 -08:00
Enrico Ros e91c0bb554 AIX: stripUndefined 2026-01-23 00:28:05 -08:00
Enrico Ros 5e306d9598 AIX: XAI: models update 2026-01-23 00:25:33 -08:00
Enrico Ros 42ebc81cbb AIX: XAI: models update 2026-01-22 23:56:11 -08:00
Enrico Ros f624c37db5 AIX: XAI: models update 2026-01-22 17:51:03 -08:00
Enrico Ros 22b6f42936 AIX: OAI: Responses wiretypes changes 2026-01-22 17:25:54 -08:00
Enrico Ros 760c66cac8 Attachments: reposition menu to see the tokens bar 2026-01-22 17:18:08 -08:00
Enrico Ros 1d91e9da03 Attachments: client-side Markdown conversion and Text/HTML cleanup & Markdown conversion 2026-01-22 17:17:57 -08:00
Enrico Ros 7eac409ec6 AIX: XAI: model removal 2026-01-22 13:54:36 -08:00
Enrico Ros 128558420c AIX: XAI: model validation 2026-01-22 13:54:17 -08:00
Enrico Ros ca3e664690 AIX: spill part type cleanup 2026-01-22 13:12:29 -08:00
Enrico Ros 7eb37462d7 LLMs: update ollama models 2026-01-22 03:06:17 -08:00
Enrico Ros 31e02c2d39 CC: slashcommands: update llms:ollama 2026-01-22 03:02:55 -08:00
Enrico Ros 003a68b9b8 CustomMarkdownRenderer.tsx: allow for <br/> tags inside Table Cells. Fixes #939 2026-01-22 02:14:42 -08:00
Enrico Ros f418708389 AIX: Inspector: improve render 2026-01-22 01:55:14 -08:00
Enrico Ros d23a564035 AIX: Inspector: reassembler 'transport' data and ui 2026-01-22 01:45:20 -08:00
Enrico Ros 7fe586244c AIX: Inspector: debugger object client sync 2026-01-22 01:38:33 -08:00
Enrico Ros f1a597cdc6 AIX: move Inspector frames selector 2026-01-22 01:36:54 -08:00
Enrico Ros 9b68c8f58c CC: slashcommands: update llms:ollama 2026-01-22 01:06:11 -08:00
Enrico Ros be5b57ea71 LLMs: shared model definition validators 2026-01-22 00:54:00 -08:00
Enrico Ros 425c82f26d CC: slashcommands: improve parsing of ollama models 2026-01-22 00:08:07 -08:00
Enrico Ros 942421c1fb LLMs: together: fix key validation 2026-01-21 23:52:15 -08:00
Enrico Ros b1184f6928 AIX: CSF: also support client-side exceptions 2026-01-21 23:39:10 -08:00
Enrico Ros ffeb6d1b98 LLMs: models bits 2026-01-21 23:34:31 -08:00
Enrico Ros b2718b56b7 CC: shashcommands: improve kimi 2026-01-21 23:16:57 -08:00
Enrico Ros 455f834957 CC: allow llms:update 2026-01-21 23:05:56 -08:00
Enrico Ros 8a14c80ff8 CC: slashcommands: llms:groq update 2026-01-21 23:04:09 -08:00
Enrico Ros e268e733c7 LLMs: Groq: overlap check 2026-01-21 23:03:29 -08:00
Enrico Ros 8933a8dfb3 LLMs: Gemini: deprecations 2026-01-21 22:53:15 -08:00
Enrico Ros 9796cc525c LLMs: xAI verified 2026-01-21 22:37:11 -08:00
Enrico Ros cdbf9a9190 Speex: CSF support - auto-detected from linked services - client-unbundled-dynamic 2026-01-21 18:38:43 -08:00
Enrico Ros c26792292d Speex: extract rpc common core 2026-01-21 18:14:03 -08:00
Enrico Ros 4698e0ee03 LLMs: OpenAI: remove /v1/ from the user input, to immedialy give feedback on the correct way of doing it 2026-01-21 16:58:57 -08:00
Enrico Ros 68afcb2f4b AIX: OpenAI: disable reasoning summaries when disabling Streaming as well. Fixes #932 2026-01-21 16:52:36 -08:00
Enrico Ros e8f61e46e3 AIX: Fetchers: don't retry on 'request too large' 2026-01-21 16:22:50 -08:00
Enrico Ros 317bb2b7c8 useLLMSelect: preserve scrolling on cat toggle 2026-01-21 16:00:43 -08:00
Enrico Ros d1b3c6b468 ContentFragments: improve zero state add text look. #934 2026-01-21 16:00:42 -08:00
Enrico Ros b35eccc984 AIX: increase resilience to new values across 3 parsers. Fixes #918
Note: relaxed throws on Anthropic and Gemini (all throws), while on OAI-Responses we are raising to throws in dev.
2026-01-21 15:08:17 -08:00
Enrico Ros a780c92047 AIX: speculative support for keepalives on chatCompletions API. Fixes #918 2026-01-21 14:36:40 -08:00
Enrico Ros 5fc65698ba Collapsible Model Groups, #936 2026-01-21 14:10:30 -08:00
Enrico Ros c923b5ec4c Restyle llm select model groups #936 2026-01-21 13:36:43 -08:00
Enrico Ros 609b2b9a7b Group models correctly by service in Beam drop down list. Fixes #936 2026-01-21 13:36:31 -08:00
Enrico Ros a257278004 dMessageUtils: OSB 2026-01-21 13:15:26 -08:00
Enrico Ros 273daed634 Chat Messages: label heuristic to ignore the vendor-N- 2026-01-21 13:13:49 -08:00
Enrico Ros a6862d8c58 LLMs: OpenAI Responses API by default for new models. Fixes #937 2026-01-21 13:07:58 -08:00
Enrico Ros 323e5b4ea7 LLMs: OpenAI OSB speculative support 2026-01-21 13:06:47 -08:00
Enrico Ros 89217a5308 CC: allow gh issue view 2026-01-21 13:03:05 -08:00
Enrico Ros a45e995d2f AIX: OpenRouter: improve reasoning through OpenAI-completions across models. Fixes #893 2026-01-20 11:32:11 -08:00
Enrico Ros 8700b4c8ca Roll AIX 2026-01-20 02:07:55 -08:00
Enrico Ros 1f7f5fb488 Data (personas): more concise default 2026-01-20 01:43:45 -08:00
Enrico Ros afde8ee864 LLMs: OpenRouter: 'verbosity' support for OpenAI gpt-5 models, and Anthropic Claude Opus 4.5 (remapped to reasoning effort by OpenRouter). Fixes #927 2026-01-20 01:42:54 -08:00
Enrico Ros 3884c26b15 LLMs: OpenAI: Reviewed 5.1 Instant / 5.2 Params support. Closes #930 2026-01-20 01:21:45 -08:00
Enrico Ros 24dce7eae9 LLMs: further improve the LLMParametersEditor's sync with the dialog. #926 2026-01-20 01:09:04 -08:00
Enrico Ros 1db4e9b771 LLMs: Anthropic: Fix Effort in the LLM Options Dialog. Fixes #926 2026-01-20 00:31:39 -08:00
Enrico Ros b2ed7eae00 CC: CMDs: llms:align-params-uis 2026-01-20 00:31:39 -08:00
Enrico Ros 3169fd67e8 LLMs: OpenRouter: fix Antrhopic thinking models. Fixes #925 2026-01-20 00:15:00 -08:00
Enrico Ros 773ceb1396 Tools: data/llm: remove obsolete registry 2026-01-19 23:40:56 -08:00
Enrico Ros 8c62ee1720 LLMs/AIX: fully remove moderation
Doesn't seem to be used anymore by anyone, nor was active in the code
2026-01-19 23:32:15 -08:00
Enrico Ros 5fa1f52922 LLMs: openai service setup: bits 2026-01-19 23:09:20 -08:00
Enrico Ros d2180c010c LLMs: Helicone: start unformize 2026-01-19 22:42:16 -08:00
Enrico Ros b73df7b2ce LLMs: OpenAI: Autocomplete + suggest hosts for Chutes, Fireworks, Novita. #921 2026-01-19 22:42:16 -08:00
Enrico Ros 971f737846 LLMs: support Novita.ai models with capability auto-detection. #921 2026-01-19 22:39:32 -08:00
Enrico Ros a393353907 LLMs: Azure: rename custom field 2026-01-19 22:39:32 -08:00
Enrico Ros 751f609554 LLMs: OpenAI: unify paths 2026-01-19 22:39:31 -08:00
Enrico Ros e8cd5c6552 LLMs: Ant: unify paths 2026-01-19 22:39:31 -08:00
Enrico Ros 86e387b270 LLMs: allow OpenAI/Azure OpenAI services renaming. Fixes #922 2026-01-19 15:54:13 -08:00
Enrico Ros 32f15aa621 FormTextField: allow for end decorator 2026-01-19 15:54:13 -08:00
Enrico Ros bfc889a9e5 LLMs: fix reset of non-declared params at models update. Fixes #924 2026-01-19 14:53:46 -08:00
Enrico Ros bd907625a8 UpDate 2026-01-15 17:48:00 -08:00
Enrico Ros 60004926d7 Recommend DC (CSF) for Local services 2026-01-15 16:28:00 -08:00
Enrico Ros ac751dfd1a Roll AIX 2026-01-14 16:42:06 -08:00
Enrico Ros 6828eee17f LLMs: Perplexity: sync 2026-01-14 16:40:01 -08:00
Enrico Ros 19c97f397b LLMs: OpenAI: sync a GPT Audio model 2026-01-14 16:38:32 -08:00
Enrico Ros 0167a8bdd8 LLMs: Mistral: update 2026-01-14 16:36:31 -08:00
Enrico Ros 93e5044603 LLMs: Groq: strings 2026-01-14 16:36:19 -08:00
Enrico Ros 024d930677 LLMs: Gemini: small update 2026-01-14 16:35:22 -08:00
Enrico Ros 98873446a8 LLMs: Ollama: update tags 2026-01-14 16:35:15 -08:00
Enrico Ros 5318b7a406 OCR: cache tesseract's import 2026-01-14 16:30:57 -08:00
Enrico Ros 4a6c3cbcd2 Roll AIX 2026-01-14 15:41:58 -08:00
Enrico Ros ac0a39c202 LLMs: OpenAI: GPT-5.2 Codex support 2026-01-14 15:40:40 -08:00
Enrico Ros 88d39345a5 Attachments: PDF: add a PDF->OCR (via interim images) and an 'Auto' (default)
The Auto mode tries plain Text, then Text to Images to OCR, then falls back to pure images.
2026-01-14 15:09:40 -08:00
Enrico Ros 7aa9cb07b2 OCR: extract util functions 2026-01-14 13:18:44 -08:00
Enrico Ros ef30c8d28d Tesseract: roll package 2026-01-14 13:17:11 -08:00
Enrico Ros 2727f690b4 Attachments: PDFs: use Density to attach images 2026-01-14 10:59:43 -08:00
Enrico Ros 5945c24301 Speex: RPC: fallback to full buffer play if streaming is unavailable (Firefox) 2026-01-14 10:32:41 -08:00
Enrico Ros 7b6aff1f95 AudioLivePlayer: Firefix doesn't support MPEG streaming 2026-01-14 10:32:39 -08:00
Enrico Ros cb0fe3aadd AIX: OpenAI Responses: support for keepalive packets 2026-01-13 19:39:47 -08:00
Enrico Ros 4f9d69f9c2 AIX: Anthropic: Fix for out-of-order block start. Fixes #917 2026-01-12 16:40:16 -08:00
Enrico Ros c18aeabe06 CC: changelog cmd update 2026-01-12 04:18:02 -08:00
Enrico Ros 550742323a CC: changelog cmd 2026-01-12 04:02:17 -08:00
Enrico Ros c71f789a08 Draw PromptComposer: fix for CJK 2026-01-12 03:44:00 -08:00
Enrico Ros a9b4b195bf Extend #916 to InlineTextArea and editing of a Text Fragment 2026-01-12 03:37:02 -08:00
Enrico Ros 52e8177f42 Simplify #916 2026-01-12 03:36:33 -08:00
Enrico Ros b0743efc48 Merge branch 'fork/tantanorange/feat/bug-Input-Method-Editor' 2026-01-12 03:30:25 -08:00
Enrico Ros 6dfd652dac LLMParametersEditor: shrink label 2026-01-12 03:23:42 -08:00
Enrico Ros 3f93cb2e6d Phone: work over an empty conversation id 2026-01-12 03:11:49 -08:00
Enrico Ros 8f7b9b7f19 Optima Page Heading: support disabled 2026-01-12 02:10:38 -08:00
Enrico Ros abff89ab6b CC: ignore temp files 2026-01-12 02:10:20 -08:00
tantanorange d4f03f743a bug(issue-784): fixed 'Enter' trigger unexpected Chat under Iput Method Editor. 2026-01-11 18:32:55 -08:00
Enrico Ros c3714f6651 Camera Capture: darker 2026-01-10 15:20:30 -08:00
Enrico Ros 9b4d0ddf2f Merge pull request #913 from enricoros/dependabot/github_actions/actions/download-artifact-7.0.0
chore(deps): bump actions/download-artifact from 4.3.0 to 7.0.0
2026-01-10 14:46:32 -08:00
Enrico Ros 2c9ac2f549 Merge pull request #914 from enricoros/dependabot/github_actions/actions/upload-artifact-6.0.0
chore(deps): bump actions/upload-artifact from 4.6.2 to 6.0.0
2026-01-10 14:46:20 -08:00
dependabot[bot] c1292de2a0 chore(deps): bump actions/upload-artifact from 4.6.2 to 6.0.0
Bumps [actions/upload-artifact](https://github.com/actions/upload-artifact) from 4.6.2 to 6.0.0.
- [Release notes](https://github.com/actions/upload-artifact/releases)
- [Commits](https://github.com/actions/upload-artifact/compare/ea165f8d65b6e75b540449e92b4886f43607fa02...b7c566a772e6b6bfb58ed0dc250532a479d7789f)

---
updated-dependencies:
- dependency-name: actions/upload-artifact
  dependency-version: 6.0.0
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
2026-01-10 11:54:05 +00:00
dependabot[bot] 21d5e4cd29 chore(deps): bump actions/download-artifact from 4.3.0 to 7.0.0
Bumps [actions/download-artifact](https://github.com/actions/download-artifact) from 4.3.0 to 7.0.0.
- [Release notes](https://github.com/actions/download-artifact/releases)
- [Commits](https://github.com/actions/download-artifact/compare/d3f86a106a0bac45b974a628896c90dbdf5c8093...37930b1c2abaa49bbe596cd826c3c89aef350131)

---
updated-dependencies:
- dependency-name: actions/download-artifact
  dependency-version: 7.0.0
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
2026-01-10 11:54:01 +00:00
Enrico Ros a9495a3e15 Roll eslint 2026-01-09 19:47:24 -08:00
Enrico Ros bff5b3d765 Roll posthog 2026-01-09 19:46:49 -08:00
Enrico Ros a4ff37eecc Roll safe packages 2026-01-09 19:40:14 -08:00
Enrico Ros 460209f486 GH Actions: fix manifest format 2026-01-09 17:36:28 -08:00
Enrico Ros 96c68c86a4 GH Actions: fix tag case 2026-01-09 17:16:22 -08:00
Enrico Ros 8b152fdff8 GH Actions: improve parallelism of the docker image build 2026-01-09 17:10:53 -08:00
Enrico Ros 25c9a52873 Dependabot: basic configuration 2026-01-09 16:40:28 -08:00
Enrico Ros 44302d903c CC Actions: checkout v6 2026-01-09 16:24:31 -08:00
Enrico Ros c7b8668609 GH Actions: docker-image: pin versions 2026-01-09 14:36:28 -08:00
Enrico Ros 7d60df6266 Docker: save 0.5GB 2026-01-09 14:27:04 -08:00
Enrico Ros b7f898a5e5 Docker: move to Node 24 and cleanup. #907 2026-01-09 13:55:02 -08:00
Enrico Ros 04c4dbe4b8 Docker: remove compose version 2026-01-09 13:53:31 -08:00
Enrico Ros 8d04c494df Docker: negate .dockerignore 2026-01-09 13:53:31 -08:00
Enrico Ros a6aadf76f3 Revert "Fix Node 25 build..." - breaks other Node versions.
This reverts commit b70d57d878.
2026-01-09 13:53:30 -08:00
Enrico Ros a685ef97bf AIX: chatGenerate executor: object-string-ellipsize when replaying input (received from the llm) 2026-01-09 13:36:06 -08:00
Enrico Ros d46c29689f AIX: OpenRouter: support image generation through OpenAI's OR-extended API, including supporting advanced Gemini params. Fixes #906 2026-01-09 13:36:06 -08:00
Enrico Ros 65ce07395b Fix drag-to-update on mobile 2026-01-09 13:36:06 -08:00
Enrico Ros cc1542fe95 Wire: improve debug print with object-ellipsize-strings 2026-01-09 13:36:06 -08:00
Enrico Ros b70d57d878 Fix Node 25 build, by using --no-webstorage with node - otherwise Zustand's persist middleware will break the build 2026-01-09 10:45:50 -08:00
Enrico Ros 5aa857362b Merge pull request #912 from enricoros/claude/issue-909-20260109-1034
feat(deepseek): add API Host field to DeepSeek settings
2026-01-09 02:42:29 -08:00
Enrico Ros c92fc34051 Merge pull request #911 from enricoros/claude/issue-902-20251229-1701
docs: remove Midori AI Subsystem section from installation guide
2026-01-09 02:39:46 -08:00
claude[bot] b01e66f12a feat(deepseek): add API Host field to DeepSeek settings
Add the ability for users to configure a custom API host for DeepSeek,
allowing them to use alternative endpoints like https://api.deepseek.com/beta.

Changes:
- Add `deepseekHost` to DDeepseekServiceSettings interface
- Wire deepseekHost to oaiHost in transport layer
- Add API Host form field visible in advanced settings

Closes #909

Co-authored-by: Enrico Ros <enricoros@users.noreply.github.com>
2026-01-09 10:38:11 +00:00
Enrico Ros a88d20784a Roll AIX 2026-01-09 02:27:54 -08:00
Enrico Ros 63486ed6cf LLMs/AIX: support 4 levels for Gemini 3 Flash. Fixes #905 2026-01-09 02:27:53 -08:00
Enrico Ros 3ceec773f2 LLMs: DeepSeek: remove 3.2 Speciale 2026-01-09 02:08:54 -08:00
Enrico Ros 817fa56ec4 LLMs: Anthropic: remove Claude 3 Opus 2026-01-09 02:05:33 -08:00
Enrico Ros 088fb21a90 LLMs: Gemini: update cache pricing 2026-01-09 01:55:53 -08:00
Enrico Ros 79c755a469 Wire: ellipsize AixDemuxers.DemuxedEvents 2026-01-09 01:31:39 -08:00
Enrico Ros a091d3f011 OpenAI: support for gpt-image-1.5 2026-01-09 01:22:26 -08:00
Enrico Ros c7c01a5d7c AIX: Gemini: sync API: FC with multimodal responses, validated FC-Config, retrieval config, multiple voices, seed, responseId 2026-01-09 01:03:29 -08:00
Enrico Ros cdc0f48973 AIX: Gemini: support MISSING_THOUGHT_SIGNATURE 2026-01-09 00:46:10 -08:00
Enrico Ros e884f6b962 LLMs: Gemini: mark Deep Research Pro Preview as hidden for now - Interactions API coming 2026-01-08 20:46:15 -08:00
Enrico Ros 485a9bea71 LLMs: Gemini: remove removed models 2026-01-08 20:44:57 -08:00
Enrico Ros f3c3b667ca LLMs: OpenAI: remove chatgpt-image from llms 2026-01-08 17:51:58 -08:00
Enrico Ros 3b0c4f31b6 LLMs: Gemini: add 3-flash-preview 2026-01-08 17:47:06 -08:00
Enrico Ros 5e54600766 Deps: set peers 2026-01-08 17:43:41 -08:00
claude[bot] c3e54f69b7 docs: remove Midori AI Subsystem section from installation guide
The Midori AI Subsystem is being sunset as announced in issue #902.
This removes the deployment section from the installation documentation.

Closes #902

Co-authored-by: Enrico Ros <enricoros@users.noreply.github.com>
2025-12-29 17:02:49 +00:00
Enrico Ros c4022d1c9b Roll small 2025-12-20 21:53:18 +01:00
Enrico Ros 6e13a78a24 Roll next 2025-12-20 21:40:15 +01:00
Enrico Ros c7cacd9727 LLMs: Gemini: remove 'medium' thinking level support - too early. Fixes #900 2025-12-20 21:08:41 +01:00
Enrico Ros a77110f704 LLMs: remove parameters set to a former enum value if then removed 2025-12-20 21:07:38 +01:00
Enrico Ros 83a6069de5 CC: update workflows 2025-12-14 03:45:26 +01:00
Enrico Ros e9a1890e54 Beam: rays/fusions can edit, delete 2025-12-11 23:50:06 +01:00
Enrico Ros bf928aa06e AIX: 'searching' item state with 5.2 Pro 2025-12-11 20:38:54 +01:00
Enrico Ros b2dc50590c LLMs: OpenAI GPT 5.2 Pro: enable web search 2025-12-11 20:33:53 +01:00
Enrico Ros 229e53ac32 Roll AIX 2025-12-11 20:20:26 +01:00
Enrico Ros 51e8a47615 OpenAI: Support X-Hight reasoning modes for 5.2-class models 2025-12-11 20:19:57 +01:00
Enrico Ros e80b58a412 UI: support 52 / 52Pro X-High/None reasoning efforts 2025-12-11 20:19:37 +01:00
Enrico Ros 48ced8b079 LLMs: support 52 / 52Pro X-High/None reasoning efforts 2025-12-11 20:19:30 +01:00
Enrico Ros c07e2aea1e AIX: support 52 / 52Pro X-High/None reasoning efforts 2025-12-11 20:19:19 +01:00
Enrico Ros f3194aa30e LLMs: Add GPT 5.2 2025-12-11 19:30:23 +01:00
Enrico Ros cb3e4cd951 LLMs: Mistral: update models 2025-12-09 13:33:01 -08:00
Enrico Ros f5d8d029ea CC: update workflows 2025-12-08 16:54:27 -08:00
Enrico Ros 7c946c4126 CC: add cherry-pick 2025-12-07 12:12:30 -08:00
Enrico Ros ded4ea0d69 Personas: disable YouTube transcript (unsupported API) 2025-12-07 12:10:27 -08:00
Enrico Ros c180c549fe BYOM: improve message 2025-12-07 11:53:49 -08:00
Enrico Ros 1f30f1168f Friction: Model Wizard: also warn if some keys are not saved 2025-12-07 11:52:51 -08:00
Enrico Ros 9446f15922 Friction: better remember Wizard model data 2025-12-07 11:51:35 -08:00
Enrico Ros e13b2c9cd9 Tutorial: 'BYOM' message 2025-12-07 11:49:18 -08:00
Enrico Ros e9e14e0292 LLMs: OpenRouter: add to the wizard 2025-12-07 11:46:04 -08:00
Enrico Ros added19656 Roll posthogs 2025-12-05 19:37:01 -08:00
Enrico Ros 4fa3c4d479 Remove old material (wrong) path 2025-12-05 19:32:34 -08:00
Enrico Ros 690738de9a Fix CVE-2025-55182 2025-12-05 18:41:13 -08:00
Enrico Ros cb31d27e68 Copy: strip background/colors on copy (keep font size and structure) 2025-12-05 11:17:17 -08:00
Enrico Ros e6658df123 Attachment: show dl issues on console 2025-12-05 11:15:15 -08:00
Enrico Ros 0b7154a14c LLMs: OpenAI: remove obsoleted models 2025-12-04 15:52:42 -08:00
Enrico Ros 02c1838de5 LLMs: OpenAI: add gpt-5.1-codex-max 2025-12-04 15:52:18 -08:00
Enrico Ros fc455fceb8 LLMs: Mistral: rmeove obsoleted models 2025-12-02 07:56:29 -08:00
Enrico Ros 8d40cdd234 LLMs: Mistral: sort 2025-12-02 07:53:15 -08:00
Enrico Ros 40145c669a LLMs: Mistral: add Mistral-Large and Ministral 2025-12-02 07:53:09 -08:00
Enrico Ros 34d2fc233f LLMs: Mistral: adapt wires 2025-12-02 07:52:55 -08:00
Enrico Ros 670ec0381a Speex: collapse configure - recollapse with shift 2025-12-01 09:11:48 -08:00
Enrico Ros 2128f255fe Speex: collapse configure when not used 2025-12-01 09:09:50 -08:00
Enrico Ros b717bd9a9a Settings: max-height to not jump around too much 2025-12-01 08:58:20 -08:00
Enrico Ros 8aab9311f5 Roll AIX 2025-12-01 08:33:32 -08:00
Enrico Ros ff3e16ea67 DeepSeek: still images are not supported 2025-12-01 08:30:16 -08:00
Enrico Ros 1de039c315 AIX: OpenAI ChatCompletion: remove multipart hotfix (not needed anymore) 2025-12-01 08:27:55 -08:00
Enrico Ros d05e1786d7 Model namings: speciale 2025-12-01 08:16:44 -08:00
Enrico Ros e34b5a7372 AIX: support Deepseek Speciale 2025-12-01 08:16:29 -08:00
Enrico Ros a1b3d1b508 DeepSeek: 3.2 and Speciale 2025-12-01 08:16:18 -08:00
Enrico Ros 1ebccdf420 Speex: Readmes 2025-12-01 02:59:56 -08:00
Enrico Ros e5f674509c 2.0.2 News 2025-11-30 16:54:56 -08:00
Enrico Ros 197a4ae5c0 2.0.2 Package 2025-11-30 16:53:25 -08:00
Enrico Ros 64d2dcf39c AudioLivePlayer: tryfix for the persistent android notification 2025-11-30 15:05:17 -08:00
Enrico Ros caf54c736b Speex: do not stop the playback too early 2025-11-30 14:31:43 -08:00
Enrico Ros 423c2cce28 speakText: port to Speex 2025-11-30 12:51:55 -08:00
Enrico Ros a1af51efcb Call: port to Speex 2025-11-30 06:55:51 -08:00
Enrico Ros ffc1bf9c58 Remove src/modules/elevenlabs 2025-11-30 06:55:51 -08:00
Enrico Ros a54bfdb342 Settings: port to Speex 2025-11-30 06:55:51 -08:00
Enrico Ros 03861d2dbd Speex: map instead of array 2025-11-30 06:38:14 -08:00
Enrico Ros 8c080da6bf Speex: Autoconfig WebSpeech best 2025-11-30 06:38:14 -08:00
Enrico Ros a8c98056b6 Speex: Config UI Done 2025-11-30 06:38:14 -08:00
Enrico Ros 78e663f955 Speex: important fixes 2025-11-30 06:38:14 -08:00
Enrico Ros 70546a5039 Speex: Almost Done 2025-11-30 06:38:14 -08:00
Enrico Ros 30f78b33cb Speex: diable Azure 2025-11-30 06:38:14 -08:00
Enrico Ros 712e8c1f16 Speex: UI update: Selects and Persona Voice changer 2025-11-30 06:38:14 -08:00
Enrico Ros 933dfdfb53 Speex: improve types 2025-11-30 06:38:14 -08:00
Enrico Ros 9ce86b029f Speex: UI settings modal 2025-11-30 06:38:14 -08:00
Enrico Ros 13580cc69d Speex: UI config improvements 2025-11-30 06:38:14 -08:00
Enrico Ros a7dee0002d Speex: debug instrumentation 2025-11-30 06:38:14 -08:00
Enrico Ros c84b2df3fa Speex: fix elevenlabs 2025-11-30 06:38:14 -08:00
Enrico Ros d9471a8684 Speex: fix types 2025-11-30 06:38:14 -08:00
Enrico Ros ef630c2272 Speex: improve UI and errors 2025-11-30 06:38:14 -08:00
Enrico Ros e188c71652 Speex: RPC: shared downstreaming 2025-11-30 06:38:14 -08:00
Enrico Ros 910260c2c8 Speex: UI: credentials edit and add new 2025-11-30 06:38:14 -08:00
Enrico Ros 22752abc38 Speex: relax engine validation 2025-11-30 06:38:14 -08:00
Enrico Ros 92bc3a5d64 Speex: DVoice -> wire_Voice 2025-11-30 06:38:14 -08:00
Enrico Ros 1383752cc1 Speex: reduce logging 2025-11-30 06:38:13 -08:00
Enrico Ros 66af16fb81 Speex: manual refactor 2025-11-30 06:38:13 -08:00
Enrico Ros fc019d7b46 Speex: client cleanups 2025-11-30 06:38:13 -08:00
Enrico Ros ac4f0fcb12 Speex: LocalAI: Preview 2025-11-30 06:38:13 -08:00
Enrico Ros a6c2bc663d Speex: arrange files 2025-11-30 06:38:13 -08:00
Enrico Ros e62ffa02e9 Speex: LocalAI vendor 2025-11-30 06:38:13 -08:00
Enrico Ros a003600839 Speex: some UI 2025-11-30 06:38:13 -08:00
Enrico Ros ea73feb06d Speex: remove elevenlabs, with key migration 2025-11-30 06:38:13 -08:00
Enrico Ros 3bdf69e1b7 Speex: ui: begin 2025-11-30 06:38:13 -08:00
Enrico Ros 590fe78bd1 Speex: client cleanup 2025-11-30 06:38:13 -08:00
Enrico Ros 76187ba0e7 Speex: rpc backend 2025-11-30 06:38:13 -08:00
Enrico Ros 5eba375f4d Speex: add webspeech (with detection) and synthesize-openai 2025-11-30 06:38:13 -08:00
Enrico Ros 8fa6a8251f Speex: vendors, engine store, client, router, skel-synthesize 2025-11-30 06:38:13 -08:00
Enrico Ros 75fa046f30 Speex: centralize capability 2025-11-30 06:38:13 -08:00
Enrico Ros 08a8cd1430 Speex: Types & Client 2025-11-30 06:38:13 -08:00
Enrico Ros 3afbb78a39 Icons: port to PhVoice 2025-11-30 06:38:12 -08:00
Enrico Ros fca6ccd816 Badge: transparent BG to not overlap text. Fixes #889 2025-11-29 14:52:13 -08:00
Enrico Ros 8d351822c1 Niy 2025-11-29 13:25:36 -08:00
Enrico Ros 7d274a31fe AIX: CGR: use shared objectUtils 2025-11-29 12:40:04 -08:00
Enrico Ros e36dde0d25 objectUtils: estimate JSON size, deep clone with string limit, find largestStringPaths 2025-11-29 12:17:28 -08:00
Enrico Ros 51cc6e5ae5 CSF: only show the option for server-side (not client-side) disconnect 2025-11-29 11:12:30 -08:00
Enrico Ros 28d911c617 ElevenLabsIcon: add icon 2025-11-28 05:49:33 -08:00
Enrico Ros b1e9fe58fb objectUtils: add stripUndefined 2025-11-28 04:23:11 -08:00
Enrico Ros 16ba014ade GoodBadge: for 'new' 2025-11-28 04:23:11 -08:00
Enrico Ros e9d5a20c1a FormTextField: support inputSx 2025-11-28 04:23:11 -08:00
Enrico Ros 6e0036f9c4 FormSecretField: crystal clear keys input 2025-11-28 04:23:11 -08:00
Enrico Ros d7e189aa1c FormSliderControl: allow sliderSx 2025-11-28 04:23:11 -08:00
Enrico Ros ea2b444fb2 FormChipControl: alignEnd 2025-11-28 04:23:11 -08:00
Enrico Ros cd1efaf26e FormChipControl: support descriptions 2025-11-28 04:23:11 -08:00
Enrico Ros e47f0e5d43 LanguageSelect: imrove select 2025-11-28 04:23:11 -08:00
Enrico Ros 5284d37984 AudioLivePlayer: ignore a closure error 2025-11-28 04:23:11 -08:00
Enrico Ros 1bf6fa0e4d Browse service: improve error reporting 2025-11-27 19:12:08 -08:00
Enrico Ros fc294c82f1 Pdfjs: lock to 5.4.54
more recent 5.4 have trouble with await import('pdfjs-dist'), throwing.
2025-11-27 18:33:20 -08:00
Enrico Ros 7b1dc49dda Roll pdfjs 2025-11-27 18:19:51 -08:00
Enrico Ros d15ddeea24 Roll react-player 2025-11-27 18:15:19 -08:00
Enrico Ros eaac213859 Ph: add Voice 2025-11-27 18:07:54 -08:00
Enrico Ros 02c1460351 Roll posthog 2025-11-27 18:04:06 -08:00
Enrico Ros 2fff35b7d9 Roll superjson 2025-11-27 18:03:37 -08:00
Enrico Ros c5b9072bde LLMs: LocalAI publish interface 2025-11-26 19:01:44 -08:00
Enrico Ros 8a570e912a CSF: docs 2025-11-26 07:37:56 -08:00
Enrico Ros 1dcc40afb8 CSF: Propagate everywhere 2025-11-26 07:37:09 -08:00
Enrico Ros c2092f8035 BlockPartError: vendor name 2025-11-26 06:50:11 -08:00
Enrico Ros 886c4b411e Revert "Test Edge on node"
This reverts commit 8888fd40cd.
2025-11-26 06:13:28 -08:00
Enrico Ros 8888fd40cd Test Edge on node 2025-11-26 04:56:26 -08:00
Enrico Ros 31cd01bccf BlockPartError: CSF enabled 2025-11-26 04:42:50 -08:00
Enrico Ros c59b221004 BlockPartError: allow retrying disconnected errors too 2025-11-26 04:27:52 -08:00
Enrico Ros cb3cc3e74c PostHog: disable the info level 2025-11-26 04:05:03 -08:00
Enrico Ros 9e90015fcc PostHog: disable the info level 2025-11-26 03:56:55 -08:00
Enrico Ros 95e0517056 60s - disable any maxDuration 2025-11-26 03:56:25 -08:00
Enrico Ros 2b2f47915f AIX: OpenAI: Fix CSF! 2025-11-26 03:11:12 -08:00
Enrico Ros 9acd178ce1 AudioPlayer: safe end of stream 2025-11-26 03:11:08 -08:00
Enrico Ros f381f80184 AIX: Anthropic: add strict to tool defs on wiretypes 2025-11-24 16:44:13 -08:00
Enrico Ros c83be61343 AIX: Anthropic: newlines for text broken by tool calls 2025-11-24 16:05:44 -08:00
Enrico Ros f6e49d31ec PWA-Desktop detect. Fixes #887 2025-11-24 15:48:50 -08:00
Enrico Ros cc0429a362 Update readme 2025-11-24 15:14:49 -08:00
282 changed files with 15275 additions and 4716 deletions
+1 -1
View File
@@ -46,4 +46,4 @@ Focus on discrepancies and gaps:
Report differences in wire types, adapter logic, parser handling, or dialect-specific quirks.
Prioritize new capabilities that improve user experience (reasoning visibility, better tool use, etc.).
When making changes, add comments with date: `// [OpenRouter, 2025-MM-DD]: explanation`
When making changes, add comments with date: `// [OpenRouter, 2026-MM-DD]: explanation`
+63
View File
@@ -0,0 +1,63 @@
---
description: Sync LLM parameter options between full model dialog and chat side panel
---
Audit and sync LLM parameter configurations between the two UI editors. Goal: identical `value` fields in option arrays + equivalent onChange logic. Labels/descriptions can differ for UI space.
**Files to Compare:**
1. **Full Model Dialog**: `src/modules/llms/models-modal/LLMParametersEditor.tsx` (main branch)
2. **Chat Side Panel**: `src/apps/chat/components/layout-panel/ChatPanelModelParameters.tsx` (main derived branches only)
**Reference Documentation:**
- Parameter system: `kb/systems/LLM-parameters-system.md`
- Parameter registry: `src/common/stores/llms/llms.parameters.ts`
**Task: Perform a comprehensive audit**
1. **Read both files** and extract all option arrays (e.g., `_reasoningEffortOptions`, `_antEffortOptions`, `_geminiThinkingLevelOptions`, etc.)
2. **Check for missing parameters:**
- Parameters handled in `LLMParametersEditor.tsx` but NOT in `ChatPanelModelParameters.tsx`
- Parameters in `ChatPanelModelParameters.tsx`'s `_interestingParameters` array but missing UI controls
- Note: The side panel intentionally shows only "interesting" parameters - focus on those listed in `_interestingParameters`
3. **Check for value mismatches** between corresponding option arrays:
- Different number of options (e.g., 3 vs 4 options)
- Same label but different `value` (this causes the bug in issue #926)
- Different labels for the same `value`
- Missing `_UNSPECIFIED`/Default option in one but not the other
4. **Check onChange handler consistency:**
- Both should remove parameter on `_UNSPECIFIED` selection
- Both should set explicit values the same way
- Watch for conditions like `value === 'high'` that may differ
**Output Format:**
```
## Parameter Sync Audit Report
### Missing Parameters
- [ ] `llmVndXyz` - In full dialog, missing from side panel
### Value Mismatches
- [ ] `_xyzOptions`:
- Full dialog: [values...]
- Side panel: [values...]
- Issue: [description]
### Handler Inconsistencies
- [ ] `llmVndXyz` onChange differs: [explanation]
### Recommended Fixes
1. [Specific fix with code snippet if needed]
```
**Fix Direction:** Full dialog is source of truth. Update side panel to match its values when mismatched.
**Notes:**
- Side panel uses shorter descriptions (space-constrained) - that's fine
- Variable names may differ (e.g., `_anthropicEffortOptions` vs `_antEffortOptions`) - that's fine, but same is better
- `value` fields must be identical sets
- `_UNSPECIFIED` must mean the same thing in both
- onChange: remove on `_UNSPECIFIED`, set explicit value otherwise
+3 -3
View File
@@ -6,11 +6,11 @@ Update `src/modules/llms/server/openai/models/groq.models.ts` with latest model
Reference `src/modules/llms/server/llm.server.types.ts` and `src/modules/llms/server/models.mappings.ts` for context only. Focus on the model file, do not descend into other code.
**Primary Sources:**
- Models: https://console.groq.com/docs/models
**Primary Source:**
- Fetch https://console.groq.com/docs/models.md directly (markdown format, no search needed)
- Pricing: https://groq.com/pricing/
**Fallbacks if blocked:** Search "groq models latest pricing", "groq latest models", "groq api models", or search GitHub for latest model prices and context windows
**Do NOT use web search.** The `.md` endpoint provides structured markdown content directly.
**Important:**
- Review the full model list for additions, removals, and price changes
+2 -2
View File
@@ -6,11 +6,11 @@ Update `src/modules/llms/server/openai/models/moonshot.models.ts` with latest mo
Reference `src/modules/llms/server/llm.server.types.ts` and `src/modules/llms/server/models.mappings.ts` for context only. Focus on the model file, do not descend into other code.
**Primary Sources:**
**Primary Sources (fetch directly, no search needed):**
- Pricing: https://platform.moonshot.ai/docs/pricing/chat
- API Reference: https://platform.moonshot.ai/docs/api/chat
**Fallbacks if blocked:** Search "moonshot kimi models latest pricing", "kimi k2 models", "moonshot api models", or search GitHub for latest model prices and context windows
**Do NOT use web search.** Fetch the URLs directly, or ask the user to provide data, if unaccessible.
**Important:**
- Review the full model list for additions, removals, and price changes
@@ -8,8 +8,8 @@ Reference `src/modules/llms/server/llm.server.types.ts` and `src/modules/llms/se
**Automated Workflow:**
```bash
# 1. Fetch the HTML
curl -s "https://ollama.com/library?sort=featured" -o /tmp/ollama-featured.html
# 1. Fetch the HTML (sorted by newest for stable ordering)
curl -s "https://ollama.com/library?sort=newest" -o /tmp/ollama-newest.html
# 2. Parse it with the script
node .claude/scripts/parse-ollama-models.js > /tmp/ollama-parsed.txt 2>&1
@@ -22,15 +22,18 @@ The parser outputs: `modelName|pulls|capabilities|sizes`
- Example: `deepseek-r1|66200000|tools,thinking|1.5b,7b,8b,14b,32b,70b,671b`
**Primary Sources:**
- Model Library: https://ollama.com/library?sort=featured
- Model Library: https://ollama.com/library?sort=newest
- Parser script: `.claude/scripts/parse-ollama-models.js`
**Fallbacks if blocked:** Check https://github.com/ollama/ollama, search "ollama featured models", "ollama latest models", or search GitHub for latest model info
**Important:**
- Skip models below 50,000 pulls (parser does this automatically)
- Skip embedding models (parser does not do this automatically)
- Sort them in the EXACT same order as the source (featured models)
- Parser filtering rules:
- Top 30 newest models are always included (regardless of pull count)
- After top 30, only models with 50K+ pulls are included
- Models with 'cloud' capability are automatically excluded
- Models with 'embedding' capability are automatically excluded
- Sort them in the EXACT same order as the source (newest first, for stable ordering)
- Extract tags: 'tools' → hasTools, 'vision' → hasVision, 'embedding' → isEmbeddings (note the 's'), 'thinking' → tags only
- Extract 'b' tags (1.5b, 7b, 32b) to tags field
- Set today's date (YYYYMMDD format) for newly added models only
+56
View File
@@ -0,0 +1,56 @@
---
description: Generate changelog bullets for big-agi.com/changes
argument-hint: date like "2026-01-10" or empty for auto-detect
---
Generate changelog bullets for a single entry in https://big-agi.com/changes
**Step 1: Find the starting date**
IMPORTANT: This repo rebases frequently, so commits are INTERLEAVED throughout history.
New commits can appear at line 10, 500, or 1800. Use AUTHOR DATE (`%ad`) to filter - it's preserved during rebases.
If `$ARGUMENTS` provided, use it as the cutoff date.
If NO argument:
1. Fetch https://big-agi.com/changes to get the most recent changelog date
2. Use that date as the cutoff
**Step 2: Get commits by author date**
Filter commits by author date to catch ALL new commits regardless of position in history:
```bash
# For commits after Jan 10, 2026 (adjust date pattern as needed)
git log --oneline --no-merges --format="%h %ad %s" --date=short | grep "2026-01-1[1-9]\|2026-01-2\|2026-02"
# Verify interleaving by checking line numbers
git log --oneline --no-merges --format="%h %ad %s" --date=short | grep -n "2026-01-1[1-9]"
```
The line numbers prove commits are scattered (e.g., lines 14, 638, 1156, 1803 = interleaved).
**Step 3: Write bullets**
Real examples from big-agi.com/changes:
- "Gemini 3 Flash support with 4-level thinking: high, medium, low, minimal"
- "Cloud Sync launched! - long awaited and top requested"
- "Deepseek V3.2 Speciale comes with almost Gemini 3 Pro performance but 20 times cheaper"
- "Anthropic Opus 4.5 with controls for effort (speed tradeoff), thinking budget, search"
- "Login with email, via magic link"
- "Mobile UX fixes for popups drag/interaction"
**Rules:**
1. **Order by importance** - most significant changes first, minor fixes last
2. **Feature-first, no verb prefixes** - "Gemini 3 support" not "Add Gemini 3 support"
3. **Model names lead** when it's about LLMs
4. **Specific details** - "4-level thinking: high, medium, low, minimal" not "multiple thinking levels"
5. **One-liners** - short, no fluff
6. **Consolidate commits** - 10 persona editor commits = 1 bullet
7. **No corporate speak** - no "enhanced", "streamlined", "robust", "leverage"
**Skip:** WIP, internal refactors, KB docs, automation, review cleanups, trivial fixes, deps bumps, CI changes.
**Output:** Just bullets, ready to paste. 2-5 bullets but adapt depending on scope, especially
in relation to the usual https://big-agi.com/changes entries.
+41 -9
View File
@@ -1,23 +1,36 @@
#!/usr/bin/env node
/**
* Parse Ollama featured models from HTML
* Parse Ollama models from HTML (sorted by newest for stable ordering)
*
* Usage:
* 1. Fetch HTML: curl -s "https://ollama.com/library?sort=featured" -o /tmp/ollama-featured.html
* 1. Fetch HTML: curl -s "https://ollama.com/library?sort=newest" -o /tmp/ollama-newest.html
* 2. Parse: node .claude/scripts/parse-ollama-models.js
*
* Outputs: pipe-delimited format: modelName|pulls|capabilities|sizes
* Example: deepseek-r1|66200000|tools,thinking|1.5b,7b,8b,14b,32b,70b,671b
*
* Filtering rules:
* - Top 30 newest models are always included (regardless of pull count)
* - After top 30, only models with 50K+ pulls are included
* - Models with 'cloud' capability are always excluded
* - Models with 'embedding' capability are always excluded
*
* Pull counts are rounded to significant figures for stable diffs:
* - >=10M: round to 100K (e.g., 109,123,456 -> 109,100,000)
* - >=1M: round to 10K (e.g., 5,432,100 -> 5,430,000)
* - <1M: round to 1K (e.g., 88,700 -> 89,000)
*/
const fs = require('fs');
const htmlPath = process.argv[2] || '/tmp/ollama-featured.html';
const htmlPath = process.argv[2] || '/tmp/ollama-newest.html';
const TOP_N_ALWAYS_INCLUDE = 30;
const MIN_PULLS_THRESHOLD = 50000;
if (!fs.existsSync(htmlPath)) {
console.error(`Error: HTML file not found at ${htmlPath}`);
console.error('Please fetch it first with:');
console.error(' curl -s "https://ollama.com/library?sort=featured" -o /tmp/ollama-featured.html');
console.error(' curl -s "https://ollama.com/library?sort=newest" -o /tmp/ollama-newest.html');
process.exit(1);
}
@@ -25,7 +38,7 @@ const html = fs.readFileSync(htmlPath, 'utf8');
// Split into model sections - each starts with <a href="/library/
const modelSections = html.split(/<a href="\/library\//);
const models = [];
const allParsedModels = [];
for (let i = 1; i < modelSections.length; i++) {
const section = modelSections[i].substring(0, 5000); // Large enough window to capture all data
@@ -65,10 +78,27 @@ for (let i = 1; i < modelSections.length; i++) {
sizes.push(sizeMatch[1].trim());
}
// Only include models with 50K+ pulls
if (pulls >= 50000) {
models.push({ name, pulls, capabilities, sizes });
// Skip models with 'cloud' or 'embedding' capability
if (capabilities.includes('cloud') || capabilities.includes('embedding')) {
continue;
}
allParsedModels.push({ name, pulls: roundPulls(pulls), capabilities, sizes });
}
// Apply filtering: top 30 always included, rest need 50K+ pulls
const models = allParsedModels.filter((model, index) => {
return index < TOP_N_ALWAYS_INCLUDE || model.pulls >= MIN_PULLS_THRESHOLD;
});
/**
* Round pulls to significant figures for stable output.
* This reduces churn from daily fluctuations while preserving magnitude.
*/
function roundPulls(pulls) {
if (pulls >= 10000000) return Math.round(pulls / 100000) * 100000; // >=10M: round to 100K
if (pulls >= 1000000) return Math.round(pulls / 10000) * 10000; // >=1M: round to 10K
return Math.round(pulls / 1000) * 1000; // <1M: round to 1K
}
// Output in pipe-delimited format (in the order they appear on the page)
@@ -78,4 +108,6 @@ models.forEach(m => {
console.log(`${m.name}|${m.pulls}|${caps}|${tags}`);
});
console.error(`\nTotal models with 50K+ pulls: ${models.length}`);
const topNCount = Math.min(TOP_N_ALWAYS_INCLUDE, allParsedModels.length);
const thresholdCount = models.length - topNCount;
console.error(`\nTotal models: ${models.length} (top ${topNCount} newest + ${thresholdCount} with ${MIN_PULLS_THRESHOLD / 1000}K+ pulls)`);
+5 -1
View File
@@ -5,11 +5,14 @@
"Bash(cp:*)",
"Bash(curl:*)",
"Bash(find:*)",
"Bash(gh issue list:*)",
"Bash(gh issue view:*)",
"Bash(git branch:*)",
"Bash(git cherry-pick:*)",
"Bash(git describe:*)",
"Bash(git grep:*)",
"Bash(git log:*)",
"Bash(git log:*)",
"Bash(git ls-tree:*)",
"Bash(git show:*)",
"Bash(grep:*)",
"Bash(ls:*)",
@@ -25,6 +28,7 @@
"Bash(sed:*)",
"Bash(tree:*)",
"Read(//tmp/**)",
"Skill(llms:update-models*)",
"WebFetch",
"WebFetch(domain:big-agi.com)",
"WebSearch",
+15 -40
View File
@@ -1,43 +1,18 @@
# big-AGI non-code files
/docs/
/dist/
README.md
*
# Ignore build and log files
Dockerfile
/.dockerignore
!app/
!kb/
!pages/
!public/
!src/
!tools/
# Node build artifacts
/node_modules
/.pnp
.pnp.js
!*.mjs
!middleware_BASIC_AUTH.ts
!middleware.ts
!next.config.ts
!package*.json
!tsconfig.json
# next.js
/.next/
/out/
# production
/build
# versioning
.git/
.github/
# IDEs
.idea/
# debug
npm-debug.log*
yarn-debug.log*
yarn-error.log*
.pnpm-debug.log*
# local env files
.env*.local
# vercel
.vercel
# typescript
*.tsbuildinfo
next-env.d.ts
!LICENSE
!README.md
+69
View File
@@ -0,0 +1,69 @@
version: 2
updates:
- package-ecosystem: docker
directory: /
schedule:
interval: weekly
commit-message:
prefix: "chore(deps)"
ignore:
- dependency-name: "node"
versions: [">=25", "<26"] # Node 25 breaks the build because of a dummy localStorage object
- package-ecosystem: github-actions
directory: /
schedule:
interval: weekly
commit-message:
prefix: "chore(deps)"
# Disabled npm updates for now - will need precise package pinning, as some packages changed behavior upstream
# - package-ecosystem: npm
# directory: /
# schedule:
# interval: weekly
# commit-message:
# prefix: "chore(deps)"
# cooldown:
# semver-patch: 3
# semver-minor: 7
# semver-major: 14
# # Ignore packages intentionally pinned due to upstream issues
# ignore:
# # Issue #857: v11.6+ breaks streaming; tried 11.4.4/11.6/11.7, only 11.5.1 works
# - dependency-name: "@trpc/*"
# versions: [">=11.5.1", "<12"]
# # Pinned during tRPC #857 debugging - may be safe to unpin, test first
# - dependency-name: "@tanstack/react-query"
# versions: [">=5.90.10", "<6"]
# # Pinned because 5.0.8 changes signatures so return set({ .. }) != void;
# - dependency-name: "zustand"
# versions: [">=5.0.7", "<6"]
# groups:
# next:
# patterns:
# - "@next/*"
# - "eslint-config-next"
# - "next"
# react:
# patterns:
# - "react"
# - "react-dom"
# - "@types/react"
# - "@types/react-dom"
# emotion:
# patterns:
# - "@emotion/*"
# mui:
# patterns:
# - "@mui/*"
# dnd-kit:
# patterns:
# - "@dnd-kit/*"
# prisma:
# patterns:
# - "@prisma/*"
# - "prisma"
# vercel:
# patterns:
# - "@vercel/*"
+4 -3
View File
@@ -30,7 +30,7 @@ jobs:
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v6
with:
fetch-depth: 1
@@ -51,7 +51,8 @@ jobs:
# or https://docs.claude.com/en/docs/claude-code/cli-reference for available options
# claude_args: '--allowed-tools Bash(gh pr:*)'
# disabling opus for now claude-opus-4-1-20250805
# former: claude-sonnet-4-5-20250929
claude_args: |
--model claude-sonnet-4-5-20250929
--model claude-opus-4-5-20251101
--max-turns 100
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(npm install),Bash(npm install:*),Bash(npm run:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools,SlashCommand"
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(npm run:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools,SlashCommand"
+3 -3
View File
@@ -23,7 +23,7 @@ jobs:
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v6
with:
fetch-depth: 1
@@ -72,6 +72,6 @@ jobs:
# See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
# or https://docs.claude.com/en/docs/claude-code/cli-reference for available options
claude_args: |
--model claude-sonnet-4-5-20250929
--model claude-opus-4-5-20251101
--max-turns 75
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(npm install),Bash(npm install:*),Bash(npm run:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools,SlashCommand"
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(npm run:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools,SlashCommand"
+3 -3
View File
@@ -31,7 +31,7 @@ jobs:
steps:
- name: Checkout repository
uses: actions/checkout@v4
uses: actions/checkout@v6
with:
fetch-depth: 1
@@ -72,6 +72,6 @@ jobs:
# See https://github.com/anthropics/claude-code-action/blob/main/docs/usage.md
# or https://docs.claude.com/en/docs/claude-code/cli-reference for available options
claude_args: |
--model claude-sonnet-4-5-20250929
--model claude-opus-4-5-20251101
--max-turns 100
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(npm install),Bash(npm install:*),Bash(npm run:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools"
--allowedTools "Edit,Read,Write,WebFetch,WebSearch,Bash(cat:*),Bash(cp:*),Bash(find:*),Bash(git branch:*),Bash(grep:*),Bash(ls:*),Bash(mkdir:*),Bash(gh issue:*),Bash(gh search:*),Bash(gh label:*),Bash(gh pr:*),mcp__chrome-devtools"
+120 -37
View File
@@ -20,29 +20,122 @@ env:
REGISTRY: ghcr.io
IMAGE_NAME: ${{ github.repository }}
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
jobs:
build-and-push-image:
runs-on: ubuntu-latest
timeout-minutes: 60 # Max 1 hour (expected: ~25min)
# Build job: runs on native runners for each platform (no QEMU emulation)
build:
strategy:
fail-fast: false
matrix:
include:
- platform: linux/amd64
runner: ubuntu-latest
- platform: linux/arm64
runner: ubuntu-24.04-arm
runs-on: ${{ matrix.runner }}
name: Build ${{ matrix.platform }}
timeout-minutes: 30
permissions:
contents: read
packages: write
security-events: write
steps:
- name: Checkout repository
uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Prepare
run: |
platform=${{ matrix.platform }}
echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
echo "IMAGE_NAME_LC=${IMAGE_NAME,,}" >> $GITHUB_ENV
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
- name: Checkout repository
uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 # v6.0.1
with:
fetch-depth: 1
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3.12.0
- name: Log in to the Container registry
uses: docker/login-action@v3
uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9 # v3.7.0
with:
registry: ${{ env.REGISTRY }}
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Extract metadata for Docker
id: meta
uses: docker/metadata-action@c299e40c65443455700f0fdfc63efafe5b349051 # v5.10.0
with:
images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
labels: |
org.opencontainers.image.title=Big-AGI Open
org.opencontainers.image.description=Big-AGI Open - Multi-model AI workspace for experts who need to think broader, decide smarter, and build with confidence.
org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
org.opencontainers.image.documentation=https://big-agi.com
- name: Build and push by digest
id: build
uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83 # v6.18.0
with:
context: .
file: Dockerfile
platforms: ${{ matrix.platform }}
labels: ${{ steps.meta.outputs.labels }}
tags: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME_LC }}
build-args: |
NEXT_PUBLIC_GA4_MEASUREMENT_ID=${{ secrets.GA4_MEASUREMENT_ID }}
NEXT_PUBLIC_BUILD_HASH=${{ github.sha }}
NEXT_PUBLIC_BUILD_REF_NAME=${{ github.ref_name }}
outputs: type=image,push-by-digest=true,name-canonical=true,push=true,oci-mediatypes=true
provenance: false
cache-from: type=gha,scope=${{ github.repository }}-${{ matrix.platform }}
cache-to: type=gha,scope=${{ github.repository }}-${{ matrix.platform }},mode=max
- name: Export digest
run: |
mkdir -p ${{ runner.temp }}/digests
digest="${{ steps.build.outputs.digest }}"
touch "${{ runner.temp }}/digests/${digest#sha256:}"
- name: Upload digest
uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6.0.0
with:
name: digests-${{ env.PLATFORM_PAIR }}
path: ${{ runner.temp }}/digests/*
if-no-files-found: error
retention-days: 1
# Merge job: combines platform-specific images into a unified multi-arch manifest
merge:
name: Merge manifests
runs-on: ubuntu-latest
timeout-minutes: 10
needs: build
permissions:
contents: read
packages: write
steps:
- name: Prepare
run: echo "IMAGE_NAME_LC=${IMAGE_NAME,,}" >> $GITHUB_ENV
- name: Download digests
uses: actions/download-artifact@37930b1c2abaa49bbe596cd826c3c89aef350131 # v7.0.0
with:
path: ${{ runner.temp }}/digests
pattern: digests-*
merge-multiple: true
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@8d2750c68a42422c14e847fe6c8ac0403b4cbd6f # v3.12.0
- name: Log in to the Container registry
uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9 # v3.7.0
with:
registry: ${{ env.REGISTRY }}
username: ${{ github.actor }}
@@ -50,7 +143,7 @@ jobs:
- name: Extract metadata (tags, labels) for Docker
id: meta
uses: docker/metadata-action@v5
uses: docker/metadata-action@c299e40c65443455700f0fdfc63efafe5b349051 # v5.10.0
with:
images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
tags: |
@@ -66,28 +159,18 @@ jobs:
# Version tags (v2.0.0, 2.0.0)
type=ref,event=tag
type=semver,pattern={{version}}
labels: |
org.opencontainers.image.title=Big-AGI Open
org.opencontainers.image.description=Big-AGI Open - Multi-model AI workspace for experts who need to think broader, decide smarter, and build with confidence.
org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
org.opencontainers.image.documentation=https://big-agi.com
- name: Build and push Docker image
uses: docker/build-push-action@v6
with:
context: .
file: Dockerfile
platforms: linux/amd64,linux/arm64
push: true
tags: ${{ steps.meta.outputs.tags }}
labels: ${{ steps.meta.outputs.labels }}
build-args: |
NEXT_PUBLIC_GA4_MEASUREMENT_ID=${{ secrets.GA4_MEASUREMENT_ID }}
NEXT_PUBLIC_BUILD_HASH=${{ github.sha }}
NEXT_PUBLIC_BUILD_REF_NAME=${{ github.ref_name }}
# Enable build cache (future)
#cache-from: type=gha
#cache-to: type=gha,mode=max
# Enable provenance and SBOM (future)
#provenance: true
#sbom: true
- name: Create manifest list and push
working-directory: ${{ runner.temp }}/digests
run: |
docker buildx imagetools create \
$(jq -cr '.tags | map("-t " + .) | join(" ")' <<< "$DOCKER_METADATA_OUTPUT_JSON") \
--annotation='index:org.opencontainers.image.title=Big-AGI Open' \
--annotation='index:org.opencontainers.image.description=Big-AGI Open - Multi-model AI workspace for experts who need to think broader, decide smarter, and build with confidence.' \
--annotation='index:org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}' \
--annotation='index:org.opencontainers.image.documentation=https://big-agi.com' \
$(printf '${{ env.REGISTRY }}/${{ env.IMAGE_NAME_LC }}@sha256:%s ' *)
- name: Inspect image
run: |
docker buildx imagetools inspect ${{ env.REGISTRY }}/${{ env.IMAGE_NAME_LC }}:${{ steps.meta.outputs.version }}
+3
View File
@@ -53,3 +53,6 @@ next-env.d.ts
.env*.local
/.run/dev (ENV).run.xml
/src/modules/3rdparty/aider/scratch*
# Ignore temporary CC files
/tmpclaude*
+1 -1
View File
@@ -228,7 +228,7 @@ The server uses a split architecture with two tRPC routers:
Distributed edge runtime for low-latency AI operations:
- **AIX** - AI streaming and communication
- **LLM Routers** - Direct vendor integrations (OpenAI, Anthropic, Gemini, Ollama)
- **External Services** - ElevenLabs (TTS), Google Search, YouTube transcripts
- **External Services** - ElevenLabs (TTS), Inworld (TTS), Google Search, YouTube transcripts
Located at `/src/server/trpc/trpc.router-edge.ts`
+19 -10
View File
@@ -1,5 +1,8 @@
# syntax=docker/dockerfile:1
# check=skip=CopyIgnoredFile
# Base
FROM node:22-alpine AS base
FROM node:24-alpine AS base
ENV NEXT_TELEMETRY_DISABLED=1
# Dependencies
@@ -39,19 +42,20 @@ ENV NEXT_PUBLIC_GA4_MEASUREMENT_ID=${NEXT_PUBLIC_GA4_MEASUREMENT_ID}
ARG NEXT_PUBLIC_POSTHOG_KEY
ENV NEXT_PUBLIC_POSTHOG_KEY=${NEXT_PUBLIC_POSTHOG_KEY}
# Optional argument to configure Google Drive Picker at build time (can reuse AUTH_GOOGLE_ID value)
ARG NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID
ENV NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID=${NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID}
# Copy development deps and source
COPY --from=deps /app/node_modules ./node_modules
COPY . .
# link ssl3 for latest Alpine
RUN sh -c '[ ! -e /lib/libssl.so.3 ] && ln -s /usr/lib/libssl.so.3 /lib/libssl.so.3 || echo "Link already exists"'
# Build the application
ENV NODE_ENV=production
RUN npm run build
# Reduce installed packages to production-only
RUN npm prune --production
RUN npm prune --omit=dev
# Runner
@@ -59,18 +63,23 @@ FROM base AS runner
WORKDIR /app
# As user
RUN addgroup --system --gid 1001 nodejs
RUN adduser --system --uid 1001 nextjs
RUN addgroup --system --gid 1001 nodejs \
&& adduser --system --uid 1001 nextjs \
&& apk add --no-cache openssl
# Copy Built app
COPY --from=builder --chown=nextjs:nodejs /app/public ./public
COPY --from=builder --chown=nextjs:nodejs /app/.next ./.next
COPY --from=builder --chown=nextjs:nodejs /app/node_modules ./node_modules
COPY --from=builder --chown=nextjs:nodejs /app/src/server/prisma ./src/server/prisma
# Instead of `COPY --from=builder --chown=nextjs:nodejs /app/.next ./.next`, we only extract some parts, excluding .next/cache which is build time only:
COPY --from=builder --chown=nextjs:nodejs /app/.next/BUILD_ID ./.next/
COPY --from=builder --chown=nextjs:nodejs /app/.next/static ./.next/static
COPY --from=builder --chown=nextjs:nodejs /app/.next/server ./.next/server
COPY --from=builder --chown=nextjs:nodejs /app/.next/types ./.next/types
COPY --from=builder --chown=nextjs:nodejs /app/.next/*.json ./.next/
# Minimal ENV for production
ENV NODE_ENV=production
ENV PATH=$PATH:/app/node_modules/.bin
# Run as non-root user
USER nextjs
@@ -79,4 +88,4 @@ USER nextjs
EXPOSE 3000
# Start the application
CMD ["next", "start"]
CMD ["/app/node_modules/.bin/next", "start"]
+1 -1
View File
@@ -1,6 +1,6 @@
MIT License
Copyright (c) 2023-2025 Enrico Ros
Copyright (c) 2023-2026 Enrico Ros
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
+13 -6
View File
@@ -42,7 +42,8 @@ It comes packed with **world-class features** like Beam, and is praised for its
[![Feature Inspector](https://img.shields.io/badge/Expert_Mode-AI_Inspector-000?style=for-the-badge&labelColor=purple)](https://big-agi.com/inspector)
### What makes Big-AGI different:
**Intelligence**: with [Beam & Merge](https://big-agi.com/beam) for multi-model de-hallucination, native search, and bleeding-edge AI models like Nano Banana, Kimi K2 Thinking or GPT 5.1 -
**Intelligence**: with [Beam & Merge](https://big-agi.com/beam) for multi-model de-hallucination, native search, and bleeding-edge AI models like Opus 4.5, Nano Banana Pro, Kimi K2.5 or GPT 5.2 -
**Control**: with personas, data ownership, requests inspection, unlimited usage with API keys, and *no vendor lock-in* -
and **Speed**: with a local-first, over-powered, zero-latency, madly optimized web app.
@@ -138,9 +139,15 @@ so you **are not vendor locked-in**, and obsessed over a powerful UI that works,
NOTE: this is a powerful tool - if you need a toy UI or clone, this ain't it.
## What's New in 2.0 · Oct 31, 2025 · Open
---
👉 **[See the full changelog](https://big-agi.com/changes)**
## Release Notes
👉 **[See the Live Release Notes](https://big-agi.com/changes)**
- Open 2.0.3: **Red Carpet** **Kimi K2.5**, **Gemini 3 Flash**, **GPT 5.2**, Google Drive, Inworld, Novita.ai, Speech/UX improvements
- Open 2.0.2: **Speex** multi-vendor speech synthesis, **Opus 4.5**, **Gemini 3 Pro**, **Nano Banana Pro**, **Grok 4.1**, **GPT-5.1**, **Kimi K2** + 280 fixes
### What's New in 2.0 · Oct 31, 2025 · Open
- **Big-AGI Open** is ready and more productive and faster than ever, with:
- **Beam 2**: multi-modal, program-based, follow-ups, save presets
@@ -153,7 +160,7 @@ NOTE: this is a powerful tool - if you need a toy UI or clone, this ain't it.
<img width="830" height="385" alt="image" src="https://github.com/user-attachments/assets/ad52761d-7e3f-44d8-b41e-947ce8b4faa1" />
### Open links: 👉 [changelog](https://big-agi.com/changes) 👉 [installation](docs/installation.md) 👉 [roadmap](https://github.com/users/enricoros/projects/4/views/2) 👉 [documentation](docs/README.md)
#### **Open** links: 👉 [changelog](https://big-agi.com/changes) 👉 [installation](docs/installation.md) 👉 [roadmap](https://github.com/users/enricoros/projects/4/views/2) 👉 [documentation](docs/README.md)
**For teams and institutions:** Need shared prompts, SSO, or managed deployments? Reach out at enrico@big-agi.com. We're actively collecting requirements from research groups and IT departments.
@@ -326,7 +333,7 @@ Configure 100s of AI models from 18+ providers:
| Multimodal services | [Azure](https://azure.microsoft.com/en-us/products/ai-services/openai-service) · [Anthropic](https://anthropic.com) · [Google Gemini](https://ai.google.dev/) · [OpenAI](https://platform.openai.com/docs/overview) |
| LLM services | [Alibaba](https://www.alibabacloud.com/en/product/modelstudio) · [DeepSeek](https://deepseek.com) · [Groq](https://wow.groq.com/) · [Mistral](https://mistral.ai/) · [Moonshot](https://www.moonshot.cn/) · [OpenPipe](https://openpipe.ai/) · [OpenRouter](https://openrouter.ai/) · [Perplexity](https://www.perplexity.ai/) · [Together AI](https://www.together.ai/) · [xAI](https://x.ai/) |
| Image services | OpenAI · Google Gemini |
| Speech services | [ElevenLabs](https://elevenlabs.io) (Voice synthesis / cloning) |
| Speech services | [ElevenLabs](https://elevenlabs.io) · [Inworld](https://inworld.ai) · [OpenAI TTS](https://platform.openai.com/docs/guides/text-to-speech) · LocalAI · Browser (Web Speech API) |
### Additional Integrations
@@ -382,4 +389,4 @@ When you open an issue, our custom AI triage system (powered by [Claude Code](ht
MIT License · [Third-Party Notices](src/modules/3rdparty/THIRD_PARTY_NOTICES.md)
**2023-2025** · Enrico Ros × [Big-AGI](https://big-agi.com)
**2023-2026** · Enrico Ros × [Big-AGI](https://big-agi.com)
+1 -1
View File
@@ -33,7 +33,7 @@ const handlerNodeRoutes = (req: Request) => fetchRequestHandler({
// NOTE: the following statement breaks the build on non-pro deployments, and conditionals don't work either
// so we resorted to raising the timeout from 10s to 60s in the vercel.json file instead
export const maxDuration = 60;
// export const maxDuration = 60;
export const runtime = 'nodejs';
export const dynamic = 'force-dynamic';
export { handlerNodeRoutes as GET, handlerNodeRoutes as POST };
+2
View File
@@ -14,5 +14,7 @@ const handlerEdgeRoutes = (req: Request) => fetchRequestHandler({
: undefined,
});
// NOTE: we don't set maxDuration explicitly here - however we set it in the Vercel project settings, raising to the limit of 300s
// export const maxDuration = 60;
export const runtime = 'edge';
export { handlerEdgeRoutes as GET, handlerEdgeRoutes as POST };
+1 -3
View File
@@ -2,8 +2,6 @@
#
# For more examples, such running big-AGI alongside a web browsing service, see the `docs/docker` folder.
version: '3.9'
services:
big-agi:
image: ghcr.io/enricoros/big-agi:latest
@@ -11,4 +9,4 @@ services:
- "3000:3000"
env_file:
- .env
command: [ "next", "start", "-p", "3000" ]
command: [ "next", "start", "-p", "3000" ]
+1 -1
View File
@@ -43,7 +43,7 @@ How to set up AI models and features in big-AGI.
- **[Web Browsing](config-feature-browse.md)**: Enable web page download through third-party services or your own cloud
- **Web Search**: Google Search API (see '[Environment Variables](environment-variables.md)')
- **Image Generation**: GPT Image (gpt-image-1), DALL·E 3 and 2
- **Voice Synthesis**: ElevenLabs API for voice generation
- **Voice Synthesis**: ElevenLabs, Inworld, OpenAI TTS, LocalAI, or browser Web Speech API
## Deployment & Customization
+55
View File
@@ -0,0 +1,55 @@
# Google Drive Integration
Attach files from Google Drive directly in the chat composer.
## Setup
### 1. Enable APIs
In [Google Cloud Console](https://console.cloud.google.com/):
1. Go to **APIs & Services > Library**
2. Enable **Google Drive API** and **Google Picker API**
### 2. Configure OAuth
1. Go to **APIs & Services > OAuth consent screen**
2. Create consent screen (External or Internal)
3. Add scope: `https://www.googleapis.com/auth/drive.file`
4. Add test users if in testing mode
### 3. Create Credentials
1. Go to **APIs & Services > Credentials**
2. Create **OAuth client ID** (Web application)
3. Add JavaScript origins:
- `http://localhost:3000` (dev)
- `https://your-domain.com` (prod)
### 4. Set Environment Variable
```bash
NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID=your-client-id.apps.googleusercontent.com
```
## Usage
- Click **Drive** button in attachment menu
## Supported Files
| Type | Export Format |
|-----------------|---------------------|
| Regular files | Downloaded directly |
| Google Docs | Markdown (.md) |
| Google Sheets | CSV (.csv) |
| Google Slides | PDF (.pdf) |
| Google Drawings | SVG (.svg) |
## Troubleshooting
**Picker won't open**: Check `NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID` is set and APIs are enabled.
**OAuth errors**: Verify your domain is in authorized JavaScript origins. Add yourself as test user if app is in testing mode.
**Download fails**: Check file permissions and that Drive API is enabled.
+6 -3
View File
@@ -66,8 +66,9 @@ HTTP_BASIC_AUTH_PASSWORD=
# Frontend variables
NEXT_PUBLIC_MOTD=
NEXT_PUBLIC_GA4_MEASUREMENT_ID=
NEXT_PUBLIC_POSTHOG_KEY=
NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID=
NEXT_PUBLIC_PLANTUML_SERVER_URL=
NEXT_PUBLIC_POSTHOG_KEY=
```
## Backend Variables
@@ -132,10 +133,11 @@ Enable the app to Talk, Draw, and Google things up.
| Variable | Description |
|:---------------------------|:------------------------------------------------------------------------------------------------------------------------|
| **Text-To-Speech** | [ElevenLabs](https://elevenlabs.io/) is a high quality speech synthesis service |
| **Text-To-Speech** | ElevenLabs, Inworld, OpenAI TTS, LocalAI, and browser Web Speech API are supported |
| `ELEVENLABS_API_KEY` | ElevenLabs API Key - used for calls, etc. |
| `ELEVENLABS_API_HOST` | Custom host for ElevenLabs |
| `ELEVENLABS_VOICE_ID` | Default voice ID for ElevenLabs |
| | *Note: OpenAI TTS and LocalAI TTS reuse credentials from your configured LLM services (no separate env vars needed)* |
| **Google Custom Search** | [Google Programmable Search Engine](https://programmablesearchengine.google.com/about/) produces links to pages |
| `GOOGLE_CLOUD_API_KEY` | Google Cloud API Key, used with the '/react' command - [Link to GCP](https://console.cloud.google.com/apis/credentials) |
| `GOOGLE_CSE_ID` | Google Custom/Programmable Search Engine ID - [Link to PSE](https://programmablesearchengine.google.com/) |
@@ -154,8 +156,9 @@ The value of these variables are passed to the frontend (Web UI) - make sure the
| `NEXT_PUBLIC_DEBUG_BREAKS` | (optional, development) When set to 'true', enables automatic debugger breaks on DEV/error/critical logs in development builds |
| `NEXT_PUBLIC_MOTD` | Message of the Day - displays a dismissible banner at the top of the app (see [customizations](customizations.md) for the template variables). Example: 🔔 Welcome to our deployment! Version {{app_build_pkgver}} built on {{app_build_time}}. |
| `NEXT_PUBLIC_GA4_MEASUREMENT_ID` | (optional) The measurement ID for Google Analytics 4. (see [deploy-analytics](deploy-analytics.md)) |
| `NEXT_PUBLIC_POSTHOG_KEY` | (optional) Key for PostHog analytics. (see [deploy-analytics](deploy-analytics.md)) |
| `NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID` | (optional) Google OAuth Client ID for Drive Picker. Can reuse `AUTH_GOOGLE_ID`. See [Google Drive](config-feature-google-drive.md) |
| `NEXT_PUBLIC_PLANTUML_SERVER_URL` | The URL of the PlantUML server, used for rendering UML diagrams. Allows using custom local servers. |
| `NEXT_PUBLIC_POSTHOG_KEY` | (optional) Key for PostHog analytics. (see [deploy-analytics](deploy-analytics.md)) |
> Important: these variables must be set at build time, which is required by Next.js to pass them to the frontend.
> This is in contrast to the backend variables, which can be set when starting the local server/container.
-5
View File
@@ -136,11 +136,6 @@ Deploy big-AGI on a Kubernetes cluster for enhanced scalability and management.
For more detailed instructions on Kubernetes deployment, including updating and troubleshooting, refer to our [Kubernetes Deployment Guide](deploy-k8s.md).
### Midori AI Subsystem for Docker Deployment
Follow the instructions found on [Midori AI Subsystem Site](https://io.midori-ai.xyz/subsystem/manager/)
for your host OS. After completing the setup process, install the Big-AGI docker backend to the Midori AI Subsystem.
## Enterprise-Grade Installation
For businesses seeking a fully-managed, scalable solution, consider our managed installations.
+3
View File
@@ -14,6 +14,9 @@ Internal documentation for Big-AGI architecture and systems, for use by AI agent
- **[AIX.md](modules/AIX.md)** - AIX streaming architecture documentation
- **[AIX-callers-analysis.md](modules/AIX-callers-analysis.md)** - Analysis of AIX entry points, call chains, common and different rendering, error handling, etc.
#### CSF - Client-Side Fetch
- **[CSF.md](systems/client-side-fetch.md)** - Direct browser-to-API communication for LLM requests
### Systems Documentation
#### Core Platform Systems
+1 -1
View File
@@ -105,7 +105,7 @@ When a model is loaded:
The system maintains type safety through:
- `DModelParameterId` union from registry keys
- `DModelParameterValue<T>` conditional types for values
- `DModelParameterSpec<T>` interfaces for specifications
- `DModelParameterSpecAny` interfaces for specifications
- Runtime validation via Zod schemas at API boundaries
## Model Variant Pattern
+13
View File
@@ -0,0 +1,13 @@
# CSF - Client-Side Fetch
Client-Side Fetch (CSF) enables direct browser-to-API communication, bypassing the server for LLM requests. When enabled, the browser makes requests directly to vendor APIs (e.g., `api.openai.com`, `api.groq.com`) instead of routing through the Next.js server. This reduces latency, decreases server load, and is particularly useful for local models where the browser can communicate directly with Ollama or LM Studio.
## Implementation
CSF is implemented as an opt-in setting stored as `csf: boolean` in each vendor's service settings. The vendor interface exposes `csfAvailable?: (setup) => boolean` to determine if CSF can be enabled (typically checking if an API key or host is configured). The actual execution happens in `aix.client.direct-chatGenerate.ts` which dynamically imports when CSF is active, making direct fetch calls using the same wire protocols as the server.
All 16 supported vendors (OpenAI, Anthropic, Gemini, Ollama, LocalAI, Deepseek, Groq, Mistral, xAI, OpenRouter, Perplexity, Together AI, Alibaba, Moonshot, OpenPipe, LM Studio) support CSF. Cloud vendors require CORS support from the API provider (all tested vendors return `access-control-allow-origin: *`). Local vendors (Ollama, LocalAI, LM Studio) require CORS to be enabled on the local server.
## UI
The CSF toggle appears in each vendor's setup panel under "Advanced" settings, labeled "Direct Connection". It becomes visible when the prerequisites are met (API key present for cloud vendors, host configured for local vendors). The setting is managed through `useModelServiceClientSideFetch` hook which provides `csfAvailable`, `csfActive`, `csfToggle`, and `csfReset` for UI consumption.
+1 -1
View File
@@ -141,7 +141,7 @@ if (process.env.POSTHOG_API_KEY && process.env.POSTHOG_ENV_ID) {
personalApiKey: process.env.POSTHOG_API_KEY,
envId: process.env.POSTHOG_ENV_ID,
host: 'https://us.i.posthog.com', // backtrace upload host
logLevel: 'info',
logLevel: 'error', // lowered, too noisy
sourcemaps: {
enabled: process.env.NODE_ENV === 'production',
project: 'big-agi',
+1150 -985
View File
File diff suppressed because it is too large Load Diff
+22 -21
View File
@@ -1,6 +1,6 @@
{
"name": "big-agi",
"version": "2.0.1",
"version": "2.0.3",
"private": true,
"author": "Enrico Ros <enrico.ros@gmail.com>",
"repository": "https://github.com/enricoros/big-agi",
@@ -29,38 +29,39 @@
"@emotion/react": "^11.14.0",
"@emotion/server": "^11.11.0",
"@emotion/styled": "^11.14.1",
"@googleworkspace/drive-picker-react": "^0.2.0",
"@mui/icons-material": "^5.18.0",
"@mui/joy": "^5.0.0-beta.52",
"@next/bundle-analyzer": "~15.1.8",
"@next/bundle-analyzer": "~15.1.12",
"@prisma/client": "~5.22.0",
"@tanstack/react-query": "5.90.10",
"@tanstack/react-virtual": "^3.13.12",
"@tanstack/react-virtual": "^3.13.18",
"@trpc/client": "11.5.1",
"@trpc/next": "11.5.1",
"@trpc/react-query": "11.5.1",
"@trpc/server": "11.5.1",
"@vercel/analytics": "^1.5.0",
"@vercel/speed-insights": "^1.2.0",
"@vercel/analytics": "^1.6.1",
"@vercel/speed-insights": "^1.3.1",
"browser-fs-access": "^0.38.0",
"cheerio": "^1.1.2",
"csv-stringify": "^6.6.0",
"dexie": "~4.0.11",
"dexie-react-hooks": "~1.1.7",
"diff": "^8.0.2",
"eventemitter3": "^5.0.1",
"diff": "^8.0.3",
"eventemitter3": "^5.0.4",
"idb-keyval": "^6.2.2",
"mammoth": "^1.11.0",
"nanoid": "^5.1.6",
"next": "~15.1.8",
"next": "~15.1.12",
"nprogress": "^0.2.0",
"pdfjs-dist": "5.4.54",
"posthog-js": "^1.298.0",
"posthog-node": "^5.14.0",
"posthog-js": "^1.336.4",
"posthog-node": "^5.24.7",
"prismjs": "^1.30.0",
"puppeteer-core": "^24.31.0",
"puppeteer-core": "^24.36.1",
"react": "^18.3.1",
"react-dom": "^18.3.1",
"react-hook-form": "^7.66.1",
"react-hook-form": "^7.71.1",
"react-markdown": "^10.1.0",
"react-player": "^3.4.0",
"react-resizable-panels": "^3.0.6",
@@ -70,26 +71,26 @@
"remark-mark-highlight": "^0.1.1",
"remark-math": "^6.0.0",
"sharp": "^0.34.5",
"superjson": "^2.2.5",
"tesseract.js": "^6.0.1",
"superjson": "^2.2.6",
"tesseract.js": "^7.0.0",
"tiktoken": "^1.0.22",
"turndown": "^7.2.2",
"zod": "^4.1.13",
"zod": "^4.3.6",
"zustand": "5.0.7"
},
"devDependencies": {
"@posthog/nextjs-config": "^1.6.0",
"@types/node": "^24.10.1",
"@posthog/nextjs-config": "~1.6.4",
"@types/node": "^25.1.0",
"@types/nprogress": "^0.2.3",
"@types/prismjs": "^1.26.5",
"@types/react": "^19.2.7",
"@types/react": "^19.2.10",
"@types/react-csv": "^1.1.10",
"@types/react-dom": "^19.2.3",
"@types/turndown": "^5.0.6",
"cross-env": "^10.1.0",
"eslint": "^9.39.1",
"eslint-config-next": "~15.1.8",
"prettier": "^3.6.2",
"eslint": "^9.39.2",
"eslint-config-next": "~15.1.12",
"prettier": "^3.8.1",
"prisma": "~5.22.0",
"typescript": "^5.9.3"
},
+1 -2
View File
@@ -18,7 +18,7 @@ import { ROUTE_APP_CHAT, ROUTE_INDEX } from '~/common/app.routes';
import { Release } from '~/common/app.release';
// capabilities access
import { useCapabilityBrowserSpeechRecognition, useCapabilityElevenLabs, useCapabilityTextToImage } from '~/common/components/useCapabilities';
import { useCapabilityBrowserSpeechRecognition, useCapabilityTextToImage } from '~/common/components/useCapabilities';
// stores access
import { getLLMsDebugInfo } from '~/common/stores/llms/store-llms';
@@ -95,7 +95,6 @@ function AppDebug() {
const cProduct = {
capabilities: {
mic: useCapabilityBrowserSpeechRecognition(),
elevenLabs: useCapabilityElevenLabs(),
textToImage: useCapabilityTextToImage(),
},
models: getLLMsDebugInfo(),
+16 -13
View File
@@ -6,13 +6,15 @@ import ChatIcon from '@mui/icons-material/Chat';
import CheckRoundedIcon from '@mui/icons-material/CheckRounded';
import CloseRoundedIcon from '@mui/icons-material/CloseRounded';
import MicIcon from '@mui/icons-material/Mic';
import RecordVoiceOverTwoToneIcon from '@mui/icons-material/RecordVoiceOverTwoTone';
import WarningRoundedIcon from '@mui/icons-material/WarningRounded';
import { useSpeexGlobalEngine } from '~/modules/speex/store-module-speex';
import { PhVoice } from '~/common/components/icons/phosphor/PhVoice';
import { animationColorRainbow } from '~/common/util/animUtils';
import { navigateBack } from '~/common/app.routes';
import { optimaOpenPreferences } from '~/common/layout/optima/useOptima';
import { useCapabilityBrowserSpeechRecognition, useCapabilityElevenLabs } from '~/common/components/useCapabilities';
import { useCapabilityBrowserSpeechRecognition } from '~/common/components/useCapabilities';
import { useChatStore } from '~/common/stores/chat/store-chats';
import { useUICounter } from '~/common/stores/store-ui';
@@ -45,7 +47,7 @@ export function CallWizard(props: { strict?: boolean, conversationId: string | n
// external state
const recognition = useCapabilityBrowserSpeechRecognition();
const synthesis = useCapabilityElevenLabs();
const speexGlobalEngine = useSpeexGlobalEngine();
const chatIsEmpty = useChatStore(state => {
if (!props.conversationId)
return false;
@@ -56,17 +58,18 @@ export function CallWizard(props: { strict?: boolean, conversationId: string | n
// derived state
const outOfTheBlue = !props.conversationId;
const overriddenEmptyChat = chatEmptyOverride || !chatIsEmpty;
const overriddenEmptyChat = outOfTheBlue || chatEmptyOverride || !chatIsEmpty;
const overriddenRecognition = recognitionOverride || recognition.mayWork;
const allGood = overriddenEmptyChat && overriddenRecognition && synthesis.mayWork;
const fatalGood = overriddenRecognition && synthesis.mayWork;
const synthesisShallWork = !!speexGlobalEngine;
const allGood = overriddenEmptyChat && overriddenRecognition && synthesisShallWork;
const fatalGood = overriddenRecognition && synthesisShallWork;
const handleOverrideChatEmpty = React.useCallback(() => setChatEmptyOverride(true), []);
const handleOverrideRecognition = React.useCallback(() => setRecognitionOverride(true), []);
const handleConfigureElevenLabs = React.useCallback(() => optimaOpenPreferences('voice'), []);
const handleConfigureVoice = React.useCallback(() => optimaOpenPreferences('voice'), []);
const handleFinishButton = React.useCallback(() => {
if (!allGood)
@@ -128,17 +131,17 @@ export function CallWizard(props: { strict?: boolean, conversationId: string | n
{/* Text to Speech status */}
<StatusCard
icon={<RecordVoiceOverTwoToneIcon />}
icon={<PhVoice />}
text={
(synthesis.mayWork ? 'Voice synthesis should be ready.' : 'There might be an issue with ElevenLabs voice synthesis.')
+ (synthesis.isConfiguredServerSide ? '' : (synthesis.isConfiguredClientSide ? '' : ' Please add your API key in the settings.'))
(synthesisShallWork ? 'Voice synthesis should be ready.' : 'There might be an issue with voice synthesis.')
// + (synthesis.isConfiguredServerSide ? '' : (synthesis.isConfiguredClientSide ? '' : ' Please add your API key in the settings.'))
}
button={synthesis.mayWork ? undefined : (
<Button variant='outlined' onClick={handleConfigureElevenLabs} sx={{ mx: 1 }}>
button={synthesisShallWork ? undefined : (
<Button variant='outlined' onClick={handleConfigureVoice} sx={{ mx: 1 }}>
Configure
</Button>
)}
hasIssue={!synthesis.mayWork}
hasIssue={!synthesisShallWork}
/>
{/*<Typography>*/}
+1 -1
View File
@@ -317,7 +317,7 @@ export function Contacts(props: { setCallIntent: (intent: AppCallIntent) => void
issue={354}
text='Call App: Support thread and compatibility matrix'
note={<>
Voice input uses the HTML Web Speech API, and speech output requires an ElevenLabs API Key.
Voice input uses the HTML Web Speech API.
</>}
// note2='Please report any issues you encounter'
sx={{
+19 -31
View File
@@ -7,23 +7,24 @@ import CallEndIcon from '@mui/icons-material/CallEnd';
import CallIcon from '@mui/icons-material/Call';
import MicIcon from '@mui/icons-material/Mic';
import MicNoneIcon from '@mui/icons-material/MicNone';
import RecordVoiceOverTwoToneIcon from '@mui/icons-material/RecordVoiceOverTwoTone';
import { ScrollToBottom } from '~/common/scroll-to-bottom/ScrollToBottom';
import { ScrollToBottomButton } from '~/common/scroll-to-bottom/ScrollToBottomButton';
import { useChatLLMDropdown } from '../chat/components/layout-bar/useLLMDropdown';
import { SystemPurposeId, SystemPurposes } from '../../data';
import { elevenLabsSpeakText } from '~/modules/elevenlabs/elevenlabs.client';
import { AixChatGenerateContent_DMessageGuts, aixChatGenerateContent_DMessage_FromConversation } from '~/modules/aix/client/aix.client';
import { useElevenLabsVoiceDropdown } from '~/modules/elevenlabs/useElevenLabsVoiceDropdown';
import { aixChatGenerateContent_DMessage_FromConversation, AixChatGenerateContent_DMessageGuts } from '~/modules/aix/client/aix.client';
import { speakText } from '~/modules/speex/speex.client';
import type { OptimaBarControlMethods } from '~/common/layout/optima/bar/OptimaBarDropdown';
import { AudioPlayer } from '~/common/util/audio/AudioPlayer';
import { Link } from '~/common/components/Link';
import { OptimaPanelGroupedList } from '~/common/layout/optima/panel/OptimaPanelGroupedList';
import { OptimaPanelIn, OptimaToolbarIn } from '~/common/layout/optima/portals/OptimaPortalsIn';
import { PhVoice } from '~/common/components/icons/phosphor/PhVoice';
import { SpeechResult, useSpeechRecognition } from '~/common/components/speechrecognition/useSpeechRecognition';
import { clipboardInterceptCtrlCForCleanup } from '~/common/util/clipboardUtils';
import { conversationTitle, remapMessagesSysToUsr } from '~/common/stores/chat/chat.conversation';
import { createDMessageFromFragments, createDMessageTextContent, DMessage, messageFragmentsReduceText, messageWasInterruptedAtStart } from '~/common/stores/chat/chat.message';
import { createErrorContentFragment } from '~/common/stores/chat/chat.fragments';
@@ -43,18 +44,13 @@ import { useAppCallStore } from './state/store-app-call';
function CallMenu(props: {
pushToTalk: boolean,
setPushToTalk: (pushToTalk: boolean) => void,
override: boolean,
setOverride: (overridePersonaVoice: boolean) => void,
}) {
// external state
const { grayUI, toggleGrayUI } = useAppCallStore();
const { voicesDropdown } = useElevenLabsVoiceDropdown(false, !props.override);
const handlePushToTalkToggle = () => props.setPushToTalk(!props.pushToTalk);
const handleChangeVoiceToggle = () => props.setOverride(!props.override);
return <OptimaPanelGroupedList title='Call'>
<MenuItem onClick={handlePushToTalkToggle}>
@@ -63,17 +59,6 @@ function CallMenu(props: {
<Switch checked={props.pushToTalk} onChange={handlePushToTalkToggle} sx={{ ml: 'auto' }} />
</MenuItem>
<MenuItem onClick={handleChangeVoiceToggle}>
<ListItemDecorator><RecordVoiceOverTwoToneIcon /></ListItemDecorator>
Change Voice
<Switch checked={props.override} onChange={handleChangeVoiceToggle} sx={{ ml: 'auto' }} />
</MenuItem>
<MenuItem>
<ListItemDecorator>{' '}</ListItemDecorator>
{voicesDropdown}
</MenuItem>
<ListDivider />
<MenuItem onClick={toggleGrayUI}>
@@ -98,7 +83,6 @@ export function Telephone(props: {
const [avatarClickCount, setAvatarClickCount] = React.useState<number>(0);// const [micMuted, setMicMuted] = React.useState(false);
const [callElapsedTime, setCallElapsedTime] = React.useState<string>('00:00');
const [callMessages, setCallMessages] = React.useState<DMessage[]>([]);
const [overridePersonaVoice, setOverridePersonaVoice] = React.useState<boolean>(false);
const [personaTextInterim, setPersonaTextInterim] = React.useState<string | null>(null);
const [pushToTalk, setPushToTalk] = React.useState(true);
const [stage, setStage] = React.useState<'ring' | 'declined' | 'connected' | 'ended'>('ring');
@@ -118,7 +102,7 @@ export function Telephone(props: {
}));
const persona = SystemPurposes[props.callIntent.personaId as SystemPurposeId] ?? undefined;
const personaCallStarters = persona?.call?.starters ?? undefined;
const personaVoiceId = overridePersonaVoice ? undefined : (persona?.voices?.elevenLabs?.voiceId ?? undefined);
// const personaVoiceSelector = React.useMemo(() => personaGetVoiceSelector(persona), [persona]);
const personaSystemMessage = persona?.systemMessage ?? undefined;
// hooks and speech
@@ -165,7 +149,6 @@ export function Telephone(props: {
};
// [E] pickup -> seed message and call timer
// FIXME: Overriding the voice will reset the call - not a desired behavior
React.useEffect(() => {
if (!isConnected) return;
@@ -185,11 +168,14 @@ export function Telephone(props: {
setCallMessages([createDMessageTextContent('assistant', firstMessage)]); // [state] set assistant:hello message
// fire/forget
void elevenLabsSpeakText(firstMessage, personaVoiceId, true, true);
// fire/forget - use 'fast' priority for real-time conversation
void speakText(firstMessage,
undefined,
{ label: 'Call', priority: 'fast' },
);
return () => clearInterval(interval);
}, [isConnected, personaCallStarters, personaVoiceId]);
}, [isConnected, personaCallStarters]);
// [E] persona streaming response - upon new user message
React.useEffect(() => {
@@ -270,9 +256,12 @@ export function Telephone(props: {
fullMessage.generator = status.lastDMessage.generator;
setCallMessages(messages => [...messages, fullMessage]); // [state] append assistant:call_response
// fire/forget
// fire/forget - use 'fast' priority for real-time conversation
if (status.outcome === 'success' && finalText?.length >= 1)
void elevenLabsSpeakText(finalText, personaVoiceId, true, true);
void speakText(finalText,
undefined,
{ label: 'Call', priority: 'fast' },
);
}).catch((err: DOMException) => {
if (err?.name !== 'AbortError') {
@@ -288,7 +277,7 @@ export function Telephone(props: {
responseAbortController.current?.abort();
responseAbortController.current = null;
};
}, [isConnected, callMessages, modelId, personaVoiceId, personaSystemMessage, reMessages]);
}, [callMessages, isConnected, modelId, personaSystemMessage, reMessages]);
// [E] Message interrupter
const abortTrigger = isConnected && recognitionState.hasSpeech;
@@ -325,7 +314,6 @@ export function Telephone(props: {
<OptimaPanelIn>
<CallMenu
pushToTalk={pushToTalk} setPushToTalk={setPushToTalk}
override={overridePersonaVoice} setOverride={setOverridePersonaVoice}
/>
</OptimaPanelIn>
@@ -373,7 +361,7 @@ export function Telephone(props: {
<ScrollToBottom stickToBottomInitial>
<Box sx={{ minHeight: '100%', p: 1, display: 'flex', flexDirection: 'column', gap: 1 }}>
<Box onCopy={clipboardInterceptCtrlCForCleanup} sx={{ minHeight: '100%', p: 1, display: 'flex', flexDirection: 'column', gap: 1 }}>
{/* Call Messages [] */}
{callMessages.map((message) =>
-7
View File
@@ -10,7 +10,6 @@ import type { DiagramConfig } from '~/modules/aifn/digrams/DiagramsModal';
import type { TradeConfig } from '~/modules/trade/TradeModal';
import { downloadSingleChat, importConversationsFromFilesAtRest, openConversationsAtRestPicker } from '~/modules/trade/trade.client';
import { imaginePromptFromTextOrThrow } from '~/modules/aifn/imagine/imaginePromptFromText';
import { elevenLabsSpeakText } from '~/modules/elevenlabs/elevenlabs.client';
import { useAreBeamsOpen } from '~/modules/beam/store-beam.hooks';
import { useCapabilityTextToImage } from '~/modules/t2i/t2i.client';
@@ -346,11 +345,6 @@ export function AppChat() {
});
}, [handleExecuteAndOutcome]);
const handleTextSpeak = React.useCallback(async (text: string): Promise<void> => {
await elevenLabsSpeakText(text, undefined, true, true);
}, []);
// Chat actions
const handleConversationNewInFocusedPane = React.useCallback((forceNoRecycle: boolean, isIncognito: boolean) => {
@@ -725,7 +719,6 @@ export function AppChat() {
onConversationNew={handleConversationNewInFocusedPane}
onTextDiagram={handleTextDiagram}
onTextImagine={handleImagineFromText}
onTextSpeak={handleTextSpeak}
sx={chatMessageListSx}
/>
)}
+12 -10
View File
@@ -7,17 +7,18 @@ import { Box, List } from '@mui/joy';
import type { SystemPurposeExample } from '../../../data';
import type { DiagramConfig } from '~/modules/aifn/digrams/DiagramsModal';
import { speakText } from '~/modules/speex/speex.client';
import type { ConversationHandler } from '~/common/chat-overlay/ConversationHandler';
import type { DLLMContextTokens } from '~/common/stores/llms/llms.types';
import { DConversationId, excludeSystemMessages } from '~/common/stores/chat/chat.conversation';
import { ShortcutKey, useGlobalShortcuts } from '~/common/components/shortcuts/useGlobalShortcuts';
import { clipboardInterceptCtrlCForCleanup } from '~/common/util/clipboardUtils';
import { convertFilesToDAttachmentFragments } from '~/common/attachment-drafts/attachment.pipeline';
import { createDMessageFromFragments, createDMessageTextContent, DMessage, DMessageId, DMessageUserFlag, DMetaReferenceItem, MESSAGE_FLAG_AIX_SKIP, messageHasUserFlag } from '~/common/stores/chat/chat.message';
import { createTextContentFragment, DMessageFragment, DMessageFragmentId } from '~/common/stores/chat/chat.fragments';
import { openFileForAttaching } from '~/common/components/ButtonAttachFiles';
import { optimaOpenPreferences } from '~/common/layout/optima/useOptima';
import { useCapabilityElevenLabs } from '~/common/components/useCapabilities';
import { useChatOverlayStore } from '~/common/chat-overlay/store-perchat_vanilla';
import { useChatStore } from '~/common/stores/chat/store-chats';
import { useScrollToBottom } from '~/common/scroll-to-bottom/useScrollToBottom';
@@ -50,7 +51,6 @@ export function ChatMessageList(props: {
onConversationNew: (forceNoRecycle: boolean, isIncognito: boolean) => void,
onTextDiagram: (diagramConfig: DiagramConfig | null) => void,
onTextImagine: (conversationId: DConversationId, selectedText: string) => Promise<void>,
onTextSpeak: (selectedText: string) => Promise<void>,
setIsMessageSelectionMode: (isMessageSelectionMode: boolean) => void,
sx?: SxProps,
}) {
@@ -75,10 +75,9 @@ export function ChatMessageList(props: {
_composerInReferenceToCount: state.inReferenceTo?.length ?? 0,
ephemerals: state.ephemerals?.length ? state.ephemerals : null,
})));
const { mayWork: isSpeakable } = useCapabilityElevenLabs();
// derived state
const { conversationHandler, conversationId, capabilityHasT2I, onConversationBranch, onConversationExecuteHistory, onTextDiagram, onTextImagine, onTextSpeak } = props;
const { conversationHandler, conversationId, capabilityHasT2I, onConversationBranch, onConversationExecuteHistory, onTextDiagram, onTextImagine } = props;
const composerCanAddInReferenceTo = _composerInReferenceToCount < 5;
const composerHasInReferenceto = _composerInReferenceToCount > 0;
@@ -212,12 +211,15 @@ export function ChatMessageList(props: {
}, [capabilityHasT2I, conversationId, onTextImagine]);
const handleTextSpeak = React.useCallback(async (text: string) => {
if (!isSpeakable)
return optimaOpenPreferences('voice');
// sandwich the speaking with the indicator
setIsSpeaking(true);
await onTextSpeak(text);
const result = await speakText(text, undefined, { label: 'Chat speak' });
setIsSpeaking(false);
}, [isSpeakable, onTextSpeak]);
// open voice preferences
if (!result.success && (result.errorType === 'tts-no-engine' || result.errorType === 'tts-unconfigured'))
optimaOpenPreferences('voice');
}, []);
// operate on the local selection set
@@ -322,7 +324,7 @@ export function ChatMessageList(props: {
);
return (
<List role='chat-messages-list' sx={listSx}>
<List role='chat-messages-list' sx={listSx} onCopy={clipboardInterceptCtrlCForCleanup}>
{props.isMessageSelectionMode && (
<MessagesSelectionHeader
@@ -377,7 +379,7 @@ export function ChatMessageList(props: {
onMessageTruncate={handleMessageTruncate}
onTextDiagram={handleTextDiagram}
onTextImagine={capabilityHasT2I ? handleTextImagine : undefined}
onTextSpeak={isSpeakable ? handleTextSpeak : undefined}
onTextSpeak={handleTextSpeak}
/>
);
@@ -220,7 +220,7 @@ export function CameraCaptureModal(props: {
backdropFilter: 'none', // using none because this is heavy
// backdropFilter: 'blur(4px)',
// backgroundColor: 'rgba(11 13 14 / 0.75)',
backgroundColor: 'rgba(var(--joy-palette-neutral-darkChannel) / 0.5)',
backgroundColor: 'rgba(var(--joy-palette-neutral-darkChannel) / 0.67)',
},
},
}}
+20 -2
View File
@@ -63,8 +63,10 @@ import { chatExecuteModeCanAttach, useChatExecuteMode } from '../../execute-mode
import { ButtonAttachCameraMemo, useCameraCaptureModalDialog } from './buttons/ButtonAttachCamera';
import { ButtonAttachClipboardMemo } from './buttons/ButtonAttachClipboard';
import { ButtonAttachGoogleDriveMemo } from './buttons/ButtonAttachGoogleDrive';
import { ButtonAttachScreenCaptureMemo } from './buttons/ButtonAttachScreenCapture';
import { ButtonAttachWebMemo } from './buttons/ButtonAttachWeb';
import { hasGoogleDriveCapability, useGoogleDrivePicker } from '~/common/attachment-drafts/useGoogleDrivePicker';
import { ButtonBeamMemo } from './buttons/ButtonBeam';
import { ButtonCallMemo } from './buttons/ButtonCall';
import { ButtonGroupDrawRepeat } from './buttons/ButtonGroupDrawRepeat';
@@ -197,7 +199,7 @@ export function Composer(props: {
const showChatAttachments = chatExecuteModeCanAttach(chatExecuteMode, props.capabilityHasT2IEdit);
const {
/* items */ attachmentDrafts,
/* append */ attachAppendClipboardItems, attachAppendDataTransfer, attachAppendEgoFragments, attachAppendFile, attachAppendUrl,
/* append */ attachAppendClipboardItems, attachAppendCloudFile, attachAppendDataTransfer, attachAppendEgoFragments, attachAppendFile, attachAppendUrl,
/* take */ attachmentsRemoveAll, attachmentsTakeAllFragments, attachmentsTakeFragmentsByType,
} = useAttachmentDrafts(conversationOverlayStore, enableLoadURLsInComposer, chatLLMSupportsImages, handleFilterAGIFile, showChatAttachments === 'only-images');
@@ -545,6 +547,9 @@ export function Composer(props: {
// Enter: primary action
if (e.key === 'Enter') {
// Skip if composing (e.g., CJK input methods) - issue #784
if (e.nativeEvent.isComposing)
return;
// Alt (Windows) or Option (Mac) + Enter: append the message instead of sending it
if (e.altKey && !e.metaKey && !e.ctrlKey) {
@@ -620,6 +625,8 @@ export function Composer(props: {
const { openWebInputDialog, webInputDialogComponent } = useWebInputModal(handleAttachWebLinks, composeText);
const { openGoogleDrivePicker, googleDrivePickerComponent } = useGoogleDrivePicker(attachAppendCloudFile, isMobile);
// Attachments Down
@@ -799,6 +806,11 @@ export function Composer(props: {
<ButtonAttachWebMemo disabled={!hasComposerBrowseCapability} onOpenWebInput={openWebInputDialog} />
</MenuItem>
{/* Responsive Google Drive button */}
{hasGoogleDriveCapability && <MenuItem>
<ButtonAttachGoogleDriveMemo onOpenGoogleDrivePicker={openGoogleDrivePicker} fullWidth />
</MenuItem>}
{/* Responsive Paste button */}
{supportsClipboardRead() && <MenuItem>
<ButtonAttachClipboardMemo onAttachClipboard={attachAppendClipboardItems} />
@@ -828,6 +840,9 @@ export function Composer(props: {
{/* Responsive Web button */}
{showChatAttachments !== 'only-images' && <ButtonAttachWebMemo color={showTint} disabled={!hasComposerBrowseCapability} onOpenWebInput={openWebInputDialog} />}
{/* Responsive Google Drive button */}
{hasGoogleDriveCapability && showChatAttachments !== 'only-images' && <ButtonAttachGoogleDriveMemo color={showTint} onOpenGoogleDrivePicker={openGoogleDrivePicker} />}
{/* Responsive Paste button */}
{supportsClipboardRead() && showChatAttachments !== 'only-images' && <ButtonAttachClipboardMemo color={showTint} onAttachClipboard={attachAppendClipboardItems} />}
@@ -905,7 +920,7 @@ export function Composer(props: {
)}
{!showChatInReferenceTo && !isDraw && tokenLimit > 0 && (
<TokenBadgeMemo hideBelowDollars={0.005} chatPricing={tokenChatPricing} direct={tokensComposer} history={tokensHistory} responseMax={tokensResponseMax} limit={tokenLimit} showCost={labsShowCost} enableHover={!isMobile} showExcess absoluteBottomRight />
<TokenBadgeMemo hideBelowDollars={0.01} chatPricing={tokenChatPricing} direct={tokensComposer} history={tokensHistory} responseMax={tokensResponseMax} limit={tokenLimit} showCost={labsShowCost} enableHover={!isMobile} showExcess absoluteBottomRight />
)}
</Box>
@@ -1123,6 +1138,9 @@ export function Composer(props: {
{/* Camera (when open) */}
{cameraCaptureComponent}
{/* Google Drive Picker (when open) */}
{googleDrivePickerComponent}
{/* Web Input Dialog (when open) */}
{webInputDialogComponent}
@@ -0,0 +1,49 @@
import * as React from 'react';
import { Box, Button, ColorPaletteProp, IconButton, Tooltip } from '@mui/joy';
import AddToDriveRoundedIcon from '@mui/icons-material/AddToDriveRounded';
import { buttonAttachSx } from '~/common/components/ButtonAttachFiles';
import { KeyStroke } from '~/common/components/KeyStroke';
export const ButtonAttachGoogleDriveMemo = React.memo(ButtonAttachGoogleDrive);
function ButtonAttachGoogleDrive(props: {
color?: ColorPaletteProp,
isMobile?: boolean,
disabled?: boolean,
fullWidth?: boolean,
noToolTip?: boolean,
onOpenGoogleDrivePicker: () => void,
}) {
const button = props.isMobile ? (
<IconButton color={props.color} disabled={props.disabled} onClick={props.onOpenGoogleDrivePicker}>
<AddToDriveRoundedIcon />
</IconButton>
) : (
<Button
variant={props.color ? 'soft' : 'plain'}
color={props.color || 'neutral'}
disabled={props.disabled}
fullWidth={props.fullWidth}
startDecorator={<AddToDriveRoundedIcon />}
onClick={props.onOpenGoogleDrivePicker}
sx={buttonAttachSx.desktop}
>
Drive
</Button>
);
return (props.noToolTip || props.isMobile) ? button : (
<Tooltip arrow disableInteractive placement='top-start' title={
<Box sx={buttonAttachSx.tooltip}>
<b>Add from Google Drive</b><br />
Attach files from your Drive
</Box>
}>
{button}
</Tooltip>
);
}
@@ -91,8 +91,11 @@ function InputErrorIndicator() {
const converterTypeToIconMap: { [key in AttachmentDraftConverterType]: React.ComponentType<any> | null } = {
'text': TextFieldsIcon,
'text-cleaner': CodeIcon,
'text-markdown': TextFieldsIcon,
'rich-text': CodeIcon,
'rich-text-cleaner': CodeIcon,
'rich-text-markdown': TextFieldsIcon,
'rich-text-table': PivotTableChartIcon,
'image-original': ImageOutlinedIcon,
'image-resized-high': PhotoSizeSelectLargeOutlinedIcon,
@@ -100,8 +103,10 @@ const converterTypeToIconMap: { [key in AttachmentDraftConverterType]: React.Com
'image-to-default': ImageOutlinedIcon,
'image-caption': AbcIcon,
'image-ocr': AbcIcon,
'pdf-auto': PictureAsPdfIcon,
'pdf-text': PictureAsPdfIcon,
'pdf-images': PermMediaOutlinedIcon,
'pdf-images-ocr': AbcIcon,
'pdf-text-and-images': PermMediaOutlinedIcon,
'docx-to-html': DescriptionOutlinedIcon,
'url-page-text': TextFieldsIcon, // was LanguageIcon
@@ -199,13 +204,21 @@ function attachmentIcons(attachmentDraft: AttachmentDraft, noTooltips: boolean,
function attachmentLabelText(attachmentDraft: AttachmentDraft): string {
const converter = attachmentDraft.converters.find(c => c.isActive) ?? null;
if (converter && attachmentDraft.label === 'Rich Text') {
if (converter.id === 'rich-text-table')
return 'Rich Table';
if (converter.id === 'rich-text-cleaner')
if (converter && attachmentDraft.label === 'Text') {
if (converter.id === 'text-markdown')
return 'Markdown';
if (converter.id === 'text-cleaner')
return 'Clean HTML';
}
if (converter && attachmentDraft.label === 'Rich Text') {
if (converter.id === 'rich-text')
return 'Rich HTML';
if (converter.id === 'rich-text-markdown')
return 'Markdown';
if (converter.id === 'rich-text-cleaner')
return 'Clean HTML';
if (converter.id === 'rich-text-table')
return 'Rich Table';
}
return ellipsizeFront(attachmentDraft.label, 22);
}
@@ -228,9 +241,10 @@ function LLMAttachmentButton(props: {
const isUnconvertible = !draft.converters.length;
const isOutputLoading = draft.outputsConverting;
const isOutputMissing = !draft.outputFragments.length;
const isOutputWarned = !!draft.outputWarnings?.length;
const hasLiveFiles = draft.outputFragments.some(_f => _f.liveFileId);
const showWarning = isUnconvertible || (isOutputMissing || !llmSupportsAllFragments);
const showWarning = isUnconvertible || (isOutputMissing || !llmSupportsAllFragments) || isOutputWarned;
// handlers
@@ -257,6 +271,17 @@ function LLMAttachmentButton(props: {
if (isInputLoading)
return <InputLoadingPlaceholder label={draft.label} />;
// tooltip for truncated filenames (only show when menu is closed)
const displayedLabel = attachmentLabelText(draft);
const showFilenameTooltip = !props.menuShown && !isOutputLoading && displayedLabel !== draft.label;
// label element (reused with/without tooltip)
const labelElement = (
<Typography level='title-sm' sx={{ whiteSpace: 'nowrap' }}>
{isOutputLoading ? 'Converting... ' : displayedLabel}
</Typography>
);
return (
<Button
size='sm'
@@ -280,10 +305,11 @@ function LLMAttachmentButton(props: {
{/* Icons: Web Page Screenshot, Converter[s] */}
{attachmentIcons(draft, props.menuShown, props.onViewImageRefPart)}
{/* Label */}
<Typography level='title-sm' sx={{ whiteSpace: 'nowrap' }}>
{isOutputLoading ? 'Converting... ' : attachmentLabelText(draft)}
</Typography>
{/* Label (with tooltip for truncated filenames) */}
{showFilenameTooltip
? <TooltipOutlined title={<span style={{ wordBreak: 'break-all' }}>{draft.label}</span>}>{labelElement}</TooltipOutlined>
: labelElement
}
{/* Is Converting icon */}
{isOutputLoading && <CircularProgress color='success' size='sm' />}
@@ -1,16 +1,15 @@
import * as React from 'react';
import type { SxProps } from '@mui/joy/styles/types';
import { Box, Checkbox, Chip, CircularProgress, LinearProgress, ListDivider, ListItem, ListItemDecorator, MenuItem, Radio, Typography } from '@mui/joy';
import AttachmentIcon from '@mui/icons-material/Attachment';
import { Box, Button, ButtonGroup, Checkbox, Chip, CircularProgress, Divider, LinearProgress, ListDivider, ListItem, ListItemDecorator, MenuItem, Radio, Typography } from '@mui/joy';
import ClearIcon from '@mui/icons-material/Clear';
import ContentCopyIcon from '@mui/icons-material/ContentCopy';
import DeleteForeverIcon from '@mui/icons-material/DeleteForever';
import DeleteOutlineIcon from '@mui/icons-material/DeleteOutline';
import ExpandLessIcon from '@mui/icons-material/ExpandLess';
import ExpandMoreIcon from '@mui/icons-material/ExpandMore';
import KeyboardArrowDownIcon from '@mui/icons-material/KeyboardArrowDown';
import KeyboardArrowLeftIcon from '@mui/icons-material/KeyboardArrowLeft';
import KeyboardArrowRightIcon from '@mui/icons-material/KeyboardArrowRight';
import ReadMoreIcon from '@mui/icons-material/ReadMore';
import VerticalAlignBottomIcon from '@mui/icons-material/VerticalAlignBottom';
import VisibilityIcon from '@mui/icons-material/Visibility';
@@ -18,6 +17,7 @@ import { CloseablePopup } from '~/common/components/CloseablePopup';
import { DMessageAttachmentFragment, DMessageDocPart, DMessageImageRefPart, isDocPart, isImageRefPart, isZyncAssetImageReferencePartWithLegacyDBlob } from '~/common/stores/chat/chat.fragments';
import { LiveFileIcon } from '~/common/livefile/liveFile.icons';
import { copyToClipboard } from '~/common/util/clipboardUtils';
import { humanReadableBytes } from '~/common/util/textUtils';
import { themeZIndexOverMobileDrawer } from '~/common/app.theme';
import { useUIPreferencesStore } from '~/common/stores/store-ui';
@@ -32,12 +32,20 @@ const DEFAULT_DETAILS_OPEN = true;
const SHOW_INLINING_OPERATIONS = false;
const indicatorSx = {
fontSize: '1rem',
} as const;
// const indicatorSx = {
// fontSize: '1rem',
// } as const;
//
// const indicatorGapSx: SxProps = {
// paddingLeft: '1.375rem',
// };
const indicatorGapSx: SxProps = {
paddingLeft: '1.375rem',
const actionButtonsSx: SxProps = {
ml: 'auto',
minHeight: 0,
borderRadius: '1rem',
backgroundColor: 'background.surface',
'& button': { fontSize: 'xs', fontWeight: 'md', py: 0, minWidth: 0, minHeight: 0 },
};
@@ -82,9 +90,10 @@ export function LLMAttachmentMenu(props: {
const isUnconvertible = !draft.converters.length;
const isOutputMissing = !draft.outputFragments.length;
const isOutputMultiple = draft.outputFragments.length > 1;
const isOutputWarned = !!draft.outputWarnings?.length;
const hasLiveFiles = draft.outputFragments.some(_f => _f.liveFileId);
const showWarning = isUnconvertible || isOutputMissing || !llmSupportsAllFragments;
const showWarning = isUnconvertible || isOutputMissing || !llmSupportsAllFragments || isOutputWarned;
// hooks
@@ -157,6 +166,8 @@ export function LLMAttachmentMenu(props: {
minWidth={260}
noTopPadding
placement='top'
placementOffset={[0, 15]}
boxShadow='lg'
zIndex={themeZIndexOverMobileDrawer /* was not set, but the Attachment Menu can be used from the Personas Modal */}
>
@@ -187,9 +198,10 @@ export function LLMAttachmentMenu(props: {
<ListItem sx={{ fontSize: 'sm', my: 0.75 }}>
Attach {draftSource.media === 'url' ? 'web page'
: draftSource.media === 'file' ? 'file'
: draftSource.media === 'text'
? (draftSource.method === 'drop' ? 'drop' : draftSource.method === 'clipboard-read' ? 'clipboard' : draftSource.method === 'paste' ? 'paste' : '')
: ''} as:
: draftSource.media === 'cloud' ? 'cloud file'
: draftSource.media === 'text'
? (draftSource.method === 'drop' ? 'drop' : draftSource.method === 'clipboard-read' ? 'clipboard' : draftSource.method === 'paste' ? 'paste' : '')
: ''} as:
{uiComplexityMode === 'extra' && (
<Chip component='span' size='sm' color='neutral' variant='outlined' startDecorator={<ContentCopyIcon />} onClick={(event) => handleCopyLabelToClipboard(event, draft.label)} sx={{ ml: 'auto' }}>
copy name
@@ -197,6 +209,17 @@ export function LLMAttachmentMenu(props: {
)}
</ListItem>
)}
{/* Auto-heuristics message, with explanation */}
{!!draft.outputsHeuristic?.isAuto && (
<ListItem color={draft.outputsHeuristic.isAuto ? 'primary' : undefined} sx={{ fontSize: 'sm', fontWeight: 'lg', mb: 0.5 }}>
{draft.outputsHeuristic.isAuto ? 'Auto: ' : ''}
{draft.outputsHeuristic.actualConverterId === 'pdf-text' && 'Text'}
{draft.outputsHeuristic.actualConverterId === 'pdf-images-ocr' && 'OCR'}
{draft.outputsHeuristic.actualConverterId === 'pdf-images' && 'Images'}
{draft.outputsHeuristic.actualConverterId === 'pdf-text-and-images' && 'Text + Images'}
{draft.outputsHeuristic.explain && ` (${draft.outputsHeuristic.explain})`}
</ListItem>
)}
{!isUnconvertible && draft.converters.map((c, idx) =>
<MenuItem
disabled={c.disabled || isConverting}
@@ -213,7 +236,9 @@ export function LLMAttachmentMenu(props: {
</ListItemDecorator>
{c.unsupported
? <Box>Unsupported 🤔 <Typography level='body-xs'>{c.name}</Typography></Box>
: c.name}
: (/* auto-converted */ draft.outputsHeuristic?.isAuto && c.id === draft.outputsHeuristic.actualConverterId)
? <Box component='span' sx={{ fontWeight: 'lg', color: 'primary.softColor' }}>{c.name}</Box>
: c.name}
</MenuItem>,
)}
{/*{!isUnconvertible && <ListDivider sx={{ mb: 0 }} />}*/}
@@ -261,11 +286,19 @@ export function LLMAttachmentMenu(props: {
<Typography color={isInputError ? 'danger' : 'warning'} level='title-sm'>
{isInputError ? 'Loading Issue' : 'Warning'}
</Typography>
{/* Only show 1 warning, excluding lower priorities */}
{isInputError ? <div>{draft.inputError}</div>
: isUnconvertible ? <div>Attachments of type {draft.input?.mimeType} are not supported yet. You can request this on GitHub.</div>
: isOutputMissing ? <div>File not supported. Please try another format.</div>
: !llmSupportsAllFragments ? <div>May not be compatible with the current model. Please try another format.</div>
: <>Unknown warning</>}
: draft.outputWarnings?.length ? '' /* printed below */
: <>Unknown warning</>}
{/* Explicit output warnings */}
{!!draft.outputWarnings?.length && draft.outputWarnings.map((w, widx) =>
<Box key={'ow-' + widx} sx={{ fontSize: 'sm', color: 'warning.softColor', py: 1 }}> {w}</Box>)
}
</Box>
</MenuItem>
</Box>
@@ -294,24 +327,24 @@ export function LLMAttachmentMenu(props: {
Details
</Typography>
) : (
<Box sx={{ my: 0.5 }}>
<Box sx={{ my: 1 }}>
{/* <- inputs */}
{showInputs && !!draftInput && (
<Typography level='body-sm' textColor='text.primary' startDecorator={<AttachmentIcon sx={indicatorSx} />}>
{draftInput.mimeType}{typeof draftInput.dataSize === 'number' ? ` · ${draftInput.dataSize.toLocaleString()} bytes` : ''}
<Typography level='body-sm' textColor='success.softColor'>
Input: {draftInput.mimeType}{typeof draftInput.dataSize === 'number' ? ` · ${humanReadableBytes(draftInput.dataSize)}` : ''}
</Typography>
)}
{showInputs && !!draftInput?.altMimeType && (
<Typography level='body-sm' sx={indicatorGapSx}>
{draftInput.altMimeType} · {draftInput.altData?.length.toLocaleString()}
<Typography level='body-sm' textColor='success.softColor'>
Input: {draftInput.altMimeType}{!draftInput.altData?.length ? '' : ` · ${humanReadableBytes(draftInput.altData.length)}`}
</Typography>
)}
{showInputs && !!draftInput?.urlImage && (
<Typography level='body-sm' sx={indicatorGapSx}>
{draftInput.urlImage.mimeType} · {draftInput.urlImage.width} x {draftInput.urlImage.height} · {draftInput.urlImage.imgDataUrl?.length.toLocaleString()}
{' · '}
<Chip component='span' size='sm' color='primary' variant='outlined' startDecorator={<VisibilityIcon />} onClick={(event) => {
<Typography level='body-sm' textColor='success.softColor' sx={{ display: 'flex', alignItems: 'center' }}>
Input: {draftInput.urlImage.mimeType} · {draftInput.urlImage.width}x{draftInput.urlImage.height}{!draftInput.urlImage.imgDataUrl?.length ? '' : ` · ${humanReadableBytes(draftInput.urlImage.imgDataUrl.length)}`}
&nbsp;
<Chip component='span' size='sm' color='success' variant='soft' startDecorator={<VisibilityIcon />} onClick={(event) => {
if (draftInput?.urlImage?.imgDataUrl) {
// Invoke the viewer but with a virtual 'temp' part description to see this preview image
handleViewImageRefPart(event, {
@@ -325,8 +358,8 @@ export function LLMAttachmentMenu(props: {
height: draftInput.urlImage.height || undefined,
});
}
}}>
view
}} sx={{ ml: 'auto' }}>
view input
</Chip>
</Typography>
)}
@@ -335,45 +368,79 @@ export function LLMAttachmentMenu(props: {
{/* Converters: {draft.converters.map(((converter, idx) => ` ${converter.id}${converter.isActive ? '*' : ''}`)).join(', ')}*/}
{/*</Typography>*/}
{/* Downward arrow */}
<Divider color='success'>
<KeyboardArrowDownIcon color='success' />
</Divider>
{/* -> Outputs */}
<Box sx={{ mt: 1 }}>
<Box>
{isOutputMissing ? (
<Typography level='body-sm' startDecorator={<ReadMoreIcon sx={indicatorSx} />}>...</Typography>
<Typography level='body-sm' color={isConverting ? 'primary' : 'danger'}>{isConverting ? '...' : '... nothing ...'}</Typography>
) : (
draft.outputFragments.map(({ part }, index) => {
if (isDocPart(part)) {
return (
<Typography key={index} level='body-sm' sx={{ color: 'text.primary' }} startDecorator={<ReadMoreIcon sx={indicatorSx} />}>
<span>{part.data.mimeType /* part.type: big-agi type, not source mime */} · {part.data.text.length.toLocaleString()} bytes ·&nbsp;</span>
<Chip component='span' size='sm' color='primary' variant='outlined' startDecorator={<VisibilityIcon />} onClick={(event) => handleViewDocPart(event, part)}>
view
</Chip>
<Chip component='span' size='sm' color='success' variant='outlined' startDecorator={<ContentCopyIcon />} onClick={(event) => handleCopyToClipboard(event, part.data.text)}>
copy
</Chip>
<Typography key={index} component='div' level='body-sm' textColor='primary.softColor' sx={{ display: 'flex', alignItems: 'center' }}>
<span>{part.data.mimeType /* part.type: big-agi type, not source mime */} · {humanReadableBytes(part.data.text.length)} &nbsp;</span>
{/*<Chip component='span' size='sm' color='primary' variant='outlined' startDecorator={<VisibilityIcon />} onClick={(event) => handleViewDocPart(event, part)} sx={{ ml: 'auto' }}>*/}
{/* view*/}
{/*</Chip>*/}
{/*<Chip component='span' size='sm' color='primary' variant='outlined' startDecorator={<ContentCopyIcon />} onClick={(event) => handleCopyToClipboard(event, part.data.text)}>*/}
{/* copy*/}
{/*</Chip>*/}
<ButtonGroup size='sm' color='primary' variant='outlined' sx={actionButtonsSx}>
<Button startDecorator={<VisibilityIcon sx={{ fontSize: 'md' }} />} onClick={(event) => handleViewDocPart(event, part)}>
view
</Button>
<Button onClick={(event) => handleCopyToClipboard(event, part.data.text)}/* endDecorator={<ContentCopyIcon />} */>
copy
</Button>
</ButtonGroup>
</Typography>
);
} else if (isZyncAssetImageReferencePartWithLegacyDBlob(part) || isImageRefPart(part)) {
// Unified Image Reference handling (both Zync Asset References with legacy fallback and legacy image_ref)
const legacyImageRefPart = isZyncAssetImageReferencePartWithLegacyDBlob(part) ? part._legacyImageRefPart! : part;
const { dataRef, width, height } = legacyImageRefPart;
const resolution = width && height ? `${width} x ${height}` : 'no resolution';
const resolution = width && height ? `${width}x${height}` : 'no resolution';
const mime = dataRef.reftype === 'dblob' ? dataRef.mimeType : 'unknown image';
return (
<Typography key={index} level='body-sm' sx={{ color: 'text.primary' }} startDecorator={<ReadMoreIcon sx={indicatorSx} />}>
<span>{mime /*.replace('image/', 'img: ')*/} · {resolution} · {dataRef.reftype === 'dblob' ? (dataRef.bytesSize?.toLocaleString() || 'no size') : '(remote)'} ·&nbsp;</span>
<Chip component='span' size={isOutputMultiple ? 'sm' : 'md'} color='primary' variant='outlined' startDecorator={<VisibilityIcon />}
onClick={(event) => handleViewImageRefPart(event, legacyImageRefPart)}>
view
</Chip>
{isOutputMultiple && <Chip component='span' size={isOutputMultiple ? 'sm' : 'md'} color='danger' variant='outlined' startDecorator={<DeleteForeverIcon />} onClick={(event) => handleDeleteOutputFragment(event, index)}>
del
</Chip>}
<Typography key={index} component='div' level='body-sm' textColor='primary.softColor' sx={{ display: 'flex', alignItems: 'center' }}>
<span>{mime /*.replace('image/', 'img: ')*/} · {resolution} · {
dataRef.reftype !== 'dblob' ? '(remote)'
: !dataRef.bytesSize ? 'no size'
: humanReadableBytes(dataRef.bytesSize)} &nbsp;</span>
{/*<Chip component='span' size={isOutputMultiple ? 'sm' : 'md'} color='primary' variant='outlined' startDecorator={<VisibilityIcon />}*/}
{/* onClick={(event) => handleViewImageRefPart(event, legacyImageRefPart)}>*/}
{/* view*/}
{/*</Chip>*/}
{/*{isOutputMultiple && <Chip component='span' size={isOutputMultiple ? 'sm' : 'md'} color='danger' variant='outlined' startDecorator={<DeleteForeverIcon />} onClick={(event) => handleDeleteOutputFragment(event, index)}>*/}
{/* del*/}
{/*</Chip>}*/}
<ButtonGroup size='sm' color='primary' variant='outlined' sx={actionButtonsSx}>
<Button
startDecorator={<VisibilityIcon sx={{ fontSize: 'md' }} />}
onClick={(event) => handleViewImageRefPart(event, legacyImageRefPart)}
>
view
</Button>
{isOutputMultiple && (
<Button
color='warning'
endDecorator={<DeleteOutlineIcon sx={{ fontSize: 'md' }} />}
onClick={(event) => handleDeleteOutputFragment(event, index)}
// sx={{ width: 48 }}
>
del
</Button>
)}
</ButtonGroup>
</Typography>
);
} else {
return (
<Typography key={index} level='body-sm' sx={{ color: 'text.primary' }} startDecorator={<ReadMoreIcon sx={indicatorSx} />}>
<Typography key={index} level='body-sm' textColor='primary.softColor'>
{(part as DMessageAttachmentFragment['part']).pt}: (other)
</Typography>
);
@@ -381,8 +448,8 @@ export function LLMAttachmentMenu(props: {
})
)}
{!!llmTokenCountApprox && (
<Typography level='body-xs' mt={0.5} sx={indicatorGapSx}>
~{llmTokenCountApprox.toLocaleString()} tokens
<Typography level='body-xs' mt={0.5} textColor='primary.softColor'>
&nbsp; ~ {llmTokenCountApprox.toLocaleString()} tokens
</Typography>
)}
</Box>
@@ -47,9 +47,9 @@ function TokenBadge(props: {
const showAltCosts = !!props.showCost && !!costMax && costMin !== undefined;
if (showAltCosts) {
// Note: switched to 'min cost (>= ...)' on mobile as well, to restore the former behavior, just uncomment the !props.enableHover (a proxy for isMobile)
badgeValue = (/*!props.enableHover ||*/ isHovering)
? '< ' + formatModelsCost(costMax)
: '> ' + formatModelsCost(costMin);
badgeValue =
// (/*!props.enableHover ||*/ isHovering) ? '< ' + formatModelsCost(costMax) :
'> ' + formatModelsCost(costMin);
} else {
// show the direct tokens, unless we exceed the limit and 'showExcess' is enabled
@@ -77,7 +77,7 @@ function TokenBadge(props: {
slotProps={{
root: {
sx: {
...((props.absoluteBottomRight) && { position: 'absolute', bottom: 8, right: 8 }),
...((props.absoluteBottomRight) && { position: 'absolute', bottom: 8, right: '1rem' }),
cursor: 'help',
...(shallInvisible && {
opacity: 0,
@@ -92,6 +92,13 @@ function TokenBadge(props: {
fontFamily: 'code',
fontSize: 'xs',
...((props.absoluteBottomRight || props.inline) && { position: 'static', transform: 'none' }),
// make it transparent over text
// backgroundColor: `rgb(var(--joy-palette-${color}-lightChannel) / 15%)`, // similar to success.50
background: 'transparent',
boxShadow: 'none', // outline
'&:hover': {
backgroundColor: `${color}.softHoverBg`,
},
},
},
}}
@@ -5,7 +5,7 @@ import { useModuleBeamStore } from '~/modules/beam/store-module-beam';
import type { DFolder } from '~/common/stores/folders/store-chat-folders';
import { DMessage, DMessageUserFlag, MESSAGE_FLAG_STARRED, messageFragmentsReduceText, messageHasUserFlag, messageUserFlagToEmoji } from '~/common/stores/chat/chat.message';
import { conversationTitle, DConversationId } from '~/common/stores/chat/chat.conversation';
import { getLocalMidnightInUTCTimestamp, getTimeBucketEn } from '~/common/util/timeUtils';
import { createTimeBucketClassifierEn } from '~/common/util/timeUtils';
import { isAttachmentFragment, isContentOrAttachmentFragment, isDocPart, isImageRefPart, isZyncAssetImageReferencePart } from '~/common/stores/chat/chat.fragments';
import { shallowEquals } from '~/common/util/hooks/useShallowObject';
import { useChatStore } from '~/common/stores/chat/store-chats';
@@ -235,14 +235,14 @@ export function useChatDrawerRenderItems(
break;
}
const midnightTime = getLocalMidnightInUTCTimestamp();
const getTimeBucket = createTimeBucketClassifierEn();
const grouped = chatNavItems.reduce((acc, item) => {
// derive the bucket name
let bucket: string;
switch (grouping) {
case 'date':
bucket = getTimeBucketEn(item.updatedAt || midnightTime, midnightTime);
bucket = getTimeBucket(item.updatedAt || Date.now());
break;
case 'persona':
bucket = item.systemPurposeId;
@@ -21,7 +21,6 @@ import InsertLinkIcon from '@mui/icons-material/InsertLink';
import MoreVertIcon from '@mui/icons-material/MoreVert';
import NotificationsActiveIcon from '@mui/icons-material/NotificationsActive';
import NotificationsOutlinedIcon from '@mui/icons-material/NotificationsOutlined';
import RecordVoiceOverOutlinedIcon from '@mui/icons-material/RecordVoiceOverOutlined';
import ReplayIcon from '@mui/icons-material/Replay';
import ReplyAllRoundedIcon from '@mui/icons-material/ReplyAllRounded';
import ReplyRoundedIcon from '@mui/icons-material/ReplyRounded';
@@ -40,11 +39,12 @@ import { CloseablePopup } from '~/common/components/CloseablePopup';
import { DMessage, DMessageId, DMessageUserFlag, DMetaReferenceItem, MESSAGE_FLAG_AIX_SKIP, MESSAGE_FLAG_NOTIFY_COMPLETE, MESSAGE_FLAG_STARRED, MESSAGE_FLAG_VND_ANT_CACHE_AUTO, MESSAGE_FLAG_VND_ANT_CACHE_USER, messageFragmentsReduceText, messageHasUserFlag } from '~/common/stores/chat/chat.message';
import { KeyStroke } from '~/common/components/KeyStroke';
import { MarkHighlightIcon } from '~/common/components/icons/MarkHighlightIcon';
import { PhVoice } from '~/common/components/icons/phosphor/PhVoice';
import { Release } from '~/common/app.release';
import { TooltipOutlined } from '~/common/components/TooltipOutlined';
import { adjustContentScaling, themeScalingMap, themeZIndexChatBubble } from '~/common/app.theme';
import { avatarIconSx, makeMessageAvatarIcon, messageBackground, useMessageAvatarLabel } from '~/common/util/dMessageUtils';
import { copyToClipboard } from '~/common/util/clipboardUtils';
import { clipboardCopyDOMSelectionOrFallback } from '~/common/util/clipboardUtils';
import { createTextContentFragment, DMessageFragment, DMessageFragmentId, updateFragmentWithEditedText } from '~/common/stores/chat/chat.fragments';
import { useFragmentBuckets } from '~/common/stores/chat/hooks/useFragmentBuckets';
import { useUIPreferencesStore } from '~/common/stores/store-ui';
@@ -69,7 +69,7 @@ const ENABLE_BUBBLE = true;
export const BUBBLE_MIN_TEXT_LENGTH = 3;
// Enable the hover button to copy the whole message. The Copy button is also available in Blocks, or in the Avatar Menu.
const ENABLE_COPY_MESSAGE_OVERLAY: boolean = false;
// const ENABLE_COPY_MESSAGE_OVERLAY: boolean = false;
const messageBodySx: SxProps = {
@@ -315,8 +315,8 @@ export function ChatMessage(props: {
const handleCloseOpsMenu = React.useCallback(() => setOpsMenuAnchor(null), []);
const handleOpsCopy = (e: React.MouseEvent) => {
copyToClipboard(textSubject, 'Text');
e.preventDefault();
clipboardCopyDOMSelectionOrFallback(blocksRendererRef.current, textSubject, 'Message');
handleCloseOpsMenu();
closeContextMenu();
closeBubble();
@@ -893,18 +893,18 @@ export function ChatMessage(props: {
{/* Overlay copy icon */}
{ENABLE_COPY_MESSAGE_OVERLAY && !fromSystem && !isEditingText && (
<Tooltip title={messagePendingIncomplete ? null : (fromAssistant ? 'Copy message' : 'Copy input')} variant='solid'>
<IconButton
variant='outlined' onClick={handleOpsCopy}
sx={{
position: 'absolute', ...(fromAssistant ? { right: { xs: 12, md: 28 } } : { left: { xs: 12, md: 28 } }), zIndex: 10,
opacity: 0, transition: 'opacity 0.16s cubic-bezier(.17,.84,.44,1)',
}}>
<ContentCopyIcon />
</IconButton>
</Tooltip>
)}
{/*{ENABLE_COPY_MESSAGE_OVERLAY && !fromSystem && !isEditingText && (*/}
{/* <Tooltip title={messagePendingIncomplete ? null : (fromAssistant ? 'Copy message' : 'Copy input')} variant='solid'>*/}
{/* <IconButton*/}
{/* variant='outlined' onClick={handleOpsCopy}*/}
{/* sx={{*/}
{/* position: 'absolute', ...(fromAssistant ? { right: { xs: 12, md: 28 } } : { left: { xs: 12, md: 28 } }), zIndex: 10,*/}
{/* opacity: 0, transition: 'opacity 0.16s cubic-bezier(.17,.84,.44,1)',*/}
{/* }}>*/}
{/* <ContentCopyIcon />*/}
{/* </IconButton>*/}
{/* </Tooltip>*/}
{/*)}*/}
{/* Message Operations Menu (3 dots) */}
@@ -1027,7 +1027,7 @@ export function ChatMessage(props: {
)}
{!!props.onTextSpeak && (
<MenuItem onClick={handleOpsSpeak} disabled={!couldSpeak || props.isSpeaking}>
<ListItemDecorator>{props.isSpeaking ? <CircularProgress size='sm' /> : <RecordVoiceOverOutlinedIcon />}</ListItemDecorator>
<ListItemDecorator>{props.isSpeaking ? <CircularProgress size='sm' /> : <PhVoice />}</ListItemDecorator>
Speak
</MenuItem>
)}
@@ -1155,7 +1155,7 @@ export function ChatMessage(props: {
</Tooltip>}
{!!props.onTextSpeak && <Tooltip disableInteractive arrow placement='top' title='Speak'>
<IconButton color='success' onClick={handleOpsSpeak} disabled={!couldSpeak || props.isSpeaking}>
{!props.isSpeaking ? <RecordVoiceOverOutlinedIcon /> : <CircularProgress sx={{ '--CircularProgress-size': '16px' }} />}
{!props.isSpeaking ? <PhVoice /> : <CircularProgress sx={{ '--CircularProgress-size': '16px' }} />}
</IconButton>
</Tooltip>}
{(!!props.onTextDiagram || !!props.onTextImagine || !!props.onTextSpeak) && <Divider />}
@@ -1195,7 +1195,7 @@ export function ChatMessage(props: {
Auto-Draw
</MenuItem>}
{!!props.onTextSpeak && <MenuItem onClick={handleOpsSpeak} disabled={!couldSpeak || props.isSpeaking}>
<ListItemDecorator>{props.isSpeaking ? <CircularProgress size='sm' /> : <RecordVoiceOverOutlinedIcon />}</ListItemDecorator>
<ListItemDecorator>{props.isSpeaking ? <CircularProgress size='sm' /> : <PhVoice />}</ListItemDecorator>
Speak
</MenuItem>}
</CloseablePopup>
@@ -7,13 +7,13 @@ import CodeIcon from '@mui/icons-material/Code';
import EditRoundedIcon from '@mui/icons-material/EditRounded';
import ImageOutlinedIcon from '@mui/icons-material/ImageOutlined';
import PictureAsPdfIcon from '@mui/icons-material/PictureAsPdf';
import RecordVoiceOverOutlinedIcon from '@mui/icons-material/RecordVoiceOverOutlined';
import TextFieldsIcon from '@mui/icons-material/TextFields';
import TextureIcon from '@mui/icons-material/Texture';
import { ContentScaling, themeScalingMap } from '~/common/app.theme';
import { DMessageAttachmentFragment, DMessageFragmentId, DVMimeType, isDocPart } from '~/common/stores/chat/chat.fragments';
import { LiveFileIcon } from '~/common/livefile/liveFile.icons';
import { PhVoice } from '~/common/components/icons/phosphor/PhVoice';
import { TooltipOutlined } from '~/common/components/TooltipOutlined';
import { ellipsizeMiddle } from '~/common/util/textUtils';
import { useLiveFileMetadata } from '~/common/livefile/useLiveFileMetadata';
@@ -24,6 +24,15 @@ export const DocSelColor: ColorPaletteProp = 'primary';
const DocUnselColor: ColorPaletteProp = 'primary';
const _styles = {
label: {
whiteSpace: 'nowrap',
fontWeight: 'md',
minWidth: 48,
},
} as const;
export function buttonIconForFragment(part: DMessageAttachmentFragment['part']): React.ComponentType<any> {
const pt = part.pt;
switch (pt) {
@@ -41,7 +50,7 @@ export function buttonIconForFragment(part: DMessageAttachmentFragment['part']):
case 'image':
return ImageOutlinedIcon;
case 'audio':
return RecordVoiceOverOutlinedIcon;
return PhVoice;
default:
const _exhaustiveCheck: never = assetType;
return TextureIcon; // missing zync asset type
@@ -146,10 +155,14 @@ export function DocAttachmentFragmentButton(props: {
if (!isDocPart(fragment.part))
return 'Unexpected: ' + fragment.part.pt;
const buttonText = ellipsizeMiddle(fragment.part.l1Title || fragment.title || 'Document', 28 /* totally arbitrary length */);
const Icon = isSelected ? EditRoundedIcon : buttonIconForFragment(fragment.part);
const fullTitle = fragment.part.l1Title || fragment.title || 'Document';
const buttonText = ellipsizeMiddle(fullTitle, 28 /* totally arbitrary length */);
const showFilenameTooltip = fullTitle !== buttonText;
const labelContent = <Box sx={_styles.label}>{buttonText}</Box>;
return (
<Button
size={props.contentScaling === 'md' ? 'md' : 'sm'}
@@ -171,9 +184,10 @@ export function DocAttachmentFragmentButton(props: {
</Box>
)}
<Box sx={{ display: 'flex', flexDirection: 'column', alignItems: 'flex-start', paddingX: '0.5rem' }}>
<Box sx={{ whiteSpace: 'nowrap', fontWeight: 'md', minWidth: 48 }}>
{buttonText}
</Box>
{showFilenameTooltip
? <TooltipOutlined title={<span style={{ wordBreak: 'break-all' }}>{fullTitle}</span>}>{labelContent}</TooltipOutlined>
: labelContent
}
{/*<Box sx={{ fontSize: 'xs', fontWeight: 'sm' }}>*/}
{/* {fragment.caption}*/}
{/*</Box>*/}
@@ -1,5 +1,7 @@
import * as React from 'react';
import type { SxProps } from '@mui/joy/styles/types';
import { BlocksTextarea } from '~/modules/blocks/BlocksContainers';
import type { ContentScaling } from '~/common/app.theme';
@@ -96,6 +98,8 @@ export function BlockEdit_TextFragment(props: {
const handleEditKeyDown = React.useCallback((e: React.KeyboardEvent<HTMLTextAreaElement>) => {
if (e.key === 'Enter') {
if (e.nativeEvent.isComposing)
return;
const withControl = e.ctrlKey;
if (enterIsNewline ? e.shiftKey : !e.shiftKey) {
e.preventDefault();
@@ -120,6 +124,32 @@ export function BlockEdit_TextFragment(props: {
{ key: ShortcutKey.Esc, description: 'Cancel', level: 3, action: onEscapePressed },
], [isControlled, isEdited, isFocused, onEscapePressed, onSubmit, props.enableRestart]));
// memo style
const sx = React.useMemo((): SxProps | undefined => {
// check sources of custom, and early outs
const isXS = props.contentScaling === 'xs';
const isSquareTop = !!props.squareTopBorder;
if (!isXS && !isSquareTop) return undefined;
if (isSquareTop && !isXS) return _styles.squareTop;
return {
// scaling note: in Chat, this can go xs/sm/md, while in Beam, this is xs/xs/sm
...(isXS && {
fontSize: 'xs',
lineHeight: 'md', // was 1.75 on all
// '--Textarea-paddingBlock': 'calc(0.25rem - 0.5px - var(--variant-borderWidth, 0px))', // not used, overridden in BlocksTextarea
'--Textarea-paddingInline': '6px',
'--Textarea-minHeight': '1.75rem', // was 2rem on 'sm'
'--Icon-fontSize': 'lg', // was 'xl' on 'sm'
'--Textarea-focusedThickness': '1px',
boxShadow: 'none', // too small to show this
}),
...(isSquareTop && _styles.squareTop),
};
}, [props.contentScaling, props.squareTopBorder]);
return (
<BlocksTextarea
variant={/*props.invertedColors ? 'plain' :*/ 'soft'}
@@ -140,7 +170,7 @@ export function BlockEdit_TextFragment(props: {
onKeyDown={handleEditKeyDown}
slotProps={enterIsNewline ? _textAreaSlotPropsEnter : _textAreaSlotPropsDone}
// endDecorator={props.endDecorator}
sx={!props.squareTopBorder ? undefined : _styles.squareTop}
sx={sx}
/>
);
}
@@ -6,6 +6,7 @@ import type { ContentScaling } from '~/common/app.theme';
import type { DMessageErrorPart } from '~/common/stores/chat/chat.fragments';
import type { DMessageRole } from '~/common/stores/chat/chat.message';
import { BlockPartError_NetDisconnected } from './BlockPartError_NetDisconnected';
import { BlockPartError_RequestExceeded } from './BlockPartError_RequestExceeded';
@@ -19,6 +20,19 @@ export function BlockPartError(props: {
// special error presentation, based on hints
switch (props.errorHint) {
case 'aix-net-disconnected':
// determine the 2 'kinds' of disconnection errors in aix.client.ts
const kind =
props.errorText.includes('**network error**') ? 'net-client-closed'
: props.errorText.includes('**connection terminated**') ? 'net-server-closed'
: 'net-unknown-closed';
// For client-side error, we don't show the _NetDisconnected component
if (kind === 'net-client-closed')
break;
return <BlockPartError_NetDisconnected disconnectionKind={kind} messageGeneratorLlmId={props.messageGeneratorLlmId} contentScaling={props.contentScaling} />;
case 'aix-request-exceeded':
return <BlockPartError_RequestExceeded messageGeneratorLlmId={props.messageGeneratorLlmId} contentScaling={props.contentScaling} />;
@@ -0,0 +1,103 @@
import * as React from 'react';
import { Alert, Box, FormHelperText, Switch } from '@mui/joy';
import WifiOffRoundedIcon from '@mui/icons-material/WifiOffRounded';
import type { ContentScaling } from '~/common/app.theme';
import { useLLM } from '~/common/stores/llms/llms.hooks';
import { useModelServiceClientSideFetch } from '~/common/stores/llms/hooks/useModelServiceClientSideFetch';
/**
* Error recovery component for "Connection terminated" errors.
*/
export function BlockPartError_NetDisconnected(props: {
disconnectionKind: 'net-client-closed' | 'net-server-closed' | 'net-unknown-closed';
messageGeneratorLlmId?: string | null;
contentScaling: ContentScaling;
}) {
// external state
const model = useLLM(props.messageGeneratorLlmId) ?? null;
const isServerSideClosed = props.disconnectionKind === 'net-server-closed'; // do not show CSF option for non-server-side
const { csfAvailable, csfActive, csfToggle, vendorName } = useModelServiceClientSideFetch(isServerSideClosed, model);
return (
<Alert
size={props.contentScaling === 'xs' ? 'sm' : 'md'}
color='danger'
variant='plain'
sx={{ display: 'flex', alignItems: 'flex-start', gap: 1 }}
>
<Box sx={{ flex: 1, display: 'flex', flexDirection: 'column', gap: 0.5, alignItems: 'flex-start' }}>
{/* Header */}
<Box sx={{ display: 'flex', gap: 2 }}>
<WifiOffRoundedIcon sx={{ flexShrink: 0, mt: 0.5 }} />
<div>
<Box fontSize='larger'>
Connection Terminated
</Box>
<div>
The connection was unexpectedly closed before the response completed.
</div>
</div>
</Box>
{/* Recovery options */}
{csfAvailable ? <>
{/* Explanation */}
<Box color='text.tertiary' fontSize='sm' my={2}>
<strong>Experimental:</strong> enable direct connection to {vendorName} to bypass server timeouts - then try again.
</Box>
{/* Toggle */}
<Box
sx={{
display: 'flex',
alignItems: 'center',
gap: 2,
p: 2,
borderRadius: 'sm',
bgcolor: 'background.popup',
boxShadow: 'md',
// border: '1px solid',
// borderColor: 'divider',
}}
>
<Box sx={{ flex: 1 }}>
<Box color={!csfActive ? undefined : 'primary.solidBg'} fontWeight='lg' mb={0.5}>
Direct Connection {csfActive && '- Now Try Again'}
</Box>
<FormHelperText>
Connect directly from this client -&gt; {vendorName || 'AI service'}
</FormHelperText>
</Box>
<Switch
checked={csfActive}
onChange={(e) => csfToggle(e.target.checked)}
/>
</Box>
</> : (
<div>
<Box sx={{ color: 'text.secondary', my: 1 }}>
Suggestions:
</Box>
<Box component='ul' sx={{ color: 'text.secondary' }}>
<li>Check your internet connection and try again</li>
<li>The AI service may be experiencing issues - wait a moment and retry</li>
<li>If the issue persists, please let us know promptly on Discord or GitHib</li>
</Box>
</div>
)}
</Box>
</Alert>
);
}
@@ -1,6 +1,6 @@
import * as React from 'react';
import { Alert, Box, FormHelperText, Switch, Typography } from '@mui/joy';
import { Alert, Box, FormHelperText, Switch } from '@mui/joy';
import WarningRoundedIcon from '@mui/icons-material/WarningRounded';
import type { ContentScaling } from '~/common/app.theme';
@@ -19,13 +19,13 @@ export function BlockPartError_RequestExceeded(props: {
// external state
const model = useLLM(props.messageGeneratorLlmId) ?? null;
const { csfAvailable, csfActive, csfToggle } = useModelServiceClientSideFetch(true, model);
const { csfAvailable, csfActive, csfToggle, vendorName } = useModelServiceClientSideFetch(true, model);
return (
<Alert
size={props.contentScaling === 'xs' ? 'sm' : 'md'}
color='danger'
sx={{ display: 'flex', alignItems: 'flex-start', gap: 1 }}
color='warning'
sx={{ display: 'flex', alignItems: 'flex-start', gap: 1, border: '1px solid', borderColor: 'warning.outlinedBorder' }}
>
<WarningRoundedIcon sx={{ flexShrink: 0, mt: 0.25 }} />
@@ -36,70 +36,69 @@ export function BlockPartError_RequestExceeded(props: {
Request Too Large
</Box>
<div>
Your message or attachments exceed the limit of the Vercel edge network.
Your message or attachments exceed the limit of the Vercel edge network
</div>
{/* Recovery options */}
{csfAvailable ? (
<Box sx={{ display: 'flex', flexDirection: 'column', gap: 2 }}>
{csfAvailable ? <>
{/* Explanation */}
<Box color='text.secondary' fontSize='sm'>
<strong>Experimental:</strong> enable direct connection to the AI services, and try again.
</Box>
{/* Toggle */}
<Box
sx={{
display: 'flex',
alignItems: 'center',
gap: 2,
p: 2,
borderRadius: 'sm',
bgcolor: 'background.popup',
boxShadow: 'md',
}}
>
<Box sx={{ flex: 1 }}>
<Box color={!csfActive ? undefined : 'primary.solidBg'} fontWeight='lg' mb={0.5}>
Direct Connection {csfActive && '- Now Try Again'}
</Box>
<FormHelperText>
Bypassing servers and connect directly from this client -&gt; AI provider
</FormHelperText>
</Box>
<Switch
checked={csfActive}
onChange={(e) => csfToggle(e.target.checked)}
/>
</Box>
{/* Regenerate button */}
{/*{props.onRegenerate && (*/}
{/* <Button*/}
{/* size='sm'*/}
{/* variant={csfActive ? 'solid' : 'outlined'}*/}
{/* color={csfActive ? 'success' : 'neutral'}*/}
{/* startDecorator={<RefreshIcon />}*/}
{/* onClick={props.onRegenerate}*/}
{/* sx={{ alignSelf: 'flex-start' }}*/}
{/* >*/}
{/* {csfActive ? 'Regenerate with Direct Connection' : 'Regenerate'}*/}
{/* </Button>*/}
{/*)}*/}
{/* Explanation */}
<Box color='text.secondary' fontSize='sm' my={2}>
<strong>Experimental:</strong> enable Direct Connection to {vendorName} to work around size limitations.
</Box>
) : (
{/* Toggle */}
<Box
sx={{
display: 'flex',
alignItems: 'center',
gap: 2,
p: 2,
borderRadius: 'sm',
bgcolor: 'background.popup',
boxShadow: 'md',
}}
>
<Box sx={{ flex: 1 }}>
<Box color={!csfActive ? undefined : 'primary.solidBg'} fontWeight='lg' mb={0.5}>
Direct Connection {csfActive && '- Now Try Again'}
</Box>
<FormHelperText>
Connect directly from this client -&gt; {vendorName || 'AI service'}
</FormHelperText>
</Box>
<Switch
checked={csfActive}
onChange={(e) => csfToggle(e.target.checked)}
/>
</Box>
{/* Regenerate button */}
{/*{props.onRegenerate && (*/}
{/* <Button*/}
{/* size='sm'*/}
{/* variant={csfActive ? 'solid' : 'outlined'}*/}
{/* color={csfActive ? 'success' : 'neutral'}*/}
{/* startDecorator={<RefreshIcon />}*/}
{/* onClick={props.onRegenerate}*/}
{/* sx={{ alignSelf: 'flex-start' }}*/}
{/* >*/}
{/* {csfActive ? 'Regenerate with Direct Connection' : 'Regenerate'}*/}
{/* </Button>*/}
{/*)}*/}
</> : (
<Box>
<Typography level='body-sm' sx={{ mb: 1 }}>
<strong>Suggestions:</strong>
</Typography>
<Typography component='ul' level='body-sm' sx={{ pl: 2, m: 0 }}>
<Box sx={{ color: 'text.secondary', my: 1 }}>
Suggestions:
</Box>
<Box component='ul' sx={{ color: 'text.secondary' }}>
<li>Use the cleanup button in the right pane to hide old messages</li>
<li>Remove large attachments from the conversation</li>
{/*<li>Reduce conversation length before sending</li>*/}
</Typography>
</Box>
</Box>
)}
</Box>
@@ -96,7 +96,17 @@ export function ContentFragments(props: {
// Content Fragments Edit Zero-State: button to create a new TextContentFragment
if (isEditingText && !props.contentFragments.some(isTextContentFragment))
return !props.onFragmentAddBlank ? null : (
<Button aria-label='message body empty' variant='plain' color='neutral' onClick={props.onFragmentAddBlank} sx={{ justifyContent: 'flex-start' }}>
<Button
aria-label='message body empty'
color={fromAssistant ? 'neutral' : 'primary'}
variant='outlined'
onClick={props.onFragmentAddBlank}
sx={{
justifyContent: 'flex-start',
backgroundColor: fromAssistant ? 'neutral.softBg' : 'primary.softBg',
'&:hover': { backgroundColor: fromAssistant ? 'neutral.softHoverBg' : 'primary.softHoverBg' },
}}
>
add text ...
</Button>
);
@@ -171,6 +171,7 @@ export function BlockPartModelAnnotations(props: {
return (
<Box
data-agi-no-copy // do not copy these buttons: has its own copy functionality
sx={{ mx: 1.5 }}
>
@@ -182,7 +182,7 @@ export function BlockPartModelAux(props: {
return <Box sx={_styles.block}>
{/* Chip to expand/collapse */}
<Box sx={{ display: 'flex', flexWrap: 'wrap', gap: 1, alignItems: 'center', justifyContent: 'space-between' }}>
<Box data-agi-no-copy /* do not copy these buttons */ sx={{ display: 'flex', flexWrap: 'wrap', gap: 1, alignItems: 'center', justifyContent: 'space-between' }}>
<Chip
color={props.isLastFragment ? REASONING_COLOR : 'neutral'}
variant={expanded ? 'solid' : 'soft'}
@@ -1,9 +1,8 @@
import { elevenLabsSpeakText } from '~/modules/elevenlabs/elevenlabs.client';
import type { AixChatGenerateContent_DMessageGuts } from '~/modules/aix/client/aix.client';
import { speakText } from '~/modules/speex/speex.client';
import { isTextContentFragment } from '~/common/stores/chat/chat.fragments';
import type { AixChatGenerateContent_DMessageGuts } from '~/modules/aix/client/aix.client';
import type { PersonaProcessorInterface } from '../chat-persona';
@@ -58,7 +57,7 @@ export class PersonaChatMessageSpeak implements PersonaProcessorInterface {
#speak(text: string) {
console.log('📢 TTS:', text);
this.spokenLine = true;
// fire/forget: we don't want to stall this loop
void elevenLabsSpeakText(text, undefined, false, true);
// fire/forget: we don't want to stall streaming
void speakText(text, undefined, { label: 'Chat message' });
}
}
+4
View File
@@ -101,6 +101,10 @@ export function PromptComposer(props: {
if (e.key !== 'Enter')
return;
// Skip if composing (e.g., CJK input methods) - issue #784
if (e.nativeEvent.isComposing)
return;
// Shift: toggles the 'enter is newline'
if (enterIsNewline ? e.shiftKey : !e.shiftKey) {
if (userHasText)
+20 -4
View File
@@ -72,11 +72,27 @@ export const DevNewsItem: NewsItem = {
// news and feature surfaces
export const NewsItems: NewsItem[] = [
{
versionCode: '2.0.1',
versionName: 'Heavy Critters',
versionDate: new Date('2025-11-24T23:30:00Z'),
versionCode: '2.0.3',
versionName: 'Red Carpet',
versionDate: new Date('2026-02-03T12:00:00Z'),
items: [
{ text: <>New: <B>Opus 4.5</B>, <B>Gemini 3 Pro</B>, <B>Nano Banana Pro</B>, <B>Grok 4.1</B>, <B>GPT-5.1</B>, <B>Kimi K2</B></> },
{ text: <><B>Kimi K2.5</B>, <B>Gemini 3 Flash</B>, <B>GPT Image 1.5</B>, <B>GPT 5.2 Codex</B>, <B issue={921}>Novita.ai</B> models, and xAI search and code execution</> },
{ text: <><B issue={943}>Google Drive</B>: attach docs, sheets, images with optimal LLM conversion</> },
{ text: <>Speech: new <B href='https://inworld.ai'>Inworld</B> support, cancelable, unlimited length</> },
{ text: <>Copy as-seen, reorder messages, AI Injector, PDF auto-OCR</> },
{ text: <>Models: <B issue={941}>duplication</B>, improved parameters, cleaner UI</> },
{ text: <>Fixes, security patches, CJK/IME input</> },
{ text: <>Developers: new Docker build, faster, and smaller containers, AI request injection capabilities in the inspector</>, dev: true },
],
},
{
versionCode: '2.0.2',
versionName: 'Heavy Critters',
versionDate: new Date('2025-12-01T06:00:00Z'), // 2.0.2
// versionDate: new Date('2025-11-24T23:30:00Z'), // 2.0.1
items: [
{ text: <><B>New in 2.0.2</B> Speech synthesis with Web Speech, LocalAI, OpenAI and more</> },
{ text: <><B>Opus 4.5</B>, <B>Gemini 3 Pro</B>, <B>Nano Banana Pro</B>, <B>Grok 4.1</B>, <B>GPT-5.1</B>, <B>Kimi K2</B></> },
{ text: <><B>Image Generation</B> with Azure and LocalAI providers, in addition to OpenAI</> },
{ text: <>Enhanced <B>OpenRouter</B> integration with auto-capabilities and reasoning</> },
{ text: <>Call transcripts, generate persona images, search button in beams</> },
+38
View File
@@ -14,6 +14,37 @@ import { InlineError } from '~/common/components/InlineError';
import type { SimplePersonaProvenance } from '../store-app-personas';
// configuration
const TEMP_DISABLE_YOUTUBE_TRANSCRIPT = true;
function YouTubeDisabledCard() {
return (
<Card
variant='soft'
color='primary'
invertedColors
sx={{
p: 3,
textAlign: 'center',
border: '1px solid',
borderColor: 'primary.solidBg',
}}
>
<Typography level='title-sm' sx={{ mb: 1 }}>
Temporarily Disabled
</Typography>
<Typography level='body-sm' sx={{ mb: 2 }}>
YouTube transcript extraction is currently unavailable due to API changes.
</Typography>
<Typography level='body-xs' color='neutral'>
Download transcripts manually and use the &quot;From Text&quot; option instead.
</Typography>
</Card>
);
}
function YouTubeVideoTranscriptCard(props: { transcript: YTVideoTranscript, onClose: () => void, sx?: SxProps }) {
const { transcript } = props;
return (
@@ -109,6 +140,13 @@ export function FromYouTube(props: {
setVideoID(videoId);
};
if (TEMP_DISABLE_YOUTUBE_TRANSCRIPT)
return <>
<Typography level='title-md' startDecorator={<YouTubeIcon sx={{ color: '#f00' }} />} sx={{ mb: 3 }}>
YouTube -&gt; Persona
</Typography>
<YouTubeDisabledCard />
</>;
return <>
+48 -9
View File
@@ -1,18 +1,17 @@
import * as React from 'react';
import { Accordion, AccordionDetails, accordionDetailsClasses, AccordionGroup, AccordionSummary, accordionSummaryClasses, Avatar, Box, Button, ListItemContent, styled, Tab, TabList, TabPanel, Tabs } from '@mui/joy';
import { Accordion, AccordionDetails, AccordionGroup, AccordionSummary, accordionSummaryClasses, Avatar, Box, Button, ListItemContent, styled, Tab, TabList, TabPanel, Tabs, Typography } from '@mui/joy';
import AddIcon from '@mui/icons-material/Add';
import AutoAwesomeIcon from '@mui/icons-material/AutoAwesome';
import KeyboardCommandKeyOutlinedIcon from '@mui/icons-material/KeyboardCommandKeyOutlined';
import LanguageRoundedIcon from '@mui/icons-material/LanguageRounded';
import MicIcon from '@mui/icons-material/Mic';
import RecordVoiceOverIcon from '@mui/icons-material/RecordVoiceOver';
import ScienceIcon from '@mui/icons-material/Science';
import SearchIcon from '@mui/icons-material/Search';
import TerminalOutlinedIcon from '@mui/icons-material/TerminalOutlined';
import { BrowseSettings } from '~/modules/browse/BrowseSettings';
import { DallESettings } from '~/modules/t2i/dalle/DallESettings';
import { ElevenlabsSettings } from '~/modules/elevenlabs/ElevenlabsSettings';
import { GoogleSearchSettings } from '~/modules/google/GoogleSearchSettings';
import { T2ISettings } from '~/modules/t2i/T2ISettings';
@@ -20,14 +19,15 @@ import type { PreferencesTabId } from '~/common/layout/optima/store-layout-optim
import { AppBreadcrumbs } from '~/common/components/AppBreadcrumbs';
import { DarkModeToggleButton, darkModeToggleButtonSx } from '~/common/components/DarkModeToggleButton';
import { GoodModal } from '~/common/components/modals/GoodModal';
import { Is } from '~/common/util/pwaUtils';
import { PhVoice } from '~/common/components/icons/phosphor/PhVoice';
import { optimaActions } from '~/common/layout/optima/useOptima';
import { useIsMobile } from '~/common/components/useMatchMedia';
import { AppChatSettingsAI } from './AppChatSettingsAI';
import { AppChatSettingsUI } from './settings-ui/AppChatSettingsUI';
import { UxLabsSettings } from './UxLabsSettings';
import { VoiceSettings } from './VoiceSettings';
import { VoiceInSettings } from './VoiceInSettings';
import { VoiceOutSettings } from './VoiceOutSettings';
// configuration
@@ -130,6 +130,7 @@ const _styles = {
// modal: undefined,
modal: {
flexGrow: 1,
backgroundColor: 'background.level1',
} as const,
@@ -209,7 +210,7 @@ export function SettingsModal(props: {
<GoodModal
// title='Preferences' strongerTitle
title={
<AppBreadcrumbs size='md' rootTitle='App'>
<AppBreadcrumbs size='md' rootTitle={isMobile ? 'App' : 'Application'}>
<AppBreadcrumbs.Leaf><b>Preferences</b></AppBreadcrumbs.Leaf>
</AppBreadcrumbs>
}
@@ -271,10 +272,10 @@ export function SettingsModal(props: {
<TabPanel value='voice' color='primary' variant='outlined' sx={_styles.tabPanel}>
<Topics>
<Topic icon={/*'🎙️'*/ <MicIcon />} title='Microphone'>
<VoiceSettings />
<VoiceInSettings isMobile={isMobile} />
</Topic>
<Topic icon={/*'📢'*/ <RecordVoiceOverIcon />} title='ElevenLabs API'>
<ElevenlabsSettings />
<Topic icon={/*'📢'*/ <PhVoice />} title={'Speech'/*<>Voices <GoodBadge badge='New' /></>*/}>
<VoiceOutSettings isMobile={isMobile} />
</Topic>
</Topics>
</TabPanel>
@@ -291,6 +292,44 @@ export function SettingsModal(props: {
</TabPanel>
<TabPanel value='tools' color='primary' variant='outlined' sx={_styles.tabPanel}>
{/* Search Modifier Info */}
<Box sx={{
p: 2,
borderRadius: 'calc(var(--joy-radius-md) - 1px)',
// backgroundColor: 'background.level1',
display: 'flex',
alignItems: 'center',
gap: 2,
}}>
<Button
variant='soft'
color='success'
startDecorator={<SearchIcon />}
sx={{
// this is copied frmo ButtonSearchControl._styles.desktop
minWidth: 100,
justifyContent: 'flex-start',
borderRadius: '18px',
pointerEvents: 'none',
'[data-joy-color-scheme="light"] &': {
bgcolor: '#d5ec31',
},
boxShadow: 'inset 0 2px 4px -1px rgba(0,0,0,0.15)',
textWrap: 'nowrap',
}}
>
Search
</Button>
<Box sx={{ flex: 1 }}>
<Typography level='body-sm' sx={{ fontWeight: 'md', mb: 0.5 }}>
Use the Search button
</Typography>
<Typography level='body-xs' sx={{ color: 'text.secondary' }}>
Modern AI models have native search built-in. Click the Search button when chatting to enable real-time web search.
</Typography>
</Box>
</Box>
<Topics>
<Topic icon={<LanguageRoundedIcon />} title='Load Web Pages (with images)' startCollapsed>
<BrowseSettings />
@@ -0,0 +1,54 @@
import * as React from 'react';
import { FormControl } from '@mui/joy';
import { useChatMicTimeoutMs } from '../chat/store-app-chat';
import type { FormRadioOption } from '~/common/components/forms/FormRadioControl';
import { FormChipControl } from '~/common/components/forms/FormChipControl';
import { FormLabelStart } from '~/common/components/forms/FormLabelStart';
import { LanguageSelect } from '~/common/components/LanguageSelect';
const _minTimeouts: ReadonlyArray<FormRadioOption<string>> = [
{ value: '600', label: '0.6s', description: 'Best for quick calls' },
{ value: '2000', label: '2s', description: 'Standard' },
{ value: '5000', label: '5s', description: 'Breathe' },
{ value: '15000', label: '15s', description: 'Best for thinking' },
] as const;
export function VoiceInSettings(props: { isMobile: boolean }) {
// external state
const [chatTimeoutMs, setChatTimeoutMs] = useChatMicTimeoutMs();
// derived - converts from string keys to numbers and vice versa
const chatTimeoutValue: string = '' + chatTimeoutMs;
const setChatTimeoutValue = React.useCallback((value: string) => {
value && setChatTimeoutMs(parseInt(value));
}, [setChatTimeoutMs]);
return <>
<FormControl orientation='horizontal' sx={{ justifyContent: 'space-between', alignItems: 'center' }}>
<FormLabelStart
title='Language'
description='Mic and voice'
// tooltip='For Microphone input and Voice output. Microphone support varies by browser (iPhone/Safari lacks speech input).'
/>
<LanguageSelect />
</FormControl>
{!props.isMobile && (
<FormChipControl
title='Timeout'
// color='primary'
options={_minTimeouts}
value={chatTimeoutValue}
onChange={setChatTimeoutValue}
/>
)}
</>;
}
@@ -0,0 +1,57 @@
import { SpeexConfigureEngines } from '~/modules/speex/components/SpeexConfigureEngines';
import { useSpeexEngines, useSpeexTtsCharLimit } from '~/modules/speex/store-module-speex';
import { ChatAutoSpeakType, useChatAutoAI } from '../chat/store-app-chat';
import { FormChipControl } from '~/common/components/forms/FormChipControl';
import { FormRadioOption } from '~/common/components/forms/FormRadioControl';
import { FormSwitchControl } from '~/common/components/forms/FormSwitchControl';
const _autoSpeakOptions: FormRadioOption<ChatAutoSpeakType>[] = [
{ value: 'off', label: 'No', description: 'Off' },
{ value: 'firstLine', label: 'Start', description: 'First paragraph' },
{ value: 'all', label: 'Full', description: 'Complete response' },
] as const;
/**
* Voice output settings - Auto-speak mode and TTS engine configuration
*/
export function VoiceOutSettings(props: { isMobile: boolean }) {
// external state
const { autoSpeak, setAutoSpeak } = useChatAutoAI();
const { ttsCharLimit, setTtsCharLimit } = useSpeexTtsCharLimit();
// external state - module
const hasEngines = useSpeexEngines().length > 0;
return <>
{/* Auto-speak setting */}
<FormChipControl
title='Speak Chats'
size='md'
// color='primary'
tooltip={!hasEngines ? 'No voice engines available. Configure a TTS service or use system voice.' : undefined}
disabled={!hasEngines}
options={_autoSpeakOptions}
value={autoSpeak}
onChange={setAutoSpeak}
/>
{/* TTS character limit toggle */}
<FormSwitchControl
title='Speak Cost Guard'
description={ttsCharLimit !== null ? 'Max ~3 min' : 'Unlimited'}
tooltip='Limits text sent to TTS providers, helping prevent unexpected costs with cloud services. By default the limit is 4096 characters (~3 minutes of speech).'
checked={ttsCharLimit !== null}
onChange={(checked) => setTtsCharLimit(checked ? 4096 : null)}
/>
{/* Engine configuration */}
<SpeexConfigureEngines isMobile={props.isMobile} />
</>;
}
-47
View File
@@ -1,47 +0,0 @@
import * as React from 'react';
import { FormControl } from '@mui/joy';
import { useChatMicTimeoutMs } from '../chat/store-app-chat';
import { FormLabelStart } from '~/common/components/forms/FormLabelStart';
import { FormRadioControl } from '~/common/components/forms/FormRadioControl';
import { LanguageSelect } from '~/common/components/LanguageSelect';
import { useIsMobile } from '~/common/components/useMatchMedia';
export function VoiceSettings() {
// external state
const isMobile = useIsMobile();
const [chatTimeoutMs, setChatTimeoutMs] = useChatMicTimeoutMs();
// this converts from string keys to numbers and vice versa
const chatTimeoutValue: string = '' + chatTimeoutMs;
const setChatTimeoutValue = (value: string) => value && setChatTimeoutMs(parseInt(value));
return <>
{/* LanguageSelect: moved from the UI settings (where it logically belongs), just to group things better from an UX perspective */}
<FormControl orientation='horizontal' sx={{ justifyContent: 'space-between', alignItems: 'center' }}>
<FormLabelStart title='Language'
description='ASR and TTS'
tooltip='Currently for Microphone input and Voice output. Microphone support varies by browser (iPhone/Safari lacks speech input). We will use the ElevenLabs MultiLanguage model if a language other than English is selected.' />
<LanguageSelect />
</FormControl>
{!isMobile && <FormRadioControl
title='Mic Timeout'
description={chatTimeoutMs < 1000 ? 'Best for quick calls' : chatTimeoutMs > 5000 ? 'Best for thinking' : 'Standard'}
options={[
{ value: '600', label: '.6s' },
{ value: '2000', label: '2s' },
{ value: '5000', label: '5s' },
{ value: '15000', label: '15s' },
]}
value={chatTimeoutValue} onChange={setChatTimeoutValue}
/>}
</>;
}
+3 -3
View File
@@ -1,5 +1,5 @@
/**
* Copyright (c)2024-2025 Enrico Ros
* Copyright (c)2024-2026 Enrico Ros
*
* This file is include by both the frontend and backend, however depending on the time
* of the build, the values may be different.
@@ -23,8 +23,8 @@ export const Release = {
// this is here to trigger revalidation of data, e.g. models refresh
Monotonics: {
Aix: 42,
NewsVersion: 201,
Aix: 54,
NewsVersion: 203,
},
// Frontend: pretty features
@@ -0,0 +1,146 @@
/**
* Attachment Cloud Files
*
* For future refresh capability, the output fragments should preserve:
* - provider, fileId: to identify the file
* - mimeType: the original cloud MIME type
* - the converter used (stored in outputsHeuristic.actualConverterId)
*
* Google Workspace files (Docs, Sheets, Slides) are auto-exported during
* input loading to standard formats (HTML, CSV, PDF) and then processed
* by standard converters.
*/
import type { AttachmentCloudProviderId } from './attachment.types';
// Error handling
export class CloudFetchError extends Error {
constructor(public readonly code: _CloudFetchErrorCode, public readonly details?: string) {
super(`Cloud fetch error: ${code}${details ? ` - ${details}` : ''}`);
this.name = 'CloudFetchError';
}
}
type _CloudFetchErrorCode = 'AUTH_EXPIRED' | 'NOT_FOUND' | 'FORBIDDEN' | 'RATE_LIMITED' | 'NETWORK_ERROR' | 'NOT_IMPLEMENTED' | 'FETCH_FAILED';
// Utility functions
/**
* Google Workspace files can't be downloaded directly - they must be exported.
* We prioritize AI-friendly formats (text > binary).
*
* Docs: md, docx, pdf, txt, rtf, odt, epub, html.zip
* Sheets: xlsx, pdf, csv (1st sheet), tsv, ods
* Slides: pptx, pdf, txt, png/jpg/svg (1st slide)
* Drawings: png, pdf, jpg, svg
*
* Regular files: we'll return no conversion
*
* @see https://developers.google.com/workspace/drive/api/guides/ref-export-formats
*/
const _GOOGLE_WORKSPACE_EXPORT: Record<string, { mimeType: string; ext: string, converter: string }> = {
'application/vnd.google-apps.document': { mimeType: 'text/markdown', ext: '.md', converter: 'Doc -> ' },
'application/vnd.google-apps.spreadsheet': { mimeType: 'text/csv', ext: '.csv', converter: 'Sheet -> ' },
'application/vnd.google-apps.presentation': { mimeType: 'application/pdf', ext: '.pdf', converter: 'Slides -> ' },
'application/vnd.google-apps.drawing': { mimeType: 'image/svg+xml', ext: '.svg', converter: 'Drawing -> ' },
};
export function attachmentCloudGoogleWorkspaceExportMIME(cloudMimeType: string): string | undefined {
return _GOOGLE_WORKSPACE_EXPORT[cloudMimeType]?.mimeType;
}
export function attachmentCloudConverterPrefix(cloudMimeType: string): string {
return _GOOGLE_WORKSPACE_EXPORT[cloudMimeType]?.converter || 'Drive -> ';
}
// Fetcher
/**
* Fetch a file from a cloud provider.
*
* @param provider - The cloud provider ID
* @param fileId - The file ID in the provider's system
* @param accessToken - OAuth access token
* @param exportMimeType - For native formats (Docs/Sheets), the export format
* @returns The file content as a Blob
*/
export async function attachmentCloudFetchFile(
provider: AttachmentCloudProviderId,
fileId: string,
accessToken: string,
exportMimeType?: string,
): Promise<Blob> {
switch (provider) {
case 'gdrive':
return _fetchGoogleDriveFile(fileId, accessToken, exportMimeType);
case 'onedrive':
case 'dropbox':
throw new CloudFetchError('NOT_IMPLEMENTED', `${provider} support coming soon`);
default:
throw new CloudFetchError('NOT_IMPLEMENTED', `Unknown provider: ${provider}`);
}
}
/**
* Google Drive API - Fetch file content
* https://developers.google.com/drive/api/reference/rest/v3/files/get
* https://developers.google.com/drive/api/reference/rest/v3/files/export
*/
async function _fetchGoogleDriveFile(
fileId: string,
accessToken: string,
exportMimeType?: string,
): Promise<Blob> {
// for native Google Workspace files, use export endpoint
const url = exportMimeType
? `https://www.googleapis.com/drive/v3/files/${encodeURIComponent(fileId)}/export?mimeType=${encodeURIComponent(exportMimeType)}`
: `https://www.googleapis.com/drive/v3/files/${encodeURIComponent(fileId)}?alt=media`;
const response = await fetch(url, {
headers: {
'Authorization': `Bearer ${accessToken}`,
},
}).catch((error) => {
console.log('[DEV] Network error while fetching Google Drive file:', { error });
throw new CloudFetchError('NETWORK_ERROR', error?.message || String(error));
});
// NOTE: we shall consider moving this to use fetchResponseOrTRPCThrow instead of this custom small impl..
if (!response.ok) {
const errorCode = _mapHttpStatusToErrorCode(response.status);
let details = `${response.status}: ${response.statusText}`;
try {
const errorBody = await response.text();
if (errorBody) details += ` - ${errorBody.slice(0, 200)}`;
} catch { /* ignore */
}
throw new CloudFetchError(errorCode, details);
}
return response.blob();
}
function _mapHttpStatusToErrorCode(status: number): _CloudFetchErrorCode {
switch (status) {
case 401:
return 'AUTH_EXPIRED';
case 403:
return 'FORBIDDEN';
case 404:
return 'NOT_FOUND';
case 429:
return 'RATE_LIMITED';
default:
return 'FETCH_FAILED';
}
}
@@ -59,17 +59,35 @@ export async function imageDataToImageAttachmentFragmentViaDBlob(
origin: { // User originated
ot: 'user',
source: 'attachment',
media: source.media === 'file' ? source.origin : source.media === 'url' ? 'url' : 'unknown',
url: source.media === 'url' ? source.url : undefined,
fileName: source.media === 'file' ? source.refPath : undefined,
media:
source.media === 'file' ? source.origin
: source.media === 'url' ? 'url'
: source.media === 'cloud' ? source.provider
: 'unknown',
url:
source.media === 'url' ? source.url
: source.media === 'cloud' ? source.webViewLink
: undefined,
fileName:
source.media === 'file' ? source.refPath
: source.media === 'cloud' ? source.fileName
: undefined,
},
});
// use title if available, otherwise use the source refPath/refUrl/fileName
const refTextSummary = title || (
source.media === 'file' ? source.refPath
: source.media === 'url' ? source.refUrl
: source.media === 'cloud' ? source.fileName
: undefined
);
// Future-proof: create a Zync Image Asset reference attachment fragment, with the legacy image_ref part for compatibility for the time being
return createZyncAssetReferenceAttachmentFragment(
title, caption,
nanoidToUuidV4(dblobAssetId, 'convert-dblob-to-dasset'),
title || (source.media === 'file' ? source.refPath : source.media === 'url' ? source.refUrl : undefined), // use title if available, otherwise use the source refPath or refUrl
refTextSummary,
'image',
{
pt: 'image_ref' as const,
@@ -77,7 +95,7 @@ export async function imageDataToImageAttachmentFragmentViaDBlob(
...(title ? { altText: title } : {}),
...(imageWidth ? { width: imageWidth } : {}),
...(imageHeight ? { height: imageHeight } : {}),
}
},
);
} catch (error) {
console.error('imageAttachment: Error processing image:', error);
@@ -11,12 +11,14 @@ import { convert_Base64DataURL_To_Base64WithMimeType, convert_Base64WithMimeType
import { getDomainModelConfiguration } from '~/common/stores/llms/hooks/useModelDomain';
import { htmlTableToMarkdown } from '~/common/util/htmlTableToMarkdown';
import { humanReadableHyphenated } from '~/common/util/textUtils';
import { ocrImageWithProgress, ocrPdfPagesWithProgress } from '~/common/util/ocrUtils';
import { pdfToImageDataURLs, pdfToText } from '~/common/util/pdfUtils';
import { createDMessageDataInlineText, createDocAttachmentFragment, DMessageAttachmentFragment, DMessageDataInline, DMessageDocPart, DVMimeType, isContentOrAttachmentFragment, isDocPart, specialContentPartToDocAttachmentFragment } from '~/common/stores/chat/chat.fragments';
import type { AttachmentCreationOptions, AttachmentDraft, AttachmentDraftConverter, AttachmentDraftId, AttachmentDraftInput, AttachmentDraftSource, AttachmentDraftSourceOriginFile, DraftEgoFragmentsInputData, DraftWebInputData, DraftYouTubeInputData } from './attachment.types';
import type { AttachmentsDraftsStore } from './store-attachment-drafts_slice';
import { attachmentCloudConverterPrefix, attachmentCloudFetchFile, attachmentCloudGoogleWorkspaceExportMIME, CloudFetchError } from './attachment.cloud';
import { attachmentGetLiveFileId, attachmentSourceSupportsLiveFile } from './attachment.livefile';
import { guessInputContentTypeFromMime, heuristicMimeTypeFixup, mimeTypeIsDocX, mimeTypeIsPDF, mimeTypeIsPlainText, mimeTypeIsSupportedImage, reverseLookupMimeType } from './attachment.mimetypes';
import { imageDataToImageAttachmentFragmentViaDBlob } from './attachment.dblobs';
@@ -27,6 +29,11 @@ const PDF_IMAGE_QUALITY = 0.5;
const ENABLE_TEXT_AND_IMAGES = false; // [PROD] ?
const DOCPART_DEFAULT_VERSION = 1;
// PDF text extraction quality thresholds
const IMAGE_LOW_TEXT_THRESHOLD = 80; // chars per image - below this, consider the image as low-text (photo-like) rather than document-like
const PDF_LOW_TEXT_THRESHOLD = 160; // chars per page - below this, consider the PDF as scanned/image-based
const PDF_FALLBACK_MAX_IMAGES = 32; // max pages to convert to images when auto-falling back (to respect LLM limits)
// internal mimes, only used to route data within us (source -> input -> converters)
const INT_MIME_VND_AGI_EGO_FRAGMENTS = 'application/vnd.agi.ego.fragments';
@@ -63,7 +70,8 @@ export function attachmentCreate(source: AttachmentDraftSource): AttachmentDraft
export async function attachmentLoadInputAsync(source: Readonly<AttachmentDraftSource>, edit: (changes: Partial<Omit<AttachmentDraft, 'outputFragments'>>) => void) {
edit({ inputLoading: true });
switch (source.media) {
const sourceMedia = source.media;
switch (sourceMedia) {
// Download URL (page, file, ..) and attach as input
case 'url':
@@ -141,6 +149,7 @@ export async function attachmentLoadInputAsync(source: Readonly<AttachmentDraftS
} else
edit({ inputError: 'No content or file found at this link' });
} catch (error: any) {
console.log('[DEV] Issue downloading page for attachment:', { error });
edit({ inputError: `Issue downloading page: ${error?.message || (typeof error === 'string' ? error : JSON.stringify(error))}` });
}
break;
@@ -221,6 +230,34 @@ export async function attachmentLoadInputAsync(source: Readonly<AttachmentDraftS
}
break;
case 'cloud':
const cloudLabel = source.fileName || 'Cloud File';
const cloudRef = source.webViewLink || `${source.provider}:${source.fileId}`;
edit({ label: cloudLabel, ref: cloudRef });
try {
// fetch / export to the destination mime
const exportMime = attachmentCloudGoogleWorkspaceExportMIME(source.mimeType);
const cloudBlob = await attachmentCloudFetchFile(source.provider, source.fileId, source.accessToken, exportMime);
// use export mime if we exported, otherwise use source or detected mime
const resultMime = exportMime || source.mimeType /* provided outside */ || cloudBlob.type /* connection */ || 'application/octet-stream';
edit({
input: {
mimeType: resultMime,
data: cloudBlob,
dataSize: cloudBlob.size,
},
});
} catch (error: unknown) {
const errorMessage = error instanceof CloudFetchError
? `${error.code}: ${error.details || error.message}`
: `Failed to download: ${error instanceof Error ? error.message : String(error)}`;
edit({ inputError: errorMessage });
}
break;
case 'ego':
edit({
label: source.label,
@@ -231,6 +268,10 @@ export async function attachmentLoadInputAsync(source: Readonly<AttachmentDraftS
},
});
break;
default:
const _exhaustiveCheck: never = sourceMedia;
break;
}
edit({ inputLoading: false });
@@ -251,6 +292,7 @@ export function attachmentDefineConverters(source: AttachmentDraftSource, input:
const converters: AttachmentDraftConverter[] = [];
const autoAddImages = ENABLE_TEXT_AND_IMAGES && !!options?.hintAddImages;
const fromCloud = source.media === 'cloud';
switch (true) {
@@ -258,6 +300,7 @@ export function attachmentDefineConverters(source: AttachmentDraftSource, input:
case mimeTypeIsPlainText(input.mimeType):
// handle a secondary layer of HTML 'text' origins: drop, paste, and clipboard-read
const textOriginHtml = source.media === 'text' && input.altMimeType === 'text/html' && !!input.altData;
const textOriginClipboard = source.media === 'text' && ['clipboard-read', 'paste'].includes(source.method);
const isHtmlTable = !!input.altData?.startsWith('<table');
// p1: Tables
@@ -265,12 +308,21 @@ export function attachmentDefineConverters(source: AttachmentDraftSource, input:
converters.push({ id: 'rich-text-table', name: 'Markdown Table' });
// p2: Text
converters.push({ id: 'text', name: attachmentSourceSupportsLiveFile(source) ? 'Text (Live)' : 'Text' });
if (fromCloud && input.mimeType === 'text/markdown') {
converters.push({ id: 'text', name: 'Markdown' });
} else {
converters.push({ id: 'text', name: attachmentSourceSupportsLiveFile(source) ? 'Text (Live)' : 'Text' });
if (!textOriginHtml && textOriginClipboard) {
converters.push({ id: 'text-markdown', name: 'Text -> Markdown' });
converters.push({ id: 'text-cleaner', name: 'Text -> Clean HTML' });
}
}
// p3: Html
// p3: Html -> Markdown, and Html
if (textOriginHtml) {
converters.push({ id: 'rich-text-cleaner', name: 'Cleaner HTML' });
converters.push({ id: 'rich-text', name: 'HTML · Heavy' });
converters.push({ id: 'rich-text-markdown', name: 'HTML -> Markdown' });
converters.push({ id: 'rich-text-cleaner', name: 'HTML -> Clean HTML' });
}
break;
@@ -283,16 +335,18 @@ export function attachmentDefineConverters(source: AttachmentDraftSource, input:
converters.push({ id: 'image-original', name: 'Image (original quality)', disabled: !inputImageMimeSupported });
if (!inputImageMimeSupported)
converters.push({ id: 'image-to-default', name: `As Image (${PLATFORM_IMAGE_MIMETYPE})` });
converters.push({ id: 'image-caption', name: 'Caption (Text)', disabled: visionModelMissing });
converters.push({ id: 'image-caption', name: 'AI Caption (Text)', disabled: visionModelMissing });
converters.push({ id: 'unhandled', name: 'No Image' });
converters.push({ id: 'image-ocr', name: 'Add Text (OCR)', isCheckbox: true });
break;
// PDF
case mimeTypeIsPDF(input.mimeType):
converters.push({ id: 'pdf-text', name: 'PDF To Text', isActive: !autoAddImages || undefined });
converters.push({ id: 'pdf-images', name: 'PDF To Images' });
converters.push({ id: 'pdf-text-and-images', name: 'PDF Text & Images (best)', isActive: autoAddImages });
converters.push({ id: 'pdf-auto', name: 'Auto', isActive: !autoAddImages });
converters.push({ id: 'pdf-text', name: 'PDF Text' });
converters.push({ id: 'pdf-images-ocr', name: 'PDF -> OCR (for scans)' });
converters.push({ id: 'pdf-images', name: 'PDF -> Images' });
converters.push({ id: 'pdf-text-and-images', name: 'PDF -> Text + Images', isActive: autoAddImages });
break;
// DOCX
@@ -337,6 +391,12 @@ export function attachmentDefineConverters(source: AttachmentDraftSource, input:
break;
}
// cosmetic for cloud: prepend cloud label prefixes
const cloudLabelPrefix = source.media === 'cloud' ? attachmentCloudConverterPrefix(source.mimeType) : '';
if (cloudLabelPrefix)
for (const converter of converters)
converter.name = cloudLabelPrefix + converter.name;
edit({ converters });
}
@@ -380,7 +440,8 @@ function _prepareDocData(source: AttachmentDraftSource, input: Readonly<Attachme
srcFileSize: source.fileWithHandle.size || input.dataSize,
};
switch (source.origin) {
const sourceOrigin = source.origin;
switch (sourceOrigin) {
case 'camera':
fileTitle = source.refPath || _lowCollisionRefString('Camera Photo', 6);
break;
@@ -398,6 +459,10 @@ function _prepareDocData(source: AttachmentDraftSource, input: Readonly<Attachme
case 'drop':
fileTitle = source.refPath || _lowCollisionRefString('Dropped File', 6);
break;
default:
const _exhaustiveCheck: never = sourceOrigin;
fileTitle = 'File';
break;
}
return {
title: fileTitle,
@@ -415,6 +480,25 @@ function _prepareDocData(source: AttachmentDraftSource, input: Readonly<Attachme
refString: humanReadableHyphenated(textRef),
};
// Cloud files
case 'cloud':
const cloudFileName = source.fileName || 'Cloud File';
const cloudProviderLabel = source.provider === 'gdrive' ? 'Google Drive'
: source.provider === 'onedrive' ? 'OneDrive'
: source.provider === 'dropbox' ? 'Dropbox'
: 'Cloud';
const cloudRef = `${source.provider}-${source.fileName || _lowCollisionRefString('file', 6)}`;
return {
title: cloudFileName,
caption: `From ${cloudProviderLabel}`,
refString: humanReadableHyphenated(cloudRef),
// TODO: expand this to allow future redownload - or other location but for the same purpose
docMeta: {
srcFileName: source.fileName,
srcFileSize: source.fileSize || input.dataSize,
},
};
// The application attaching pieces of itself
case 'ego':
const egoKind = source.method === 'ego-fragments' ? 'Chat Message' : '';
@@ -478,6 +562,8 @@ export async function attachmentPerformConversion(
edit(attachment.id, {
outputsConverting: true,
outputsConversionProgress: null,
outputWarnings: undefined,
outputsHeuristic: undefined,
});
// apply converter to the input
@@ -490,35 +576,69 @@ export async function attachmentPerformConversion(
switch (converter.id) {
// text as-is
// text
case 'text':
case 'text-cleaner':
case 'text-markdown':
const possibleLiveFileId = await attachmentGetLiveFileId(source);
const textContent = await _inputDataToString(input.data, 'text');
const textualInlineData = createDMessageDataInlineText(textContent, input.mimeType);
let textContent = await _inputDataToString(input.data, 'text');
let textContentMime = input.mimeType || 'text/plain';
switch (converter.id) {
case 'text-cleaner':
textContent = _cleanPossibleHtmlText(textContent);
break;
case 'text-markdown':
try {
const { convertHtmlToMarkdown } = await import('./file-converters/HtmlToMarkdown');
textContent = convertHtmlToMarkdown(textContent);
textContentMime = 'text/markdown';
} catch (error) {
console.log('[DEV] Error converting Text (HTML) to Markdown:', error);
}
break;
}
const textualInlineData = createDMessageDataInlineText(textContent, textContentMime);
newFragments.push(createDocAttachmentFragment(title, caption, _guessDocVDT(input.mimeType), textualInlineData, refString, DOCPART_DEFAULT_VERSION, docMeta, possibleLiveFileId));
break;
// html as-is
// html
case 'rich-text':
case 'rich-text-cleaner':
case 'rich-text-markdown':
let richText: string;
if (input.altData)
richText = input.altData;
else if (input.mimeType === 'text/html')
richText = await _inputDataToString(input.data, 'rich-text');
else
richText = '';
let richTextMimeType = 'text/html';
// html -> cleaner/html or markdown
switch (converter.id) {
case 'rich-text-cleaner':
richText = _cleanPossibleHtmlText(richText);
richTextMimeType = 'text/html';
break;
case 'rich-text-markdown':
try {
const { convertHtmlToMarkdown } = await import('./file-converters/HtmlToMarkdown');
richText = convertHtmlToMarkdown(richText);
richTextMimeType = 'text/markdown';
} catch (error) {
console.log('[DEV] Error converting HTML to Markdown:', error);
}
break;
}
// NOTE: before we had the following: createTextAttachmentFragment(ref || '\n<!DOCTYPE html>', input.altData!), which
// was used to wrap the HTML in a code block to facilitate AutoRenderBlocks's parser. Historic note, for future debugging.
const richTextData = createDMessageDataInlineText(input.altData || '', input.altMimeType);
const richTextData = createDMessageDataInlineText(richText, richTextMimeType);
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.VndAgiCode, richTextData, refString, DOCPART_DEFAULT_VERSION, docMeta));
break;
// html cleaned
case 'rich-text-cleaner':
const cleanerHtml = (input.altData || '')
// remove class and style attributes
.replace(/<[^>]+>/g, (tag) =>
tag.replace(/ class="[^"]*"/g, '').replace(/ style="[^"]*"/g, ''),
)
// remove svg elements
.replace(/<svg[^>]*>.*?<\/svg>/g, '');
const cleanedHtmlData = createDMessageDataInlineText(cleanerHtml, 'text/html');
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.VndAgiCode, cleanedHtmlData, refString, DOCPART_DEFAULT_VERSION, docMeta));
break;
// html to markdown table
case 'rich-text-table':
let tableData: DMessageDataInline;
@@ -570,23 +690,14 @@ export async function attachmentPerformConversion(
case 'image-ocr':
if (!_expectBlob(input.data, 'Image OCR converter')) break;
try {
let lastProgress = -1;
const { recognize } = await import('tesseract.js');
const result = await recognize(input.data, undefined, {
errorHandler: e => console.error(e),
logger: (message) => {
if (message.status === 'recognizing text') {
if (message.progress > lastProgress + 0.01) {
lastProgress = message.progress;
edit(attachment.id, { outputsConversionProgress: lastProgress });
}
}
},
});
const imageText = result.data.text;
// Image -> OCR -> Inline text doc
const imageText = await ocrImageWithProgress(input.data, (progress) => edit(attachment.id, { outputsConversionProgress: progress }));
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(imageText, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'image' }));
// warn if very little text was extracted (likely a photo/diagram rather than text)
if (imageText.trim().length < IMAGE_LOW_TEXT_THRESHOLD)
edit(attachment.id, { outputWarnings: ['Very little text extracted - this image may not contain readable text.'] });
} catch (error) {
console.error(error);
console.error('[Image OCR Error]', error);
}
break;
@@ -615,26 +726,111 @@ export async function attachmentPerformConversion(
} catch (error: any) {
console.log('[DEV] Failed to caption image:', error);
const errorText = `[Captioning failed: ${error?.message || String(error)}]`;
edit(attachment.id, { outputWarnings: [errorText] });
newFragments.push(createDocAttachmentFragment(title, caption + ' (Error)', DVMimeType.TextPlain, createDMessageDataInlineText(errorText, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'image-caption' }));
}
break;
// pdf to text
case 'pdf-text':
if (!_expectBlob(input.data, 'PDF text converter')) break;
// Convert Blob to ArrayBuffer for PDF.js
const pdfText = await pdfToText(await input.data.arrayBuffer(), (progress: number) => {
edit(attachment.id, { outputsConversionProgress: progress });
});
if (pdfText.trim().length < 2) {
// Warn the user if no text is extracted
// edit(attachment.id, { inputError: 'No text found in the PDF file.' });
} else
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(pdfText, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' }));
// pdf-auto: intelligent conversion with fallback chain (text → OCR → images)
case 'pdf-auto':
if (!_expectBlob(input.data, 'PDF auto converter')) break;
try {
// Phase 1: Try text extraction (0-20% progress)
const pdfArrayBuffer = await input.data.arrayBuffer();
// [pdf-text] Extract text with quality metadata
const pdfTextResult = await pdfToText(pdfArrayBuffer, (progress: number) => {
// Reserve 0-20% for text extraction attempt, 20-100% for potential image fallback
edit(attachment.id, { outputsConversionProgress: progress * 0.2 });
});
// Check text density to detect scanned/image-based PDFs
if (pdfTextResult.avgCharsPerPage >= PDF_LOW_TEXT_THRESHOLD) {
// Good text extraction - use it
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(pdfTextResult.text, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' }));
edit(attachment.id, {
outputsHeuristic: { isAuto: true, actualConverterId: 'pdf-text', explain: `${pdfTextResult.avgCharsPerPage.toFixed(0)} chars/page` },
});
} else {
// Low text density - try OCR
// console.log(`[PDF Auto] Low text density (${pdfTextResult.avgCharsPerPage.toFixed(0)} chars/page), trying OCR...`);
// [pdf-images] Phase 2: Render pages to images (20-40% progress)
const pdfArrayBufferForImages = await input.data.arrayBuffer();
const imageDataURLs = await pdfToImageDataURLs(pdfArrayBufferForImages, PLATFORM_IMAGE_MIMETYPE, PDF_IMAGE_QUALITY, PDF_IMAGE_PAGE_SCALE, (progress) => {
edit(attachment.id, { outputsConversionProgress: 0.2 + progress * 0.2 });
});
// Limit pages for OCR (performance)
const pagesToProcess = Math.min(imageDataURLs.length, PDF_FALLBACK_MAX_IMAGES);
const imagesToOcr = imageDataURLs.slice(0, pagesToProcess);
// Phase 3: Try OCR on rendered pages (40-90% progress)
try {
// [pdf-images-ocr] OCR the images
const ocrResult = await ocrPdfPagesWithProgress(imagesToOcr, (progress) => {
edit(attachment.id, { outputsConversionProgress: 0.4 + progress * 0.5 });
});
if (ocrResult.avgCharsPerPage >= PDF_LOW_TEXT_THRESHOLD) {
// OCR yielded good text - use it
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(ocrResult.text, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' }));
const truncNote = pdfTextResult.pageCount > pagesToProcess ? ` (${pagesToProcess}/${pdfTextResult.pageCount} pages)` : '';
edit(attachment.id, {
outputsHeuristic: { isAuto: true, actualConverterId: 'pdf-images-ocr', explain: /*OCR extracted */`${ocrResult.avgCharsPerPage.toFixed(0)} chars/page${truncNote}` },
});
} else {
// OCR also yielded poor results - fall back to images
// console.log(`[PDF Auto] OCR also sparse (${ocrResult.avgCharsPerPage.toFixed(0)} chars/page), falling back to images`);
for (let i = 0; i < pagesToProcess; i++) {
const pdfPageImage = imageDataURLs[i];
const pdfPageImageF = await imageDataToImageAttachmentFragmentViaDBlob(pdfPageImage.mimeType, pdfPageImage.base64Data, source, `${title} (pg. ${i + 1})`, caption, false, false);
if (pdfPageImageF)
newFragments.push(pdfPageImageF);
}
const truncNote = pdfTextResult.pageCount > pagesToProcess ? ` (${pagesToProcess}/${pdfTextResult.pageCount} pages)` : '';
edit(attachment.id, {
outputsHeuristic: { isAuto: true, actualConverterId: 'pdf-images', explain: `not a text page${truncNote}` },
});
}
} catch (ocrError) {
// OCR failed - fall back to images
console.warn('[PDF Auto] OCR failed, falling back to images:', ocrError);
for (let i = 0; i < pagesToProcess; i++) {
const pdfPageImage = imageDataURLs[i];
const pdfPageImageF = await imageDataToImageAttachmentFragmentViaDBlob(pdfPageImage.mimeType, pdfPageImage.base64Data, source, `${title} (pg. ${i + 1})`, caption, false, false);
if (pdfPageImageF)
newFragments.push(pdfPageImageF);
}
edit(attachment.id, {
outputsHeuristic: { isAuto: true, actualConverterId: 'pdf-images', explain: 'OCR failed, attached as images' },
});
}
}
} catch (error) {
console.error('Error in PDF auto conversion:', error);
}
break;
// pdf to images
// pdf-text: strict text extraction, no fallback (honors user choice)
case 'pdf-text':
if (!_expectBlob(input.data, 'PDF text converter')) break;
try {
const pdfTextResult = await pdfToText(await input.data.arrayBuffer(), progress => edit(attachment.id, { outputsConversionProgress: progress }));
// Always output text, even if sparse (user explicitly chose this)
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(pdfTextResult.text, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' }));
edit(attachment.id, {
// warn if very little text was extracted (likely a scanned PDF)
outputWarnings: pdfTextResult.avgCharsPerPage >= 20 ? undefined : ['Very little text extracted - this PDF may be scanned. Try "Auto" or "OCR (for scans)" mode.'],
outputsHeuristic: { isAuto: false, actualConverterId: 'pdf-text', explain: `${pdfTextResult.avgCharsPerPage.toFixed(0)} chars/page` },
});
} catch (error) {
console.error('Error in PDF text extraction:', error);
}
break;
// pdf-images: render all pages as images (honors user choice)
case 'pdf-images':
if (!_expectBlob(input.data, 'PDF images converter')) break;
// Convert Blob to ArrayBuffer for PDF.js
@@ -647,11 +843,39 @@ export async function attachmentPerformConversion(
if (pdfPageImageF)
newFragments.push(pdfPageImageF);
}
edit(attachment.id, {
outputsHeuristic: { isAuto: false, actualConverterId: 'pdf-images', explain: `${imageDataURLs.length} pages` },
});
} catch (error) {
console.error('Error converting PDF to images:', error);
}
break;
// pdf-images-ocr: force OCR on all pages (for scanned documents)
case 'pdf-images-ocr':
if (!_expectBlob(input.data, 'PDF OCR converter')) break;
try {
// Render pages to images (0-40% progress)
const imageDataURLs = await pdfToImageDataURLs(await input.data.arrayBuffer(), PLATFORM_IMAGE_MIMETYPE, PDF_IMAGE_QUALITY, PDF_IMAGE_PAGE_SCALE, (progress) => {
edit(attachment.id, { outputsConversionProgress: progress * 0.4 });
});
// OCR all pages (40-100% progress)
const ocrResult = await ocrPdfPagesWithProgress(imageDataURLs, (progress) => {
edit(attachment.id, { outputsConversionProgress: 0.4 + progress * 0.6 });
});
newFragments.push(createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(ocrResult.text, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' }));
edit(attachment.id, {
// warn if very little text was extracted (likely a scanned PDF)
outputWarnings: ocrResult.avgCharsPerPage >= 20 ? undefined : ['Very little text extracted via OCR - this PDF may contain mostly images/diagrams.'],
outputsHeuristic: { isAuto: false, actualConverterId: 'pdf-images-ocr', explain: `${ocrResult.avgCharsPerPage.toFixed(0)} chars/page from ${ocrResult.pageCount} pages` },
});
} catch (error) {
console.error('Error in PDF OCR:', error);
}
break;
// pdf to text and images
case 'pdf-text-and-images':
if (!_expectBlob(input.data, 'PDF text and images converter')) break;
@@ -673,18 +897,21 @@ export async function attachmentPerformConversion(
}
// duplicated from 'pdf-text'
const pdfText = await pdfToText(pdfArrayBufferForText, (progress: number) => {
const pdfTextResult = await pdfToText(pdfArrayBufferForText, (progress: number) => {
edit(attachment.id, { outputsConversionProgress: 0.5 + progress / 2 }); // Update progress (50% to 100%)
});
if (pdfText.trim().length < 2) {
// Do not warn the user, as hopefully the images are useful
} else {
const textFragment = createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(pdfText, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' });
if (pdfTextResult.text.trim().length >= 2) {
// Add text fragment if there's meaningful text
const textFragment = createDocAttachmentFragment(title, caption, DVMimeType.TextPlain, createDMessageDataInlineText(pdfTextResult.text, 'text/plain'), refString, DOCPART_DEFAULT_VERSION, { ...docMeta, srcOcrFrom: 'pdf' });
newFragments.push(textFragment);
}
// Note: if text is sparse, images are still attached (user explicitly chose text+images)
// Add the text fragment first, then the image fragments
newFragments.push(...imageFragments);
edit(attachment.id, {
outputsHeuristic: { isAuto: false, actualConverterId: 'pdf-text-and-images', explain: `${pdfTextResult.avgCharsPerPage.toFixed(0)} chars/page + ${imageFragments.length} images` },
});
} catch (error) {
console.error('Error converting PDF to text and images:', error);
}
@@ -801,6 +1028,12 @@ export async function attachmentPerformConversion(
case 'unhandled':
// force the user to explicitly select 'as text' if they want to proceed
break;
default:
const _exhaustiveCheck: never = converter.id;
console.warn('[DEV] Unhandled converter type:', _exhaustiveCheck);
break;
}
}
@@ -843,6 +1076,19 @@ async function _inputDataToString(data: AttachmentDraftInput['data'], debugLocat
return '';
}
/**
* Simple Client-side cleaning of possible HTML
*/
function _cleanPossibleHtmlText(inputStr: string): string {
return inputStr
// remove class and style attributes
.replace(/<[^>]+>/g, (tag) =>
tag.replace(/ class="[^"]*"/g, '').replace(/ style="[^"]*"/g, ''),
)
// remove svg elements
.replace(/<svg[^>]*>.*?<\/svg>/g, '');
}
/**
* Special function to convert a list of files to Attachment Fragments, without passing through the attachments system
@@ -24,6 +24,16 @@ export type AttachmentDraft = {
outputsConversionProgress: number | null;
outputFragments: DMessageAttachmentFragment[];
// Warnings for poor conversions (e.g. scanned PDF with text extraction rather than OCR)
outputWarnings?: string[];
// Tracks what method was actually used (especially for Auto mode)
outputsHeuristic?: {
isAuto: boolean;
actualConverterId: AttachmentDraftConverterType;
explain?: string; // e.g., "42 chars/page detected"
};
// metadata: {
// creationDate?: Date; // Creation date of the file
// modifiedDate?: Date; // Last modified date of the file
@@ -33,6 +43,13 @@ export type AttachmentDraft = {
export type AttachmentDraftId = string;
export type AttachmentCreationOptions = {
/** Also attach an image representation of the attachment. Requires Release.Features.ENABLE_TEXT_AND_IMAGES as well. */
hintAddImages?: boolean;
}
export type AttachmentCloudProviderId = 'gdrive' | 'onedrive' | 'dropbox';
// 0. draft source (filled at the onset)
@@ -51,6 +68,23 @@ export type AttachmentDraftSource = {
method: 'clipboard-read' | AttachmentDraftSourceOriginDTO;
textPlain?: string;
textHtml?: string;
} | {
media: 'cloud';
origin: AttachmentDraftSourceOriginCloud;
// auth for fetching
accessToken: string;
// tokenExpiresAt?: number; // optional for staleness detection, unix ts
// recipe for fetching
provider: AttachmentCloudProviderId;
fileId: string;
mimeType: string; // cloud-native MIME (e.g., 'application/vnd.google-apps.document')
// decorative
fileName: string;
fileSize?: number;
webViewLink?: string; // link to view in cloud provider's UI
} | {
// special type for attachments thar are references to self (ego, application) objects
media: 'ego';
@@ -65,10 +99,7 @@ export type AttachmentDraftSourceOriginDTO = 'drop' | 'paste';
export type AttachmentDraftSourceOriginUrl = 'input-link' | 'clipboard-read' | AttachmentDraftSourceOriginDTO;
export type AttachmentCreationOptions = {
/** Also attach an image representation of the attachment. Requires Release.Features.ENABLE_TEXT_AND_IMAGES as well. */
hintAddImages?: boolean;
}
export type AttachmentDraftSourceOriginCloud = `picker-${AttachmentCloudProviderId}`;
// 1. draft input (loaded from the source)
@@ -135,9 +166,10 @@ export type AttachmentDraftConverter = {
}
export type AttachmentDraftConverterType =
| 'text' | 'rich-text' | 'rich-text-cleaner' | 'rich-text-table'
| 'text' | 'text-cleaner' | 'text-markdown'
| 'rich-text' | 'rich-text-cleaner' | 'rich-text-markdown' | 'rich-text-table'
| 'image-original' | 'image-resized-high' | 'image-resized-low' | 'image-ocr' | 'image-caption' | 'image-to-default'
| 'pdf-text' | 'pdf-images' | 'pdf-text-and-images'
| 'pdf-auto' | 'pdf-text' | 'pdf-images' | 'pdf-images-ocr' | 'pdf-text-and-images'
| 'docx-to-html'
| 'url-page-text' | 'url-page-markdown' | 'url-page-html' | 'url-page-null' | 'url-page-image'
| 'youtube-transcript' | 'youtube-transcript-simple'
@@ -0,0 +1,109 @@
import { default as TurndownService } from 'turndown';
// Cached Turndown service instance
let _turndownService: TurndownService | null = null;
function getTurndownService(): TurndownService {
if (!_turndownService) {
_turndownService = new TurndownService({
headingStyle: 'atx',
codeBlockStyle: 'fenced',
emDelimiter: '_',
});
// Remove script and style elements
_turndownService.remove(['script', 'style', 'noscript']);
}
return _turndownService;
}
/**
* Convert HTML string to Markdown using Turndown.
* Performs basic HTML cleaning before conversion.
*/
export function convertHtmlToMarkdown(html: string): string {
// Basic client-side HTML cleaning using DOMParser
const cleanedHtml = cleanHtmlForMarkdown(html);
return getTurndownService().turndown(cleanedHtml);
}
/**
* Client-side HTML cleaning optimized for Markdown conversion.
* Uses DOMParser (browser-native) instead of Cheerio (server-only).
*/
function cleanHtmlForMarkdown(html: string): string {
try {
const parser = new DOMParser();
const doc = parser.parseFromString(html, 'text/html');
// Remove unwanted elements
const unwantedSelectors = [
'script', 'style', 'link', 'noscript', 'iframe', 'svg', 'canvas',
'nav:not(main nav)', 'aside', 'footer:not(article footer)',
'.ad', '.ads', '.advertisement', '.banner', '.popup', '.modal', '.overlay',
'.cookie-banner', '.newsletter-signup', '.social-share', '.comments',
'.sidebar', '.widget', '.carousel', '.slider',
'[aria-hidden="true"]', '[hidden]',
'[data-analytics]', '[data-tracking]', '[data-gtm]',
];
for (const selector of unwantedSelectors) {
try {
doc.querySelectorAll(selector).forEach(el => el.remove());
} catch {
// Skip invalid selectors (e.g., complex :not() selectors may fail in some browsers)
}
}
// Remove hidden elements via inline styles
doc.querySelectorAll('[style]').forEach(el => {
const style = el.getAttribute('style') || '';
if (style.includes('display: none') || style.includes('display:none') ||
style.includes('visibility: hidden') || style.includes('visibility:hidden'))
el.remove();
});
// Clean up anchor hrefs (remove tracking parameters)
doc.querySelectorAll('a[href]').forEach(el => {
const href = el.getAttribute('href');
if (!href) return;
// Remove javascript: links
if (href.toLowerCase().startsWith('javascript:')) {
el.removeAttribute('href');
return;
}
// Remove tracking parameters
if (href.includes('?')) {
try {
const url = new URL(href, 'http://placeholder');
const cleanParams = new URLSearchParams();
url.searchParams.forEach((value, key) => {
if (!key.match(/^(utm_|fbclid|gclid|msclkid)/i))
cleanParams.append(key, value);
});
const cleanHref = `${url.pathname}${cleanParams.toString() ? '?' + cleanParams.toString() : ''}${url.hash}`;
el.setAttribute('href', cleanHref);
} catch {
// Keep original href if URL parsing fails
}
}
});
// Remove comments (HTML comment nodes)
const walker = document.createTreeWalker(doc.body, NodeFilter.SHOW_COMMENT);
const comments: Comment[] = [];
while (walker.nextNode())
comments.push(walker.currentNode as Comment);
comments.forEach(comment => comment.remove());
return doc.body.innerHTML;
} catch (error) {
console.error('HTML cleaning error:', error);
return html; // Return original if cleaning fails
}
}
@@ -12,7 +12,7 @@ import type { DMessageId } from '~/common/stores/chat/chat.message';
import { getAllFilesFromDirectoryRecursively, getDataTransferFilesOrPromises } from '~/common/util/fileSystemUtils';
import { useChatAttachmentsStore } from '~/common/chat-overlay/store-perchat_vanilla';
import type { AttachmentDraftSourceOriginDTO, AttachmentDraftSourceOriginFile, AttachmentDraftSourceOriginUrl } from './attachment.types';
import type { AttachmentDraftSource, AttachmentDraftSourceOriginDTO, AttachmentDraftSourceOriginFile, AttachmentDraftSourceOriginUrl } from './attachment.types';
import type { AttachmentDraftsStoreApi } from './store-attachment-drafts_slice';
@@ -26,6 +26,9 @@ function notifyOnlyImages(item: any) {
}
export type AttachmentStoreCloudInput = Omit<Extract<AttachmentDraftSource, { media: 'cloud' }>, 'media' | 'origin'>;
/**
* @param attachmentsStoreApi A Per-Chat or standalone Attachment Drafts store.
* @param enableLoadURLsOnPaste Only used if invoking attachAppendDataTransfer or attachAppendClipboardItems.
@@ -254,8 +257,8 @@ export function useAttachmentDrafts(attachmentsStoreApi: AttachmentDraftsStoreAp
// https://github.com/enricoros/big-AGI/issues/286
const textHtml = clipboardItem.types.includes('text/html')
? await clipboardItem.getType('text/html')
.then(blob => blob?.text() ?? '')
.catch(() => '')
.then(blob => blob?.text() ?? '')
.catch(() => '')
: '';
const heuristicBypassImage = textHtml.startsWith('<table ');
@@ -289,8 +292,8 @@ export function useAttachmentDrafts(attachmentsStoreApi: AttachmentDraftsStoreAp
// get the Plain text
const textPlain = clipboardItem.types.includes('text/plain')
? await clipboardItem.getType('text/plain')
.then(blob => blob?.text() ?? '')
.catch(() => '')
.then(blob => blob?.text() ?? '')
.catch(() => '')
: '';
// attach as URL
@@ -321,6 +324,27 @@ export function useAttachmentDrafts(attachmentsStoreApi: AttachmentDraftsStoreAp
}
}, [_createAttachmentDraft, attachAppendFile, attachAppendUrl, enableLoadURLsOnPaste, filterOnlyImages, hintAddImages]);
/**
* Append a cloud file (Google Drive, OneDrive, etc.) to the attachments.
* This is the entry point for cloud file picker integrations.
*/
const attachAppendCloudFile = React.useCallback((cloudFile: AttachmentStoreCloudInput) => {
if (ATTACHMENTS_DEBUG_INTAKE)
console.log('attachAppendCloudFile', cloudFile);
// only-images: ignore cloud files as they may not be images
if (filterOnlyImages && !cloudFile.mimeType.startsWith('image/')) {
notifyOnlyImages(cloudFile);
return Promise.resolve();
}
return _createAttachmentDraft({
media: 'cloud',
origin: `picker-${cloudFile.provider}`,
...cloudFile,
}, { hintAddImages });
}, [_createAttachmentDraft, filterOnlyImages, hintAddImages]);
/**
* Append ego content to the attachments.
*/
@@ -348,6 +372,7 @@ export function useAttachmentDrafts(attachmentsStoreApi: AttachmentDraftsStoreAp
// create drafts
attachAppendClipboardItems,
attachAppendCloudFile,
attachAppendDataTransfer,
attachAppendEgoFragments,
attachAppendFile,
@@ -0,0 +1,219 @@
import * as React from 'react';
import { createPortal } from 'react-dom';
import type { PickerCanceledEvent, PickerPickedEvent } from '@googleworkspace/drive-picker-element';
import { DrivePicker, DrivePickerDocsView } from '@googleworkspace/drive-picker-react';
import { IconButton } from '@mui/joy';
import LogoutIcon from '@mui/icons-material/Logout';
import { TooltipOutlined } from '~/common/components/TooltipOutlined';
import { addSnackbar } from '~/common/components/snackbar/useSnackbarsStore';
import type { AttachmentStoreCloudInput } from './useAttachmentDrafts';
// configuration
const GOOGLE_DRIVE_CLIENT_ID = process.env.NEXT_PUBLIC_GOOGLE_DRIVE_CLIENT_ID || '';
const MAX_FILE_SIZE_MB = 10; // skip files larger than this; 0 = no limit; note: Google Workspace files report 0 bytes
const MAX_PICKER_FILES = 8; // max files per picker session; 0 = unlimited
export const hasGoogleDriveCapability = !!GOOGLE_DRIVE_CLIENT_ID;
// -- Token Definitions --
export interface ICloudProviderTokenAccessor {
get: () => CloudProviderToken | null;
set: (value: CloudProviderToken | null) => void;
}
export interface CloudProviderToken {
token: string;
expiresAt?: number; // timestamp in ms; if missing, token is treated as valid (the downstream may clear it eventually)
}
function _getUnexpiredToken(stored: CloudProviderToken | null): string | undefined {
if (!stored?.token) return undefined;
// if expiresAt is set and expired (with 60s safety margin), return undefined
if (stored.expiresAt && Date.now() > stored.expiresAt - 60 * 1000) return undefined;
return stored.token;
}
// --- In-memory token storage ---
let _inMemoryToken: CloudProviderToken | null = null;
const _inMemoryTokenStorage: ICloudProviderTokenAccessor = {
get: () => _inMemoryToken,
set: (value: CloudProviderToken | null) => _inMemoryToken = value,
};
type _OauthResponseEvent = {
detail?: {
access_token: string; // xxxx.yyyyy....
expires_in?: string | number; // 3599
// scope?: string; // 'https://www.googleapis.com/auth/drive.file'
// token_type?: string; // 'Bearer'
};
}
type _OauthErrorEvent = {
detail?: {
error?: string; // 'access_denied', 'popup_closed_by_user', ...
} | {
type?: string; // 'popup_closed'
// message?: string; // 'Popup window closed'
// stack?: string;
} | object;
}
export function useGoogleDrivePicker(
onCloudFileSelected: (cloudFile: AttachmentStoreCloudInput) => void,
isMobile: boolean,
tokenStorage: ICloudProviderTokenAccessor = _inMemoryTokenStorage,
loginHint?: string,
) {
// state
const [isPickerOpen, setIsPickerOpen] = React.useState(false);
const openGoogleDrivePicker = React.useCallback(() => setIsPickerOpen(true), []);
const handleDeauthClick = React.useCallback(() => {
setIsPickerOpen(false);
tokenStorage.set(null);
}, [tokenStorage]);
// handle oauth events, to store the token for the picker callback
const handleOAuthResponse = React.useCallback((e: _OauthResponseEvent) => {
if (!e.detail?.access_token) return;
const expiresIn = typeof e.detail.expires_in === 'number' ? e.detail.expires_in : typeof e.detail.expires_in === 'string' ? parseInt(e.detail.expires_in, 10) : undefined;
tokenStorage.set({
token: e.detail.access_token,
expiresAt: expiresIn === undefined ? undefined : Date.now() + expiresIn * 1000,
});
}, [tokenStorage]);
const handleOAuthError = React.useCallback((e: _OauthErrorEvent) => {
setIsPickerOpen(false);
// ignore if user closed the popup
if (e?.detail && 'type' in e?.detail && e.detail.type === 'popup_closed') return;
const errorMsg = e?.detail && 'error' in e?.detail && typeof e.detail.error === 'string' ? e.detail.error : undefined;
addSnackbar({ key: 'gdrive-oauth-error', message: errorMsg === 'access_denied' ? 'Drive file access was denied' : 'Google Drive authentication failed.', type: 'issue' });
}, []);
// handler picker events
const handleCanceled = React.useCallback((_e: PickerCanceledEvent) => setIsPickerOpen(false), []);
const handlePicked = React.useCallback((e: PickerPickedEvent) => {
setIsPickerOpen(false);
const docs = e.detail?.docs;
if (!docs?.length) return;
// read token, just set by handleOAuthResponse
const currentToken = _getUnexpiredToken(tokenStorage.get());
if (!currentToken)
return addSnackbar({ key: 'gdrive-no-token', message: 'Unable to access Google Drive.', type: 'issue' });
// convert picker docs to cloud file metadata for the attachment system
const maxBytes = MAX_FILE_SIZE_MB * 1024 * 1024;
const skippedFiles: string[] = [];
for (const doc of docs) {
// skip files that are too large (note: Google Workspace files report 0 bytes)
if (MAX_FILE_SIZE_MB && doc.sizeBytes && doc.sizeBytes > maxBytes) {
skippedFiles.push(doc.name);
continue;
}
onCloudFileSelected({
accessToken: currentToken,
provider: 'gdrive',
fileId: doc.id,
mimeType: doc.mimeType,
fileName: doc.name,
fileSize: doc.sizeBytes,
webViewLink: doc.url,
});
}
if (skippedFiles.length)
addSnackbar({ key: 'gdrive-size-limit', message: `Skipped ${skippedFiles.length} file(s) over ${MAX_FILE_SIZE_MB} MB: ${skippedFiles.join(', ')}`, type: 'issue' });
}, [onCloudFileSelected, tokenStorage]);
// memo components (close button and picker) | null
const googleDrivePickerComponent = React.useMemo(() => !isPickerOpen || !GOOGLE_DRIVE_CLIENT_ID ? null : <>
{/* Top-level close button - portaled to body, above the Google Drive picker */}
{createPortal(
<TooltipOutlined title='Close and Switch Google Drive Account' placement='bottom'>
<IconButton
onClick={handleDeauthClick}
sx={{
'--IconButton-size': '2.75rem',
backgroundColor: 'background.popup',
borderRadius: '50%',
boxShadow: 'lg',
position: 'fixed',
top: '1rem',
left: '50%',
transform: 'translateX(-50%)',
zIndex: 2002, // above the Drive Picker (2001+)
}}
>
<LogoutIcon />
</IconButton>
</TooltipOutlined>,
document.body,
)}
<DrivePicker
app-id={GOOGLE_DRIVE_CLIENT_ID.split('-')[0] || ''}
client-id={GOOGLE_DRIVE_CLIENT_ID}
title='Attach files from Google Drive'
multiselect={true}
hide-title-bar='true'
// nav-hidden={true /* disables the 'Google Drive' nav */}
// mine-only={true}
login-hint={loginHint}
max-items={MAX_PICKER_FILES || undefined}
oauth-token={_getUnexpiredToken(tokenStorage.get())}
onOauthResponse={handleOAuthResponse}
onOauthError={handleOAuthError}
onPicked={handlePicked}
onCanceled={handleCanceled}
>
<DrivePickerDocsView
// file-ids='id1,id2,id3'
// include-folders='default'
// mime-types=
mode={isMobile ? 'LIST' : undefined /* LIST, GRID - if set hides the switch */}
// owned-by-me='default'
// select-folder-enabled='default' // does not work, while the one in DrivePicker does
// starred=
/>
</DrivePicker>
</>, [handleCanceled, handleDeauthClick, handleOAuthError, handleOAuthResponse, handlePicked, isMobile, isPickerOpen, loginHint, tokenStorage]);
return {
openGoogleDrivePicker,
googleDrivePickerComponent,
};
}
@@ -20,7 +20,7 @@ const BoxCollapsee = styled(Box)({
export function ExpanderControlledBox(props: { expanded: boolean, children: React.ReactNode, sx?: SxProps }) {
return (
<BoxCollapser aria-expanded={props.expanded} sx={props.sx}>
<BoxCollapser aria-expanded={props.expanded} data-agi-no-copy={!props.expanded || undefined} sx={props.sx}>
<BoxCollapsee>
{props.children}
</BoxCollapsee>
+38
View File
@@ -0,0 +1,38 @@
import * as React from 'react';
import { Chip, ChipProps } from '@mui/joy';
/**
* Simple badge/label component for inline status indicators like "New", "Beta", etc.
*/
export function GoodBadge(props: {
badge: React.ReactNode;
color?: ChipProps['color'];
variant?: ChipProps['variant'];
sx?: ChipProps['sx'];
}) {
return (
<Chip
size='sm'
color={props.color ?? 'success'}
variant={props.variant ?? 'soft'}
sx={{
ml: 1.5,
fontSize: 'xs',
fontWeight: 'md',
borderRadius: 'xs',
px: 1,
py: 0.25,
// default "new" color - lime/yellow-green
...(props.color === undefined && {
bgcolor: '#d5ec31',
color: 'primary.softColor',
}),
...props.sx,
}}
>
{props.badge}
</Chip>
);
}
+2
View File
@@ -40,6 +40,8 @@ export function InlineTextarea(props: {
const handleEditKeyDown = (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
if (e.key === 'Enter') {
if (e.nativeEvent.isComposing)
return;
const shiftOrAlt = e.shiftKey || e.altKey;
if (enterIsNewline ? shiftOrAlt : !shiftOrAlt) {
e.preventDefault();
+22 -8
View File
@@ -1,6 +1,6 @@
import * as React from 'react';
import { Option, Select } from '@mui/joy';
import { Option, optionClasses, Select, SelectSlotsAndSlotProps } from '@mui/joy';
import KeyboardArrowDownIcon from '@mui/icons-material/KeyboardArrowDown';
import { useUIPreferencesStore } from '~/common/stores/store-ui';
@@ -10,6 +10,20 @@ import { useUIPreferencesStore } from '~/common/stores/store-ui';
import languages from './Languages.json';
// copied from useLLMSelect.tsx - inspired by optimaSelectSlotProps.listbox
const _selectSlotProps: SelectSlotsAndSlotProps<false>['slotProps'] = {
root: { sx: { minWidth: 200 } },
listbox: {
sx: {
boxShadow: 'xl',
[`& .${optionClasses.root}`]: {
maxWidth: 'min(640px, calc(100dvw - 0.25rem))',
},
},
} as const,
} as const;
export function LanguageSelect() {
// external state
@@ -32,19 +46,19 @@ export function LanguageSelect() {
</Option>
) : (
Object.entries(localesOrCode).map(([country, code]) => (
<Option key={code} value={code}>
<Option key={code} value={code} label={language}>
{`${language} (${country})`}
</Option>
))
)), []);
return (
<Select value={preferredLanguage} onChange={handleLanguageChanged}
indicator={<KeyboardArrowDownIcon />}
slotProps={{
root: { sx: { minWidth: 200 } },
indicator: { sx: { opacity: 0.5 } },
}}>
<Select
value={preferredLanguage}
onChange={handleLanguageChanged}
indicator={<KeyboardArrowDownIcon />}
slotProps={_selectSlotProps}
>
{languageOptions}
</Select>
);
@@ -21,6 +21,13 @@ const _styles = {
gap: 1,
} as const,
chipGroupEnd: {
display: 'flex',
flexWrap: 'wrap',
justifyContent: 'flex-end',
gap: 1,
} as const,
chip: {
'--Chip-minHeight': '1.75rem', // this makes it prob better
px: 1.5,
@@ -36,6 +43,7 @@ export const FormChipControl = <TValue extends string>(props: {
// specific
size?: 'sm' | 'md' | 'lg',
color?: ColorPaletteProp,
alignEnd?: boolean,
// =FormRadioControl
title: string | React.JSX.Element;
description?: string | React.JSX.Element;
@@ -48,6 +56,9 @@ export const FormChipControl = <TValue extends string>(props: {
const { onChange } = props;
const selectedOption = props.options.find(option => option.value === props.value);
const description = selectedOption?.description ?? props.description;
const handleChipClick = React.useCallback((value: Immutable<TValue>) => {
if (!props.disabled)
onChange(value);
@@ -55,8 +66,8 @@ export const FormChipControl = <TValue extends string>(props: {
return (
<FormControl orientation='horizontal' disabled={props.disabled} sx={_styles.control}>
{(!!props.title || !!props.description) && <FormLabelStart title={props.title} description={props.description} tooltip={props.tooltip} />}
<Box sx={_styles.chipGroup}>
{(!!props.title || !!description) && <FormLabelStart title={props.title} description={description} tooltip={props.tooltip} />}
<Box sx={props.alignEnd ? _styles.chipGroupEnd : _styles.chipGroup}>
{props.options.map((option) => (
<Chip
key={'opt-' + option.value}
@@ -0,0 +1,95 @@
import * as React from 'react';
import type { SxProps } from '@mui/joy/styles/types';
import { FormControl, IconButton, Input } from '@mui/joy';
import KeyIcon from '@mui/icons-material/Key';
import VisibilityIcon from '@mui/icons-material/Visibility';
import VisibilityOffIcon from '@mui/icons-material/VisibilityOff';
import { FormLabelStart } from './FormLabelStart';
const _styles = {
formControl: {
flexWrap: 'wrap',
justifyContent: 'space-between',
alignItems: 'center',
},
inputDefault: {
flexGrow: 1,
},
} as const satisfies Record<string, SxProps>;
/**
* Secret/API key form field with visibility toggle.
* Same inline layout as FormTextField but with secret-specific features:
* - Password masking with visibility toggle
* - Key icon (customizable)
* - Password manager integration
*/
export function FormSecretField(props: {
autoCompleteId: string;
title: string | React.JSX.Element;
description?: string | React.JSX.Element;
tooltip?: string | React.JSX.Element;
placeholder?: string;
value: string;
onChange: (text: string) => void;
// Behavior
required?: boolean;
disabled?: boolean;
isError?: boolean;
// Appearance
inputSx?: SxProps;
/** Custom start decorator, or false to hide. Default: KeyIcon */
startDecorator?: React.ReactNode | false;
}) {
// state
const [isVisible, setIsVisible] = React.useState(false);
// derived
const acId = 'secret-' + props.autoCompleteId;
// password manager username
const ghost = props.autoCompleteId.replace(/-key$/, '').replace(/-/g, ' ');
const endDecorator = React.useMemo(() => !!props.value && (
<IconButton size='sm' onClick={() => setIsVisible(on => !on)}>
{isVisible ? <VisibilityIcon sx={{ fontSize: 'md' }} /> : <VisibilityOffIcon sx={{ fontSize: 'md' }} />}
</IconButton>
), [props.value, isVisible]);
return (
<FormControl
id={acId}
orientation='horizontal'
disabled={props.disabled}
sx={_styles.formControl}
>
<FormLabelStart title={props.title} description={props.description} tooltip={props.tooltip} />
{/* Hidden username field for password manager association */}
<input
type='text'
autoComplete='username'
value={ghost}
readOnly
tabIndex={-1}
style={{ display: 'none' }}
/>
<Input
name={acId}
type={isVisible ? 'text' : 'password'}
autoComplete='new-password'
variant='outlined'
placeholder={props.required && !props.placeholder ? 'required' : props.placeholder}
error={props.isError}
value={props.value}
onChange={event => props.onChange(event.target.value)}
startDecorator={props.startDecorator ?? <KeyIcon sx={{ fontSize: 'md' }} />}
endDecorator={endDecorator}
sx={props.inputSx ?? _styles.inputDefault}
/>
</FormControl>
);
}
@@ -24,6 +24,7 @@ export function FormSliderControl(props: {
startAdornment?: React.ReactNode,
endAdornment?: React.ReactNode,
styleNoTrack?: boolean,
sliderSx?: SxProps,
}) {
@@ -66,8 +67,7 @@ export function FormSliderControl(props: {
onChange={handleChange}
onChangeCommitted={handleChangeCommitted}
valueLabelDisplay={props.valueLabelDisplay}
sx={props.styleNoTrack ? _styleNoTrack : undefined}
// sx={{ py: 1, mt: 1.1 }}
sx={props.styleNoTrack ? _styleNoTrack : props.sliderSx}
/>
{props.endAdornment}
</FormControl>
+15 -7
View File
@@ -6,11 +6,16 @@ import { FormControl, Input } from '@mui/joy';
import { FormLabelStart } from './FormLabelStart';
const formControlSx: SxProps = {
flexWrap: 'wrap',
justifyContent: 'space-between',
alignItems: 'center',
};
const _styles = {
formControl: {
flexWrap: 'wrap',
justifyContent: 'space-between',
alignItems: 'center',
},
inputDefault: {
flexGrow: 1,
},
} as const satisfies Record<string, SxProps>;
/**
@@ -23,6 +28,8 @@ export function FormTextField(props: {
tooltip?: string | React.JSX.Element,
placeholder?: string, isError?: boolean, disabled?: boolean,
value: string | undefined, onChange: (text: string) => void,
endDecorator?: React.ReactNode,
inputSx?: SxProps,
}) {
const acId = 'text-' + props.autoCompleteId;
return (
@@ -30,7 +37,7 @@ export function FormTextField(props: {
id={acId}
orientation='horizontal'
disabled={props.disabled}
sx={formControlSx}
sx={_styles.formControl}
>
<FormLabelStart title={props.title} description={props.description} tooltip={props.tooltip} />
<Input
@@ -39,7 +46,8 @@ export function FormTextField(props: {
autoComplete='off'
variant='outlined' placeholder={props.placeholder} error={props.isError}
value={props.value} onChange={event => props.onChange(event.target.value)}
sx={{ flexGrow: 1 }}
endDecorator={props.endDecorator}
sx={props.inputSx ?? _styles.inputDefault}
/>
</FormControl>
);
@@ -1,11 +1,14 @@
import * as React from 'react';
import { FormSwitchControl } from './FormSwitchControl';
import { Box, FormControl, Switch, Tooltip } from '@mui/joy';
import { FormLabelStart } from './FormLabelStart';
/**
* Reusable toggle for enabling client-side API fetch.
* Appears with animation when client key is present.
* Shows a tooltip recommendation when local host is detected but CSF is off.
*/
export function SetupFormClientSideToggle(props: {
visible: boolean;
@@ -13,8 +16,12 @@ export function SetupFormClientSideToggle(props: {
onChange: (on: boolean) => void;
helpText: string;
disabled?: boolean;
localHostDetected?: boolean; // shows a tooltip to hint at using this
}) {
// show recommendation tooltip for local hosts when CSF is off
const showLocalRecommendation = !!props.localHostDetected && !props.checked;
return (
<div
style={{
@@ -24,14 +31,29 @@ export function SetupFormClientSideToggle(props: {
}}
>
<div style={{ overflow: 'hidden' }}>
<FormSwitchControl
title='Direct Connection'
description={props.checked ? 'Connect from browser' : 'Via server (default)'}
tooltip={props.helpText}
checked={props.checked}
onChange={props.onChange}
disabled={props.disabled}
/>
<FormControl orientation='horizontal' disabled={props.disabled} sx={{ flexWrap: 'wrap', justifyContent: 'space-between', alignItems: 'center' }}>
<FormLabelStart
title='Direct Connection'
description={props.checked ? 'Connect from browser' : 'Via server (default)'}
tooltip={showLocalRecommendation ? undefined : props.helpText}
/>
<Tooltip
open={showLocalRecommendation}
disableInteractive
arrow
variant='solid'
color='success'
placement='top-end'
title='Recommended ON for local services'
>
<Switch
checked={props.checked}
onChange={event => props.onChange(event.target.checked)}
endDecorator={props.checked ? 'On' : 'Off'}
slotProps={{ endDecorator: { sx: { minWidth: 26 } } }}
/>
</Tooltip>
</FormControl>
</div>
</div>
);
@@ -1,7 +1,7 @@
import * as React from 'react';
import { Box, Button, FormLabel } from '@mui/joy';
import SyncIcon from '@mui/icons-material/Sync';
import RefreshIcon from '@mui/icons-material/Refresh';
import type { ToggleableBoolean } from '~/common/util/hooks/useToggleableBoolean';
@@ -32,7 +32,7 @@ export function SetupFormRefetchButton(props: {
color={props.error ? 'warning' : 'primary'}
disabled={props.disabled}
loading={props.loading}
endDecorator={<SyncIcon />}
endDecorator={<RefreshIcon />}
onClick={props.refetch}
sx={{ minWidth: 120, ml: 'auto' }}
>
+86 -24
View File
@@ -1,21 +1,23 @@
import * as React from 'react';
import type { SxProps } from '@mui/joy/styles/types';
import { Box, Chip, ColorPaletteProp, FormControl, IconButton, ListDivider, ListItem, ListItemButton, ListItemDecorator, Option, Select, SelectSlotsAndSlotProps, SvgIconProps, VariantProp, optionClasses } from '@mui/joy';
import { Chip, ColorPaletteProp, FormControl, IconButton, ListDivider, ListItem, ListItemButton, ListItemDecorator, Option, optionClasses, Select, SelectSlotsAndSlotProps, SvgIconProps, VariantProp } from '@mui/joy';
import ArrowForwardRoundedIcon from '@mui/icons-material/ArrowForwardRounded';
import AutoModeIcon from '@mui/icons-material/AutoMode';
import BuildCircleIcon from '@mui/icons-material/BuildCircle';
import ExpandLessIcon from '@mui/icons-material/ExpandLess';
import ExpandMoreIcon from '@mui/icons-material/ExpandMore';
import type { IModelVendor } from '~/modules/llms/vendors/IModelVendor';
import { findModelVendor } from '~/modules/llms/vendors/vendors.registry';
import { llmsGetVendorIcon, LLMVendorIcon } from '~/modules/llms/components/LLMVendorIcon';
import type { DModelDomainId } from '~/common/stores/llms/model.domains.types';
import type { DModelsServiceId } from '~/common/stores/llms/llms.service.types';
import { DLLM, DLLMId, getLLMPricing, LLM_IF_OAI_Reasoning, LLM_IF_Outputs_Audio, LLM_IF_Outputs_Image, LLM_IF_Tools_WebSearch } from '~/common/stores/llms/llms.types';
import { PhGearSixIcon } from '~/common/components/icons/phosphor/PhGearSixIcon';
import { StarIconUnstyled, StarredNoXL2 } from '~/common/components/StarIcons';
import { TooltipOutlined } from '~/common/components/TooltipOutlined';
import { getChatLLMId, llmsStoreActions } from '~/common/stores/llms/store-llms';
import { findModelsServiceOrNull, getChatLLMId, llmsStoreActions } from '~/common/stores/llms/store-llms';
import { optimaActions, optimaOpenModels } from '~/common/layout/optima/useOptima';
import { useUIPreferencesStore } from '~/common/stores/store-ui';
import { useVisibleLLMs } from '~/common/stores/llms/llms.hooks';
@@ -69,12 +71,13 @@ const _styles = {
backgroundColor: 'background.surface',
zIndex: 1,
},
listVendor: {
// see OptimaBarDropdown's _styles.separator
listServiceHeaderButton: {
fontSize: 'sm',
color: 'text.tertiary',
textAlign: 'center',
my: 0.75,
fontWeight: 'md',
justifyContent: 'space-between',
},
listServiceHeaderExpand: {
fontSize: 'md',
},
listConfSep: {
mb: 0,
@@ -84,7 +87,7 @@ const _styles = {
},
} as const satisfies Record<string, SxProps>;
const _slotProps: SelectSlotsAndSlotProps<false>['slotProps'] = {
const _slotProps = {
// see the OptimaBarDropdown.listbox for a well made customization (max-height, max-width, etc.)
listbox: {
sx: {
@@ -127,7 +130,7 @@ const _slotProps: SelectSlotsAndSlotProps<false>['slotProps'] = {
minWidth: '6rem',
} as const,
} as const,
} as const;
} as const satisfies SelectSlotsAndSlotProps<false>['slotProps'];
interface LLMSelectOptions {
@@ -162,6 +165,7 @@ export function useLLMSelect(
// state
const [controlledOpen, setControlledOpen] = React.useState(false);
const [collapsedServices, setCollapsedServices] = React.useState<Set<DModelsServiceId>>(new Set());
// external state
const starredOnly = useUIPreferencesStore(state => showStarFilter && state.showModelsStarredOnly);
@@ -174,27 +178,83 @@ export function useLLMSelect(
const isReasoning = !LLM_SELECT_SHOW_REASONING_ICON ? false : llm?.interfaces?.includes(LLM_IF_OAI_Reasoning) ?? false;
// handlers
const toggleServiceCollapse = React.useCallback((serviceId: DModelsServiceId) => {
setCollapsedServices(prev => {
const next = new Set(prev);
if (next.has(serviceId)) next.delete(serviceId);
else next.add(serviceId);
return next;
});
}, []);
// Scroll preservation: MUI's useSelect auto-scrolls to highlighted item when options change - we want to preserve scroll instead
const listboxRef = React.useRef<HTMLUListElement>(null);
const listboxSlotPropsStable = React.useMemo(() => ({
..._slotProps,
listbox: { ..._slotProps.listbox, ref: listboxRef },
}), []);
React.useLayoutEffect(() => {
// restore scroll after collapse/expand - snapshot before MUI scrolls, restore via double RAF
const el = listboxRef.current;
if (!el) return;
const scrollTop = el.scrollTop;
const raf = requestAnimationFrame(() => {
// usually works, especially on expansion
el.scrollTop = scrollTop;
return requestAnimationFrame(() => {
// fixes the collapse too
el.scrollTop = scrollTop;
});
});
return () => cancelAnimationFrame(raf);
}, [collapsedServices]);
// memo LLM Options
const optimizeToSingleVisibleId = (!controlledOpen && _filteredLLMs.length > LLM_SELECT_REDUCE_OPTIONS) ? llmId : null; // id to keep visible when optimizing
const optionsArray = React.useMemo(() => {
// check if we have multiple services (to show collapsible headers)
const hasMultipleServices = _filteredLLMs.some((llm, i, arr) => i > 0 && llm.sId !== arr[i - 1].sId);
// create the option items
let formerVendor: IModelVendor | null = null;
let prevServiceId: DModelsServiceId | null = null;
return _filteredLLMs.reduce((acc, llm, _index) => {
if (optimizeToSingleVisibleId && llm.id !== optimizeToSingleVisibleId)
return acc;
const vendor = findModelVendor(llm.vId);
const vendorChanged = vendor !== formerVendor;
if (vendorChanged)
formerVendor = vendor;
const serviceVendor = findModelVendor(llm.vId);
const isServiceCollapsed = hasMultipleServices && collapsedServices.has(llm.sId);
// add separators if the vendor changed (and more than one vendor)
const addSeparator = vendorChanged && formerVendor !== null;
if (addSeparator && !optimizeToSingleVisibleId)
acc.push(<Box key={'llm-sep-' + llm.id} sx={_styles.listVendor}>{vendor?.name}</Box>);
// add collapsible service headers when changing services
if (hasMultipleServices && llm.sId !== prevServiceId) {
if (!optimizeToSingleVisibleId) {
const serviceLabel = findModelsServiceOrNull(llm.sId)?.label || serviceVendor?.name || llm.sId;
acc.push(
<ListItem key={'llm-sep-' + llm.sId}>
<ListItemButton onClick={() => toggleServiceCollapse(llm.sId)} sx={_styles.listServiceHeaderButton}>
{/*{serviceVendor?.id && <ListItemDecorator><LLMVendorIcon vendorId={serviceVendor.id} /></ListItemDecorator>}*/}
<div />
{isServiceCollapsed ? <i>{serviceLabel}</i> : serviceLabel}
{isServiceCollapsed ? <ExpandMoreIcon sx={_styles.listServiceHeaderExpand} /> : <ExpandLessIcon sx={_styles.listServiceHeaderExpand} />}
</ListItemButton>
</ListItem>,
);
}
prevServiceId = llm.sId;
}
// skip models if service is collapsed (but always show selected model)
if (isServiceCollapsed && llm.id !== llmId)
return acc;
let features = '';
const isNotSymlink = !llm.label.startsWith('🔗');
@@ -202,6 +262,8 @@ export function useLLMSelect(
if (isNotSymlink) {
// check features
if (seemsFree) features += 'free ';
if (llm.isUserClone)
features += ' '; // is clone
if (llm.interfaces.includes(LLM_IF_OAI_Reasoning))
features += '🧠 '; // can reason
if (llm.interfaces.includes(LLM_IF_Tools_WebSearch))
@@ -225,7 +287,7 @@ export function useLLMSelect(
>
{!noIcons && (
<ListItemDecorator>
{(llm.userStarred && !starredOnly) ? <StarredNoXL2 /> : vendor?.id ? <LLMVendorIcon vendorId={vendor.id} /> : null}
{(llm.userStarred && !starredOnly) ? <StarredNoXL2 /> : serviceVendor?.id ? <LLMVendorIcon vendorId={serviceVendor.id} /> : null}
</ListItemDecorator>
)}
{/*<Tooltip title={llm.description}>*/}
@@ -244,7 +306,7 @@ export function useLLMSelect(
// variant='outlined'
onClick={(e) => {
e.stopPropagation();
optimaActions().openModelOptions(llm.id);
optimaActions().openModelOptions(llm.id, 'parameters');
}}
sx={_styles.configButton}
>
@@ -260,7 +322,7 @@ export function useLLMSelect(
return acc;
}, [] as React.JSX.Element[]);
}, [_filteredLLMs, llmId, noIcons, optimizeToSingleVisibleId, starredOnly]);
}, [_filteredLLMs, collapsedServices, llmId, noIcons, optimizeToSingleVisibleId, starredOnly, toggleServiceCollapse]);
const onSelectChange = React.useCallback((_event: unknown, value: DLLMId | null) => {
@@ -289,7 +351,7 @@ export function useLLMSelect(
listboxOpen={controlledOpen}
onListboxOpenChange={hasNoModels ? optimaOpenModels : setControlledOpen}
placeholder={hasNoModels ? LLM_TEXT_CONFIGURE : placeholder}
slotProps={_slotProps}
slotProps={listboxSlotPropsStable}
endDecorator={autoRefreshDomain ?
<TooltipOutlined title='Auto-select the model'>
<IconButton onClick={() => llmsStoreActions().assignDomainModelId(autoRefreshDomain, null)}>
@@ -332,7 +394,7 @@ export function useLLMSelect(
</Select>
{/*</Box>*/}
</FormControl>
), [appendConfigureModels, autoRefreshDomain, controlledOpen, disabled, hasNoModels, hasStarred, isHorizontal, isReasoning, label, larger, llmId, onSelectChange, optimizeToSingleVisibleId, options.color, options.sx, options.variant, optionsArray, placeholder, showNoOptions, showStarFilter, starredOnly]);
), [appendConfigureModels, autoRefreshDomain, controlledOpen, disabled, hasNoModels, hasStarred, isHorizontal, isReasoning, label, larger, listboxSlotPropsStable, llmId, onSelectChange, optimizeToSingleVisibleId, options.color, options.sx, options.variant, optionsArray, placeholder, showNoOptions, showStarFilter, starredOnly]);
// Memo the vendor icon for the chat LLM
const chatLLMVendorIconFC = React.useMemo(() => {
@@ -0,0 +1,14 @@
import * as React from 'react';
import { SvgIcon, SvgIconProps } from '@mui/joy';
/*
* Source: 'https://phosphoricons.com/' - user-sound
*/
export function PhVoice(props: SvgIconProps) {
return (
<SvgIcon viewBox='0 0 256 256' stroke='none' fill='currentColor' width='24' height='24' {...props}>
<path d='M144,165.68a68,68,0,1,0-71.9,0c-20.65,6.76-39.23,19.39-54.17,37.17a8,8,0,0,0,12.25,10.3C50.25,189.19,77.91,176,108,176s57.75,13.19,77.88,37.15a8,8,0,1,0,12.25-10.3C183.18,185.07,164.6,172.44,144,165.68ZM56,108a52,52,0,1,1,52,52A52.06,52.06,0,0,1,56,108ZM207.36,65.6a108.36,108.36,0,0,1,0,84.8,8,8,0,0,1-7.36,4.86,8,8,0,0,1-7.36-11.15,92.26,92.26,0,0,0,0-72.22,8,8,0,0,1,14.72-6.29ZM248,108a139,139,0,0,1-11.29,55.15,8,8,0,0,1-14.7-6.3,124.43,124.43,0,0,0,0-97.7,8,8,0,1,1,14.7-6.3A139,139,0,0,1,248,108Z' />
</SvgIcon>
);
}
@@ -0,0 +1,9 @@
import * as React from 'react';
import { SvgIcon, SvgIconProps } from '@mui/joy';
export function ElevenLabsIcon(props: SvgIconProps) {
return <SvgIcon viewBox='0 0 24 24' width='24' height='24' fill='currentColor' {...props}>
<path d='M7 4h3v16H7V4zm7 0h3v16h-3V4z' />
</SvgIcon>;
}
+10
View File
@@ -0,0 +1,10 @@
import * as React from 'react';
import { SvgIcon, SvgIconProps } from '@mui/joy';
export function InworldIcon(props: SvgIconProps) {
return <SvgIcon viewBox='0 0 141 181' width='24' height='24' fill='currentColor' {...props}>
<path d='M48.2616 34.7993C47.9981 34.8585 47.9766 34.6058 48.1379 34.5144C53.3155 31.4874 60.6866 30.767 61.235 30.353C61.7297 29.6594 57.864 29.697 56.5199 29.7024C46.74 30.0411 38.2989 33.681 31.288 40.2941C20.5242 49.4664 16.9973 64.5582 18.4866 78.1285C20.2285 92.1611 27.9384 105.344 39.385 113.635C63.2405 129.673 96.9242 122.011 114.774 100.183C136.753 73.537 130.194 35.6004 107.215 18.2613C101.252 14.1214 94.7359 10.8524 87.7035 8.98143C84.4238 8.00828 79.8592 7.07277 77.7624 7.00288C74.5472 6.90073 75.6171 9.54596 75.0472 10.1965C74.4773 10.8471 68.5901 11.1105 75.2515 13.4493C79.1925 14.8311 76.7946 14.9117 73.7891 17.6322C71.1278 20.0409 73.4504 22.7399 72.531 23.6861C68.5847 27.7508 74.6279 29.5358 77.7462 31.31C92.8542 39.3747 101.177 53.8913 98.4242 70.9831C94.5585 93.5644 60.9286 103.317 46.0464 85.462C34.5999 71.924 38.928 49.3858 55.2241 42.0361C57.4822 40.9447 62.5792 39.3532 60.3909 39.536C51.9552 40.2403 45.369 44.7297 43.2399 46.3588C43.0571 46.4986 42.8152 46.2835 42.9281 46.0846C44.4711 43.3695 53.4122 37.036 71.0041 34.8101C74.1494 34.1327 62.3157 31.6541 48.2616 34.7993Z' />
<path d='M55.4704 170.577C56.4274 170.566 57.3791 170.706 58.2716 170.797C60.5673 170.867 63.148 170.604 65.6696 170.932C71.1536 171.244 76.6807 172.475 82.1808 172.609C85.9712 172.889 89.5896 172.717 93.4607 172.862C96.3801 173.238 99.848 173.007 102.466 173.4C104.101 173.588 105.601 173.507 107.241 173.749C111.579 174.367 115.902 174.459 120.295 174.4C121.967 174.453 123.682 174.486 125.349 174.378C127.128 174.276 129.069 174.351 130.499 173.212C131.349 172.711 131.306 171.695 131.483 170.996C131.704 170.615 132.166 170.41 132.37 170.023C132.596 169.357 132.741 168.663 132.752 167.98V167.937C132.8 165.615 131.688 163.33 130.838 161.174C129.424 158.05 127.763 155.13 125.967 152.206C123.962 149.372 122.956 145.506 119.547 143.974C113.402 141.281 107.085 141.565 100.283 140.743C87.482 139.608 73.8849 138.775 61.9867 138.614C50.8843 138.775 31.2708 139.501 21.1738 140.399C3.2378 141.996 11.5875 144.84 11.023 149.872C10.8187 151.727 8.79172 151.576 9.01753 153.394C9.39388 156.028 14.3349 159.324 16.8188 159.991C27.76 162.916 13.2435 163.351 14.8672 166.276C16.2166 168.711 22.2383 168.012 24.6954 168.98L55.4597 170.577H55.4704Z' />
</SvgIcon>;
}
+1 -1
View File
@@ -186,7 +186,7 @@ export function GoodModal(props: {
},
display: 'flex',
flexWrap: 'wrap',
gap: 1,
// gap: { md: 1 }, // Note: let the startButton decide how to space itself
justifyContent: 'space-between',
}}>
{props.startButton}
-11
View File
@@ -25,17 +25,6 @@ export interface CapabilityBrowserSpeechRecognition {
export { browserSpeechRecognitionCapability as useCapabilityBrowserSpeechRecognition } from './speechrecognition/useSpeechRecognition';
/// Speech Synthesis: ElevenLabs
export interface CapabilityElevenLabsSpeechSynthesis {
mayWork: boolean;
isConfiguredServerSide: boolean;
isConfiguredClientSide: boolean;
}
export { useCapability as useCapabilityElevenLabs } from '~/modules/elevenlabs/elevenlabs.client';
/// Image Generation
export interface TextToImageProvider {
@@ -4,7 +4,7 @@ import { Alert, IconButton } from '@mui/joy';
import CloseRoundedIcon from '@mui/icons-material/CloseRounded';
import WarningRoundedIcon from '@mui/icons-material/WarningRounded';
import { Is, isBrowser, isPwa } from '~/common/util/pwaUtils';
import { isBrowser, isPwa } from '~/common/util/pwaUtils';
import { useUICounter } from '~/common/stores/store-ui';
@@ -31,15 +31,14 @@ export function usePWADesktopModeWarning() {
// if PWA
const isInPwaMode = isPwa();
if (!isInPwaMode) return false;
// if Mobile device (detected using touch points), while desktop have 0 touch points
const isTouchDevice = navigator?.maxTouchPoints > 0;
// if Physical Screen is small - typical mobile screen size - e.g. 412 for SGS24 Ultra
const isSmallScreen = window.screen?.width < 600;
// if Desktop mode - e.g. "Desktop Site" reports a large viewport width, typically 9xx+
const isDesktopWidth = window.matchMedia('(min-width: 900px)').matches;
// if OS is mobile
const isMobileOS = Is.OS.iOS || Is.OS.Android;
if (!isMobileOS) return false;
// Check if viewport width suggests desktop mode (>= 900px)
// This matches the mobile breakpoint used in useMatchMedia.ts
return window.matchMedia('(min-width: 900px)').matches;
return isInPwaMode && isTouchDevice && isSmallScreen && isDesktopWidth;
}, []);
const showWarning = isInDesktopMode && !hideWarning && lessThanFive;
@@ -19,7 +19,7 @@ const _styles = {
textDecoration: 'underline',
},
},
accentedTagline: {
taglineAccented: {
textAlign: 'start',
mt: 0.75,
},
@@ -37,9 +37,10 @@ const _styles = {
export function OptimaAppPageHeading(props: {
title: React.ReactNode;
tagline?: React.ReactNode;
accentedTagline?: boolean;
taglineAccented?: boolean;
startDecorator?: React.ReactNode;
endDecorator?: React.ReactNode;
disabled?: boolean;
noDivider?: boolean;
noMarginBottom?: boolean;
onClick?: (event: React.MouseEvent) => void;
@@ -50,13 +51,13 @@ export function OptimaAppPageHeading(props: {
return (
<Box mb={props.noMarginBottom ? undefined : 2.25} sx={{ overflow: 'hidden', display: 'grid' }}>
{!!props.title && <Typography level={isMobile ? 'h3' : 'h2'} startDecorator={props.startDecorator} endDecorator={props.endDecorator} sx={_styles.title}>
{!!props.title && <Typography level={isMobile ? 'h3' : 'h2'} startDecorator={props.startDecorator} endDecorator={props.endDecorator} textColor={props.disabled ? 'neutral.plainDisabledColor' : undefined} sx={_styles.title}>
{props.onClick
? <Box component='span' sx={_styles.textClickable} onClick={props.onClick} className='agi-ellipsize'>{props.title}</Box>
: <span className='agi-ellipsize'>{props.title}</span>
}
</Typography>}
{!!props.tagline && <Typography level='body-sm' sx={props.accentedTagline ? _styles.accentedTagline : _styles.tagline}>
{!!props.tagline && <Typography level='body-sm' sx={props.taglineAccented ? _styles.taglineAccented : _styles.tagline}>
{props.tagline}
</Typography>}
{!props.noDivider && <ListDivider sx={_styles.divisor} />}
@@ -10,6 +10,8 @@ import { OPTIMA_OPEN_DEBOUNCE, OPTIMA_PEEK_HOVER_ENTER_DELAY, OPTIMA_PEEK_HOVER_
export type PreferencesTabId = 'chat' | 'voice' | 'draw' | 'tools' | undefined;
export type ModelOptionsContext = 'full' | 'parameters';
interface OptimaState {
@@ -27,6 +29,7 @@ interface OptimaState {
showKeyboardShortcuts: boolean;
showLogger: boolean;
showModelOptions: DLLMId | false;
showModelOptionsContext: ModelOptionsContext;
showModels: boolean;
showPreferences: boolean;
preferencesTab: PreferencesTabId;
@@ -51,6 +54,7 @@ const modalsClosedState = {
showKeyboardShortcuts: false,
showLogger: false,
showModelOptions: false,
showModelOptionsContext: 'full' as ModelOptionsContext,
showModels: false,
showPreferences: false,
} as const;
@@ -102,7 +106,7 @@ export interface OptimaActions {
openLogger: () => void;
closeModelOptions: () => void;
openModelOptions: (id: DLLMId) => void;
openModelOptions: (id: DLLMId, context?: ModelOptionsContext) => void;
closeModels: () => void;
openModels: () => void;
@@ -209,7 +213,7 @@ export const useLayoutOptimaStore = create<OptimaState & OptimaActions>((_set, _
openLogger: () => _set({ ...modalsClosedState, showLogger: true }),
closeModelOptions: () => _set({ showModelOptions: false }),
openModelOptions: (id: DLLMId) => _set({ showModelOptions: id }),
openModelOptions: (id: DLLMId, context?: ModelOptionsContext) => _set({ showModelOptions: id, showModelOptionsContext: context ?? 'full' }),
closeModels: () => _set({ showModels: false }),
openModels: () => _set({ showModels: true }),
+1
View File
@@ -91,6 +91,7 @@ export function useOptimaModals() {
showKeyboardShortcuts: state.showKeyboardShortcuts,
showLogger: state.showLogger,
showModelOptions: state.showModelOptions,
showModelOptionsContext: state.showModelOptionsContext,
showModels: state.showModels,
showPreferences: state.showPreferences,
preferencesTab: state.preferencesTab,
@@ -27,6 +27,8 @@ export type GlobalOverlayId = // string - disabled so we keep an orderliness
| 'livefile-overwrite'
| 'shortcuts-confirm-close'
| 'blocks-off-enhance-code'
| 'llms-remove-clones'
| 'llms-reset-parameters'
| 'llms-service-remove'
| 'composer-unsupported-attachments' // The LLM does not seem to support this mime type - continue anyway?
| 'composer-open-or-attach' // Open a file or attach it to the chat?
@@ -11,6 +11,10 @@ export function setupClientUncaughtErrorsLogging(): () => void {
// Handle uncaught exceptions
const handleError = (event: ErrorEvent) => {
// Ignore benign ResizeObserver errors (browser warning, not an actual error)
if (event.message?.includes('ResizeObserver loop'))
return;
logger.error('Uncaught error', {
message: event.error?.message || event.message,
stack: event.error?.stack,
@@ -54,6 +54,8 @@ const scrollableBoxSx: SxProps = {
overflowY: 'auto',
// actually make sure this scrolls & fills
height: '100%',
// prevents pull-to-refresh on mobile when scrolling up in the chat
overscrollBehaviorY: 'none',
} as const;
+7 -7
View File
@@ -208,7 +208,7 @@ export type DMessageToolInvocationPart = {
type: 'code_execution';
language: string;
code: string;
author: 'gemini_auto_inline';
author: DMessageToolCodeExecutor;
}
};
@@ -224,12 +224,12 @@ export type DMessageToolResponsePart = {
} | {
type: 'code_execution';
result: string; // The output
executor: 'gemini_auto_inline';
executor: DMessageToolCodeExecutor;
},
environment: DMessageToolEnvironment,
};
type DMessageToolEnvironment = 'upstream' | 'server' | 'client';
type DMessageToolCodeExecutor = 'gemini_auto_inline' | 'code_interpreter';
type DVoidModelAnnotationsPart = {
pt: 'annotations',
@@ -403,7 +403,7 @@ export function create_FunctionCallInvocation_ContentFragment(id: string, functi
return _createContentFragment(_create_FunctionCallInvocation_Part(id, functionName, args));
}
export function create_CodeExecutionInvocation_ContentFragment(id: string, language: string, code: string, author: 'gemini_auto_inline'): DMessageContentFragment {
export function create_CodeExecutionInvocation_ContentFragment(id: string, language: string, code: string, author: DMessageToolCodeExecutor): DMessageContentFragment {
return _createContentFragment(_create_CodeExecutionInvocation_Part(id, language, code, author));
}
@@ -411,7 +411,7 @@ export function create_FunctionCallResponse_ContentFragment(id: string, error: b
return _createContentFragment(_create_FunctionCallResponse_Part(id, error, name, result, environment));
}
export function create_CodeExecutionResponse_ContentFragment(id: string, error: boolean | string, result: string, executor: 'gemini_auto_inline', environment: DMessageToolEnvironment): DMessageContentFragment {
export function create_CodeExecutionResponse_ContentFragment(id: string, error: boolean | string, result: string, executor: DMessageToolCodeExecutor, environment: DMessageToolEnvironment): DMessageContentFragment {
return _createContentFragment(_create_CodeExecutionResponse_Part(id, error, result, executor, environment));
}
@@ -553,7 +553,7 @@ function _create_FunctionCallInvocation_Part(id: string, functionName: string, a
return { pt: 'tool_invocation', id, invocation: { type: 'function_call', name: functionName, args } };
}
function _create_CodeExecutionInvocation_Part(id: string, language: string, code: string, author: 'gemini_auto_inline'): DMessageToolInvocationPart {
function _create_CodeExecutionInvocation_Part(id: string, language: string, code: string, author: DMessageToolCodeExecutor): DMessageToolInvocationPart {
return { pt: 'tool_invocation', id, invocation: { type: 'code_execution', language, code, author } };
}
@@ -561,7 +561,7 @@ function _create_FunctionCallResponse_Part(id: string, error: boolean | string,
return { pt: 'tool_response', id, error, response: { type: 'function_call', name, result }, environment };
}
function _create_CodeExecutionResponse_Part(id: string, error: boolean | string, result: string, executor: 'gemini_auto_inline', environment: DMessageToolEnvironment): DMessageToolResponsePart {
function _create_CodeExecutionResponse_Part(id: string, error: boolean | string, result: string, executor: DMessageToolCodeExecutor, environment: DMessageToolEnvironment): DMessageToolResponsePart {
return { pt: 'tool_response', id, error, response: { type: 'code_execution', result, executor }, environment };
}
+2 -2
View File
@@ -300,8 +300,8 @@ function _messageSetGeneratorAIX(message: Pick<DMessage, 'generator'>, modelLabe
export function messageSetGeneratorAIX_AutoLabel(message: Pick<DMessage, 'generator'>, modelVendorId: ModelVendorId, modelId: DLLMId): void {
// Simply strip the first part of the modelId, which is the serviceId, before the dash.
const heuristicLabel = modelId.includes('-') ? modelId.replace(/^[^-]+-/, '') : modelId;
// Strip the serviceId prefix: 'vendor-' or 'vendor-N-' (when multiple providers of same vendor)
const heuristicLabel = modelId.includes('-') ? modelId.replace(/^[^-]+-(\d-)?/, '') : modelId;
_messageSetGeneratorAIX(message, heuristicLabel, modelVendorId, modelId);
}
@@ -7,36 +7,38 @@ import type { DModelsService } from '../llms.service.types';
import { llmsStoreActions, useModelsStore } from '../store-llms';
const CSF_KEY = 'csf';
/**
* Hook to manage client-side fetch setting for a model's service
* The CSF setting is stored as 'csf' in service settings for all vendors
*/
export function useModelServiceClientSideFetch(enabled: boolean, model: DLLM | null) {
// memo vendor
const { vendor, csfKey } = React.useMemo(() => {
if (!enabled) return { vendor: null, csfKey: '' };
const vendor = findModelVendor(model?.vId);
const csfKey = vendor?.csfKey || '';
return { vendor, csfKey };
const vendor = React.useMemo(() => {
if (!enabled) return null;
return findModelVendor(model?.vId);
}, [enabled, model?.vId]);
// external state
const service: null | DModelsService = useModelsStore(state => !model?.sId ? null : state.sources.find(s => s.id === model.sId) ?? null);
// actual state
const csfAvailable: boolean | undefined = !!csfKey && vendor?.csfAvailable?.(service?.setup);
const csfActive: boolean | undefined = csfAvailable && (service?.setup as any)?.[csfKey];
const csfAvailable: boolean | undefined = !!vendor?.csfAvailable && vendor?.csfAvailable?.(service?.setup);
const csfActive: boolean | undefined = csfAvailable && (service?.setup as any)?.[CSF_KEY];
const serviceId = service?.id || '';
const csfToggle = React.useCallback((value: boolean) => {
if (csfKey && serviceId)
llmsStoreActions().updateServiceSettings(serviceId, { [csfKey]: value });
}, [csfKey, serviceId]);
if (serviceId)
llmsStoreActions().updateServiceSettings(serviceId, { [CSF_KEY]: value });
}, [serviceId]);
const csfReset = React.useCallback(() => {
if (csfKey && serviceId)
llmsStoreActions().updateServiceSettings(serviceId, { [csfKey]: false });
}, [csfKey, serviceId]);
if (serviceId)
llmsStoreActions().updateServiceSettings(serviceId, { [CSF_KEY]: false });
}, [serviceId]);
return { csfAvailable, csfActive, csfToggle, csfReset };
return { csfAvailable, csfActive, csfToggle, csfReset, vendorName: vendor?.name || vendor?.id || 'AI Service' };
}
+4
View File
@@ -9,6 +9,10 @@ export function useLLM(llmId: undefined | DLLMId | null): DLLM | undefined {
return useModelsStore(state => !llmId ? undefined : state.llms.find(llm => llm.id === llmId));
}
export function useLLMExists(llmId: undefined | DLLMId | null): boolean {
return useModelsStore(state => !llmId ? false : state.llms.some(llm => llm.id === llmId));
}
export function useLLMs(llmIds: ReadonlyArray<DLLMId>): ReadonlyArray<DLLM | undefined> {
return useModelsStore(useShallow(state => {
return llmIds.map(llmId => !llmId ? undefined : state.llms.find(llm => llm.id === llmId));
+259 -125
View File
@@ -20,6 +20,54 @@ export const FALLBACK_LLM_PARAM_TEMPERATURE = 0.5;
// const FALLBACK_LLM_PARAM_REF_UNKNOWN = 'unknown_id';
/// Registry Entry Types (for compile-time validation)
type _ParameterRegistryEntry =
| _IntegerParamDef
| _FloatParamDef
| _StringParamDef
| _BooleanParamDef
| _EnumParamDef;
interface _ParamDefBase {
readonly label: string;
readonly description: string;
}
interface _IntegerParamDef extends _ParamDefBase {
readonly type: 'integer';
readonly range?: readonly [number, number];
readonly nullable?: { readonly meaning: string };
readonly requiredFallback?: number;
readonly initialValue?: number | null;
}
interface _FloatParamDef extends _ParamDefBase {
readonly type: 'float';
readonly range?: readonly [number, number];
readonly nullable?: { readonly meaning: string };
readonly requiredFallback?: number;
readonly initialValue?: number | null;
}
interface _StringParamDef extends _ParamDefBase {
readonly type: 'string';
readonly initialValue?: string;
}
interface _BooleanParamDef extends _ParamDefBase {
readonly type: 'boolean';
readonly initialValue?: boolean;
}
interface _EnumParamDef extends _ParamDefBase {
readonly type: 'enum';
readonly values: readonly string[];
readonly requiredFallback?: string;
readonly initialValue?: string;
}
/// Registry
export const DModelParameterRegistry = {
@@ -29,60 +77,64 @@ export const DModelParameterRegistry = {
llmRef: {
label: 'Model ID',
type: 'string' as const,
type: 'string',
description: 'Upstream model reference',
hidden: true,
} as const,
},
llmResponseTokens: {
label: 'Maximum Tokens',
type: 'integer' as const,
type: 'integer',
description: 'Maximum length of generated text',
nullable: {
meaning: 'Explicitly avoid sending max_tokens to upstream API',
} as const,
},
requiredFallback: FALLBACK_LLM_PARAM_RESPONSE_TOKENS, // if required and not specified/user overridden, use this value
} as const,
},
llmTemperature: {
label: 'Temperature',
type: 'float' as const,
type: 'float',
description: 'Controls randomness in the output',
range: [0.0, 2.0] as const,
nullable: {
meaning: 'Explicitly avoid sending temperature to upstream API',
} as const,
},
requiredFallback: FALLBACK_LLM_PARAM_TEMPERATURE,
} as const,
},
/// Extended parameters, specific to certain models/vendors
llmTopP: {
label: 'Top P',
type: 'float' as const,
type: 'float',
description: 'Nucleus sampling threshold',
range: [0.0, 1.0] as const,
requiredFallback: 1.0,
incompatibleWith: ['temperature'] as const,
} as const,
},
/**
* First introduced as a user-configurable parameter for the 'Verification' required by o3.
* [2025-04-16] Adding parameter to disable streaming for o3, and possibly more models.
*
* [2026-01-21] OpenAI Responses API: Reasoning Summaries require organization verification.
* Per OpenAI docs, both streaming AND reasoning summaries require org verification for GPT-5/5.1/5.2.
* - https://help.openai.com/en/articles/10362446-api-model-availability-by-usage-tier-and-verification-status
* - Rather than adding a separate param, we piggyback on llmForceNoStream.
* - AIX Wire type `vndOaiReasoningSummary` is derived from `llmForceNoStream` in aix.client.ts.
*/
llmForceNoStream: {
label: 'Disable Streaming',
type: 'boolean' as const,
type: 'boolean',
description: 'Disables streaming for this model',
// initialValue: false, // we don't need the initial value here, will be assumed off
} as const,
},
llmVndAnt1MContext: {
label: '1M Context Window (Beta)',
type: 'boolean' as const,
type: 'boolean',
description: 'Enable 1M token context window with premium pricing for >200K input tokens',
// No initialValue - undefined means off (e.g. default 200K context window)
} as const,
},
llmVndAntEffort: {
label: 'Effort',
@@ -94,37 +146,37 @@ export const DModelParameterRegistry = {
llmVndAntSkills: {
label: 'Document Skills',
type: 'string' as const,
type: 'string',
description: 'Comma-separated skills (xlsx,pptx,pdf,docx)',
initialValue: '', // empty string = disabled
} as const,
},
llmVndAntThinkingBudget: {
label: 'Thinking Budget',
type: 'integer' as const,
type: 'integer',
description: 'Budget for extended thinking',
range: [1024, 65536] as const,
initialValue: 16384,
nullable: {
meaning: 'Disable extended thinking',
} as const,
} as const,
},
},
llmVndAntWebFetch: {
label: 'Web Fetch',
type: 'enum' as const,
type: 'enum',
description: 'Enable fetching content from web pages and PDFs',
values: ['auto', 'off'] as const,
// No initialValue - undefined means off (same as 'off')
} as const,
},
llmVndAntWebSearch: {
label: 'Web Search',
type: 'enum' as const,
type: 'enum',
description: 'Enable web search for real-time information',
values: ['auto', 'off'] as const,
// No initialValue - undefined means off (same as 'off')
} as const,
},
// llmVndAntToolSearch: { // Not user set
// label: 'Tool Search',
@@ -136,64 +188,63 @@ export const DModelParameterRegistry = {
llmVndGeminiAspectRatio: {
label: 'Aspect Ratio',
type: 'enum' as const,
type: 'enum',
description: 'Controls the aspect ratio of generated images',
values: ['1:1', '2:3', '3:2', '3:4', '4:3', '9:16', '16:9', '21:9'] as const,
// No initial value - when undefined, the model decides the aspect ratio
} as const,
},
llmVndGeminiCodeExecution: {
label: 'Code Execution',
type: 'enum' as const,
type: 'enum',
description: 'Enable automatic Python code generation and execution by the model',
values: ['auto'] as const,
// No initialValue - undefined means off
} as const,
},
llmVndGeminiComputerUse: {
label: 'Computer Use Environment',
type: 'enum' as const,
type: 'enum',
description: 'Environment type for Computer Use tool (required for Computer Use model)',
values: ['browser'] as const,
initialValue: 'browser',
// requiredFallback: 'browser', // See `const _requiredParamId: DModelParameterId[]` in llms.parameters.ts for why custom params don't have required values at AIX invocation...
hidden: true,
} as const,
},
llmVndGeminiGoogleSearch: {
label: 'Google Search',
type: 'enum' as const,
type: 'enum',
description: 'Enable Google Search grounding with optional time filter',
values: ['unfiltered', '1d', '1w', '1m', '6m', '1y'] as const,
// No initialValue - undefined means off
} as const,
},
llmVndGeminiImageSize: { // [Gemini, 2025-11-20] Nano Banana launch
label: 'Image Size',
type: 'enum' as const,
type: 'enum',
description: 'Controls the resolution of generated images',
values: ['1K', '2K', '4K'] as const,
// No initial value - when undefined, the model decides the image size
} as const,
},
llmVndGeminiMediaResolution: {
label: 'Media Resolution',
type: 'enum' as const,
type: 'enum',
description: 'Controls vision processing quality for multimodal inputs. Higher resolution improves text reading and detail identification but increases token usage.',
values: ['mr_high', 'mr_medium', 'mr_low'] as const,
// No initialValue - undefined: "If unspecified, the model uses optimal defaults based on the media type." (Images: high, PDFs: medium, Videos: low/medium (rec: high for OCR))
} as const,
},
llmVndGeminiShowThoughts: {
label: 'Show Thoughts',
type: 'boolean' as const,
type: 'boolean',
description: 'Show Gemini\'s reasoning process',
// initialValue: true, // no initial value
} as const,
},
llmVndGeminiThinkingBudget: {
label: 'Thinking Budget',
type: 'integer' as const,
type: 'integer',
/**
* can be overwritten, as gemini models seem to have different ranges which also does not include 0
* - value = 0 disables thinking
@@ -202,15 +253,23 @@ export const DModelParameterRegistry = {
range: [0, 24576] as const,
// initialValue: unset, // auto-budgeting
description: 'Budget for extended thinking. 0 disables thinking. If not set, the model chooses automatically.',
} as const,
},
llmVndGeminiThinkingLevel: {
label: 'Thinking Level',
type: 'enum' as const,
description: 'Controls internal reasoning depth. Replaces thinking_budget for Gemini 3 models. When unset, the model decides dynamically.',
values: ['high', 'medium' /* not present at launch */, 'low' /* default when unset */] as const,
// No initialValue - undefined means 'dynamic', which for Gemini Pro is the same as 'high' (which is the equivalent of 'medium' for OpenAI's effort levels.. somehow)
} as const,
type: 'enum',
description: 'Controls internal reasoning depth for Gemini 3 Pro. When unset, the model decides dynamically.',
values: ['high', 'low'] as const,
// No initialValue - undefined means 'dynamic', which for Gemini Pro is the same as 'high'
},
llmVndGeminiThinkingLevel4: {
label: 'Thinking Level',
type: 'enum',
description: 'Controls internal reasoning depth for Gemini 3 Flash. When unset, the model decides dynamically.',
values: ['high', 'medium', 'low', 'minimal'] as const,
// No initialValue - undefined means 'dynamic'
},
// NOTE: we don't have this as a parameter, as for now we use it in tandem with llmVndGeminiGoogleSearch
// llmVndGeminiUrlContext: {
@@ -223,54 +282,84 @@ export const DModelParameterRegistry = {
// Moonshot-specific parameters
llmVndMoonReasoningEffort: {
label: 'Reasoning Effort',
type: 'enum',
description: 'Controls thinking depth for Kimi K2.5. High enables extended multi-step reasoning (default).',
values: ['none', 'high'] as const,
// No initialValue - undefined means high (thinking enabled, the default for K2.5)
},
llmVndMoonshotWebSearch: {
label: 'Web Search',
type: 'enum' as const,
type: 'enum',
description: 'Enable Kimi\'s $web_search builtin function for real-time web search ($0.005 per search)',
values: ['auto'] as const,
// No initialValue - undefined means off
} as const,
},
// OpenAI-specific parameters
// Reasoning effort levels per model:
// - GPT-5: minimal, low, medium (default), high
// - GPT-5.1: none (default), low, medium, high
// - GPT-5.2: none (default), low, medium, high, xhigh
// - GPT-5.2 Pro: medium (default), high, xhigh
llmVndOaiReasoningEffort: {
label: 'Reasoning Effort',
type: 'enum' as const,
type: 'enum',
description: 'Constrains effort on reasoning for OpenAI reasoning models',
values: ['low', 'medium', 'high'] as const,
requiredFallback: 'medium',
} as const,
},
llmVndOaiReasoningEffort4: {
label: 'Reasoning Effort',
type: 'enum' as const,
type: 'enum',
description: 'Constrains effort on reasoning for OpenAI advanced reasoning models',
values: ['minimal', 'low', 'medium', 'high'] as const,
requiredFallback: 'medium',
} as const,
},
llmVndOaiReasoningEffort52: {
label: 'Reasoning Effort',
type: 'enum',
description: 'Constrains effort on reasoning for GPT-5.2 models. When unset, defaults to none (fast responses).',
values: ['none', 'low', 'medium', 'high', 'xhigh'] as const,
// No requiredFallback - unset = none (the default for GPT-5.2)
// No initialValue - starts undefined, which the UI should display as "none"
},
llmVndOaiReasoningEffort52Pro: {
label: 'Reasoning Effort',
type: 'enum',
description: 'Constrains effort on reasoning for GPT-5.2 Pro. Defaults to medium.',
values: ['medium', 'high', 'xhigh'] as const,
// No requiredFallback - unset = medium (the default for GPT-5.2 Pro)
},
llmVndOaiRestoreMarkdown: {
label: 'Restore Markdown',
type: 'boolean' as const,
type: 'boolean',
description: 'Restore Markdown formatting in the output',
initialValue: true,
} as const,
},
llmVndOaiVerbosity: {
label: 'Verbosity',
type: 'enum' as const,
type: 'enum',
description: 'Controls response length and detail level',
values: ['low', 'medium', 'high'] as const,
requiredFallback: 'medium',
} as const,
},
llmVndOaiWebSearchContext: {
label: 'Search Context Size',
type: 'enum' as const,
type: 'enum',
description: 'Amount of context retrieved from the web',
values: ['low', 'medium', 'high'] as const,
requiredFallback: 'medium',
} as const,
},
llmVndOaiWebSearchGeolocation: {
// NOTE: for now this is a boolean to enable/disable using client-side geolocation, but
@@ -278,19 +367,27 @@ export const DModelParameterRegistry = {
// back if of type AixAPI_Model.userGeolocation, which is the AIX Wire format for the
// location payload.
label: 'Add User Location (Geolocation API)',
type: 'boolean' as const,
type: 'boolean',
description: 'Approximate location for search results',
initialValue: false,
} as const,
},
llmVndOaiImageGeneration: {
label: 'Image Generation',
type: 'enum' as const,
type: 'enum',
description: 'Image generation mode and quality',
values: ['mq', 'hq', 'hq_edit' /* precise input editing */, 'hq_png' /* uncompressed */] as const,
// No initialValue - defaults to undefined (off)
// No requiredFallback - this is optional
} as const,
},
llmVndOaiCodeInterpreter: {
label: 'Code Interpreter',
type: 'enum',
description: 'Python code execution ($0.03/container)',
values: ['off', 'auto'] as const,
// No initialValue - undefined means off (same as 'off')
},
// Perplexity-specific parameters
@@ -298,64 +395,120 @@ export const DModelParameterRegistry = {
llmVndPerplexityDateFilter: {
label: 'Date Range',
type: 'enum' as const,
type: 'enum',
description: 'Filter results by publication date',
values: ['unfiltered', '1m', '3m', '6m', '1y'] as const,
// requiredFallback: 'unfiltered',
} as const,
},
llmVndOrtWebSearch: {
label: 'Web Search',
type: 'enum' as const,
type: 'enum',
description: 'Enable OpenRouter web search (uses native search for OpenAI/Anthropic, Exa for others)',
values: ['auto'] as const,
// No initialValue - undefined means off
} as const,
},
llmVndPerplexitySearchMode: {
label: 'Search Mode',
type: 'enum' as const,
type: 'enum',
description: 'Type of sources to search',
values: ['default', 'academic'] as const,
// requiredFallback: 'default', // or leave unset for "unspecified"
} as const,
},
// xAI-specific parameters
llmVndXaiSearchMode: {
label: 'Search Mode',
type: 'enum' as const,
description: 'Controls when to use live search',
values: ['auto', 'on', 'off'] as const,
initialValue: 'auto', // we default to auto for our users, to get them search out of the box
} as const,
llmVndXaiCodeExecution: {
label: 'Code Execution',
type: 'enum',
description: 'Enable server-side code execution by the model',
values: ['off', 'auto'] as const,
// No initialValue - undefined means off (same as 'off')
},
llmVndXaiSearchSources: {
label: 'Search Sources',
type: 'string' as const,
description: 'Comma-separated sources (web,x,news,rss)',
initialValue: 'web,x', // defaults to web,x as per xAI docs
} as const,
llmVndXaiSearchDateFilter: {
label: 'Search From Date',
type: 'enum' as const,
description: 'Filter search results by publication date',
llmVndXaiSearchInterval: {
label: 'Search Interval', // "X Search only" for now, fw comp to web search
type: 'enum',
description: 'Search in this interval',
values: ['unfiltered', '1d', '1w', '1m', '6m', '1y'] as const,
// requiredFallback: 'unfiltered',
} as const,
// No initialValue - undefined means unfiltered
},
} as const;
llmVndXaiWebSearch: {
label: 'Web Search',
type: 'enum',
description: 'Enable web search for real-time information',
values: ['off', 'auto'] as const,
// No initialValue - undefined means off (same as 'off')
},
llmVndXaiXSearch: {
label: 'X Search',
type: 'enum',
description: 'Enable X/Twitter search for social media content',
values: ['off', 'auto'] as const,
// NOTE: disabling or this could be slow
// initialValue: 'auto', // we default to 'auto' for our users, as they may expect "X search" out of the box
},
llmVndXaiXSearchHandles: {
label: 'X Handles Filter',
type: 'string',
description: 'Filter X search to specific handles (comma-separated, e.g. @elonmusk, @xai)',
// initialValue: '', // empty = no filter
},
} as const satisfies Record<string, _ParameterRegistryEntry>;
/// Types
// this is the client-side typescript definition that matches ModelParameterSpec_schema in `llm.server.types.ts`
export interface DModelParameterSpec<T extends DModelParameterId> {
/** Stores runtime parameter values (initial and user overrides). */
export type DModelParameterValues = {
[K in DModelParameterId]?: DModelParameterValue<K>;
};
export type DModelParameterId = keyof typeof DModelParameterRegistry;
/** Maps a parameter ID to its TypeScript value type (with nullable handling). */
export type DModelParameterValue<T extends DModelParameterId> =
typeof DModelParameterRegistry[T] extends { nullable: object }
? _ParamTypeToBaseValue<T> | null
: _ParamTypeToBaseValue<T>;
// helper: map parameter type to base TypeScript type (before nullable handling)
type _ParamTypeToBaseValue<T extends DModelParameterId> =
typeof DModelParameterRegistry[T]['type'] extends 'integer' ? number :
typeof DModelParameterRegistry[T]['type'] extends 'float' ? number :
typeof DModelParameterRegistry[T]['type'] extends 'string' ? string :
typeof DModelParameterRegistry[T]['type'] extends 'boolean' ? boolean :
typeof DModelParameterRegistry[T]['type'] extends 'enum' ? _EnumValues<typeof DModelParameterRegistry[T]> :
never;
type _EnumValues<T> = T extends { readonly type: 'enum'; readonly values: readonly (infer U)[] } ? U : never;
/**
* Union of all possible model parameter specifications.
*/
export type DModelParameterSpecAny = {
[K in DModelParameterId]: DModelParameterSpec<K>;
}[DModelParameterId];
/**
* Model-specific parameter configuration
* Defines which parameters a model supports and their per-model settings.
*
* Note: This is the client-side TypeScript definition that matches
* ModelParameterSpec_schema in `llm.server.types.ts`.
*/
interface DModelParameterSpec<T extends DModelParameterId> {
paramId: T;
required?: boolean;
hidden?: boolean;
initialValue?: boolean | number | string | null;
initialValue?: DModelParameterValue<T>;
// upstreamDefault?: DModelParameterValue<T>;
/**
* (optional, rare) Special: [min, max] range override for this parameter.
@@ -364,44 +517,21 @@ export interface DModelParameterSpec<T extends DModelParameterId> {
rangeOverride?: [number, number];
}
export type DModelParameterValues = {
[K in DModelParameterId]?: DModelParameterValue<K>;
}
export type DModelParameterId = keyof typeof DModelParameterRegistry;
// type _ExtendedParameterId = keyof typeof _ExtendedParameterRegistry;
type _EnumValues<T> = T extends { type: 'enum', values: readonly (infer U)[] } ? U : never;
type DModelParameterValue<T extends DModelParameterId> =
typeof DModelParameterRegistry[T]['type'] extends 'integer'
? typeof DModelParameterRegistry[T] extends { nullable: any }
? number | null
: number :
typeof DModelParameterRegistry[T]['type'] extends 'float'
? typeof DModelParameterRegistry[T] extends { nullable: any }
? number | null
: number :
typeof DModelParameterRegistry[T]['type'] extends 'string' ? string :
typeof DModelParameterRegistry[T]['type'] extends 'boolean' ? boolean :
typeof DModelParameterRegistry[T]['type'] extends 'enum'
? _EnumValues<typeof DModelParameterRegistry[T]>
: never;
/// Utility Functions
export function applyModelParameterInitialValues(destValues: DModelParameterValues, parameterSpecs: DModelParameterSpec<DModelParameterId>[], overwriteExisting: boolean): void {
for (const param of parameterSpecs) {
const paramId = param.paramId;
export function applyModelParameterSpecsInitialValues(destValues: DModelParameterValues, modelParameterSpecs: DModelParameterSpecAny[], overwriteExisting: boolean): void {
for (const parameterSpec of modelParameterSpecs) {
const paramId = parameterSpec.paramId;
// skip if already present
// NOTE: for the currently only caller, the destValues already has llmRef, llmTemperature, llmResponseTokens
if (!overwriteExisting && paramId in destValues)
continue;
// 1. (if present) apply Spec.initialValue
if (param.initialValue !== undefined) {
destValues[paramId] = param.initialValue as DModelParameterValue<typeof paramId>;
if (parameterSpec.initialValue !== undefined) {
destValues[paramId] = parameterSpec.initialValue as DModelParameterValue<typeof paramId>;
continue;
}
@@ -416,17 +546,21 @@ export function applyModelParameterInitialValues(destValues: DModelParameterValu
}
const _requiredParamId: DModelParameterId[] = [
/**
* Implicit common parameters always supported by all models, not listed in parameterSpecs.
* Must be preserved during model refresh operations.
*/
export const LLMS_ImplicitParamIds: readonly DModelParameterId[] = [
// 'llmRef', // disabled: we know this can't have a fallback value in the registry
'llmResponseTokens', // DModelParameterRegistry.llmResponseTokens.requiredFallback = FALLBACK_LLM_PARAM_RESPONSE_TOKENS
'llmTemperature', // DModelParameterRegistry.llmTemperature.requiredFallback = FALLBACK_LLM_PARAM_TEMPERATURE
] as const;
];
export function getAllModelParameterValues(initialParameters: undefined | DModelParameterValues, userParameters?: DModelParameterValues): DModelParameterValues {
// fallback values
const fallbackParameters: DModelParameterValues = {};
for (const requiredParamId of _requiredParamId) {
for (const requiredParamId of LLMS_ImplicitParamIds) {
if ('requiredFallback' in DModelParameterRegistry[requiredParamId])
fallbackParameters[requiredParamId] = DModelParameterRegistry[requiredParamId].requiredFallback as DModelParameterValue<typeof requiredParamId>;
}
+20 -16
View File
@@ -4,7 +4,7 @@
import type { ModelVendorId } from '~/modules/llms/vendors/vendors.registry';
import type { DModelParameterId, DModelParameterSpec, DModelParameterValues } from './llms.parameters';
import type { DModelParameterSpecAny, DModelParameterValues } from './llms.parameters';
import type { DModelPricing } from './llms.pricing';
import type { DModelsServiceId } from './llms.service.types';
@@ -21,37 +21,40 @@ export type DLLMId = string;
export interface DLLM {
id: DLLMId;
// editable properties (kept on update, if isEdited)
// factory properties (overwritten on update)
label: string;
created: number | 0;
updated?: number | 0;
description: string;
hidden: boolean; // default hidden state (can change underlying between refreshes)
hidden: boolean;
// hard properties (overwritten on update)
contextTokens: DLLMContextTokens; // null: must assume it's unknown
maxOutputTokens: DLLMMaxOutputTokens; // null: must assume it's unknown
trainingDataCutoff?: string; // 'Apr 2029'
interfaces: DModelInterfaceV1[]; // if set, meaning this is the known and comprehensive set of interfaces
benchmark?: { cbaElo?: number, cbaMmlu?: number }; // benchmark values
benchmark?: { cbaElo?: number }; // benchmark values (Chat Bot Arena ELO)
pricing?: DModelPricing;
// parameters system
parameterSpecs: DModelParameterSpec<DModelParameterId>[];
// parameters system (overwritten on update)
parameterSpecs: DModelParameterSpecAny[];
initialParameters: DModelParameterValues;
// references
sId: DModelsServiceId;
vId: ModelVendorId;
// references (const, never change)
sId: DModelsServiceId; // could be weak, but they're removed at the same time
vId: ModelVendorId; // known hardcoded value
// user edited properties - if not undefined/missing, they override the others
userLabel?: string;
userHidden?: boolean;
userStarred?: boolean;
userParameters?: DModelParameterValues; // user has set these parameters
userContextTokens?: DLLMContextTokens; // user override for context window
userMaxOutputTokens?: DLLMMaxOutputTokens; // user override for max output tokens
userPricing?: DModelPricing; // user override for model pricing
userContextTokens?: DLLMContextTokens;
userMaxOutputTokens?: DLLMMaxOutputTokens;
userPricing?: DModelPricing;
userParameters?: DModelParameterValues;
// clone metadata - user-created duplicates of models with independent settings
isUserClone?: boolean; // true if this is a user-created clone
cloneSourceId?: DLLMId; // original model ID (for reference)
}
@@ -157,6 +160,7 @@ export type DModelInterfaceV1 =
| 'tools-web-search' // TEMP: ui flag - supports integrated web search tool
| 'hotfix-no-stream' // disable streaming for o1-preview (old) and o1 (20241217)
| 'hotfix-no-temperature' // disable temperature for deepseek-r1
| 'hotfix-no-webp' // convert WebP images to PNG (e.g. some local models via LM Studio)
| 'hotfix-strip-images' // strip images from the input
| 'hotfix-strip-sys0' // strip the system instruction (unsupported)
| 'hotfix-sys0-to-usr0' // cast sys0 to usr0
@@ -178,11 +182,11 @@ export const LLM_IF_Tools_WebSearch: DModelInterfaceV1 = 'tools-web-search';
export const LLM_IF_OAI_Complete: DModelInterfaceV1 = 'oai-complete';
export const LLM_IF_ANT_PromptCaching: DModelInterfaceV1 = 'ant-prompt-caching';
export const LLM_IF_OAI_PromptCaching: DModelInterfaceV1 = 'oai-prompt-caching';
export const LLM_IF_OAI_Realtime: DModelInterfaceV1 = 'oai-realtime';
export const LLM_IF_OAI_Responses: DModelInterfaceV1 = 'oai-responses';
export const LLM_IF_GEM_CodeExecution: DModelInterfaceV1 = 'gem-code-execution';
export const LLM_IF_HOTFIX_NoStream: DModelInterfaceV1 = 'hotfix-no-stream';
export const LLM_IF_HOTFIX_NoTemperature: DModelInterfaceV1 = 'hotfix-no-temperature';
export const LLM_IF_HOTFIX_NoWebP: DModelInterfaceV1 = 'hotfix-no-webp';
export const LLM_IF_HOTFIX_StripImages: DModelInterfaceV1 = 'hotfix-strip-images';
export const LLM_IF_HOTFIX_StripSys0: DModelInterfaceV1 = 'hotfix-strip-sys0';
export const LLM_IF_HOTFIX_Sys0ToUsr0: DModelInterfaceV1 = 'hotfix-sys0-to-usr0';
@@ -206,11 +210,11 @@ export const LLMS_ALL_INTERFACES = [
LLM_IF_ANT_PromptCaching, // [Anthropic] model supports anthropic-specific caching
LLM_IF_GEM_CodeExecution, // [Gemini] Tool: code execution
LLM_IF_OAI_PromptCaching, // [OpenAI] model supports OpenAI prompt caching
LLM_IF_OAI_Realtime, // [OpenAI] realtime API support - unused
LLM_IF_OAI_Responses, // [OpenAI] Responses API (new) support
// Hotfixes to patch specific model quirks
LLM_IF_HOTFIX_NoStream, // disable streaming (e.g., o1-preview(old))
LLM_IF_HOTFIX_NoTemperature,// disable temperature parameter (e.g., deepseek-r1)
LLM_IF_HOTFIX_NoWebP, // convert WebP images to PNG (e.g. LM Studio)
LLM_IF_HOTFIX_StripImages, // remove images from input (e.g. o3-mini-2025-01-31)
LLM_IF_HOTFIX_StripSys0, // strip system instruction (e.g. Gemini Image Generation 2025-03-13), excludes Sys0ToUsr0
LLM_IF_HOTFIX_Sys0ToUsr0, // downgrade system to user messages for this model (e.g. o1-mini-2024-09-12)
+147 -44
View File
@@ -7,7 +7,8 @@ import { persist } from 'zustand/middleware';
import type { DOpenRouterServiceSettings } from '~/modules/llms/vendors/openrouter/openrouter.vendor';
import type { IModelVendor } from '~/modules/llms/vendors/IModelVendor';
import type { ModelVendorId } from '~/modules/llms/vendors/vendors.registry';
import { createDLLMUserClone, getDLLMCloneId } from '~/modules/llms/llm.client';
import { findModelVendor, type ModelVendorId } from '~/modules/llms/vendors/vendors.registry';
import { hasKeys } from '~/common/util/objectUtils';
@@ -15,6 +16,7 @@ import type { DModelDomainId } from './model.domains.types';
import type { DModelParameterId, DModelParameterValues } from './llms.parameters';
import type { DModelsService, DModelsServiceId } from './llms.service.types';
import { DLLM, DLLMId, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision } from './llms.types';
import { DModelParameterRegistry, LLMS_ImplicitParamIds } from './llms.parameters';
import { createDModelConfiguration, DModelConfiguration } from './modelconfiguration.types';
import { createLlmsAssignmentsSlice, LlmsAssignmentsActions, LlmsAssignmentsSlice, LlmsAssignmentsState, llmsHeuristicUpdateAssignments } from './store-llms-domains_slice';
import { getDomainModelConfiguration } from './hooks/useModelDomain';
@@ -35,17 +37,21 @@ export interface LlmsRootState {
interface LlmsRootActions {
setServiceLLMs: (serviceId: DModelsServiceId, serviceLLMs: ReadonlyArray<DLLM>, keepUserEdits: boolean, keepMissingLLMs: boolean) => void;
setServiceLLMs: (serviceId: DModelsServiceId, serviceLLMs: ReadonlyArray<DLLM>, keepUserEdits: true, keepMissingLLMs: false) => void;
removeLLM: (id: DLLMId) => void;
removeCustomModels: (serviceId: DModelsServiceId) => void;
rerankLLMsByServices: (serviceIdOrder: DModelsServiceId[]) => void;
updateLLM: (id: DLLMId, partial: Partial<DLLM>) => void;
updateLLMs: (updates: Array<{ id: DLLMId; partial: Partial<DLLM> }>) => void;
updateLLMUserParameters: (id: DLLMId, partial: Partial<DModelParameterValues>) => void;
deleteLLMUserParameter: (id: DLLMId, parameterId: DModelParameterId) => void;
resetLLMUserParameters: (id: DLLMId) => void;
resetServiceUserParameters: (serviceId: DModelsServiceId) => void;
userCloneLLM: (sourceId: DLLMId, cloneLabel: string, cloneVariant: string) => DLLMId | null;
createModelsService: (vendor: IModelVendor) => DModelsService;
removeService: (id: DModelsServiceId) => void;
updateServiceLabel: (id: DModelsServiceId, label: string, allowEmpty?: boolean) => void;
updateServiceSettings: <TServiceSettings>(id: DModelsServiceId, partialSettings: Partial<TServiceSettings>) => void;
setConfServiceId: (id: DModelsServiceId | null) => void;
@@ -74,42 +80,75 @@ export const useModelsStore = create<LlmsStore>()(persist(
// actions
setServiceLLMs: (serviceId: DModelsServiceId, serviceLLMs: ReadonlyArray<DLLM>, keepUserEdits: boolean, keepMissingLLMs: boolean) =>
set(({ llms: existingLLMs, modelAssignments }) => {
setServiceLLMs: (serviceId: DModelsServiceId, updatedServiceLLMs: ReadonlyArray<DLLM>, keepUserEdits: true, keepMissingLLMs: false) =>
set(({ llms, modelAssignments }) => {
// keep existing model customizations
if (keepUserEdits) {
serviceLLMs = serviceLLMs.map((llm: DLLM): DLLM => {
const existing = existingLLMs.find(m => m.id === llm.id);
if (!existing) return llm;
// separate existing models
const otherServiceLLMs = llms.filter(llm => llm.sId !== serviceId);
const previousServiceLLMs = llms.filter(llm => llm.sId === serviceId);
const consumedPreviousIds = new Set<DLLMId>();
const result = {
...llm,
...(existing.userLabel !== undefined ? { userLabel: existing.userLabel } : {}),
...(existing.userHidden !== undefined ? { userHidden: existing.userHidden } : {}),
...(existing.userStarred !== undefined ? { userStarred: existing.userStarred } : {}),
...(existing.userParameters !== undefined ? { userParameters: { ...existing.userParameters } } : {}),
...(existing.userContextTokens !== undefined ? { userContextTokens: existing.userContextTokens } : {}),
...(existing.userMaxOutputTokens !== undefined ? { userMaxOutputTokens: existing.userMaxOutputTokens } : {}),
...(existing.userPricing !== undefined ? { userPricing: existing.userPricing } : {}),
};
// process updated models, re-applying user customizations where applicable
const mergedServiceLLMs: DLLM[] = updatedServiceLLMs.map((llm: DLLM): DLLM => {
// new model: as-is
const e = previousServiceLLMs.find(m => m.id === llm.id);
if (!e) return llm;
// clean up stale parameters from userParameters - e.g. was in the model spec but removed in the new version
if (result.userParameters)
for (const key of Object.keys(result.userParameters))
if (!llm.parameterSpecs.some(spec => spec.paramId === key))
delete result.userParameters[key as DModelParameterId];
// mark this previous model as matched (consumed)
consumedPreviousIds.add(e.id);
return result;
});
}
// re-apply user edits from existing model to the new model data
if (!keepUserEdits) return llm;
const result: DLLM = {
...llm,
...(e.userLabel !== undefined ? { userLabel: e.userLabel } : {}),
...(e.userHidden !== undefined ? { userHidden: e.userHidden } : {}),
...(e.userStarred !== undefined ? { userStarred: e.userStarred } : {}),
...(e.userContextTokens !== undefined ? { userContextTokens: e.userContextTokens } : {}),
...(e.userMaxOutputTokens !== undefined ? { userMaxOutputTokens: e.userMaxOutputTokens } : {}),
...(e.userPricing !== undefined ? { userPricing: e.userPricing } : {}),
...(e.userParameters !== undefined ? { userParameters: { ...e.userParameters } } : {}),
};
// remove models that are not in the new list
if (!keepMissingLLMs)
existingLLMs = existingLLMs.filter(llm => llm.sId !== serviceId);
// clean up stale parameters from userParameters -
// - e.g. was in the model spec but removed in the new version
// - or the value of an enum got removed, and so we remove ours
if (result.userParameters) {
for (const key of Object.keys(result.userParameters)) {
const paramId = key as DModelParameterId;
// replace existing llms with the same id
const newLlms = [...serviceLLMs, ...existingLLMs.filter(existingLlm => !serviceLLMs.some(newLlm => newLlm.id === existingLlm.id))];
// keep implicit common parameters (always supported, not in parameterSpecs)
if (LLMS_ImplicitParamIds.includes(paramId))
continue;
// remove parameters no longer in spec
const paramSpec = llm.parameterSpecs.find(spec => spec.paramId === paramId);
if (!paramSpec) {
delete result.userParameters[paramId];
continue;
}
// for enum types, validate the value is still in the allowed values (e.g., 'medium' was removed from thinkingLevel)
const regDef = DModelParameterRegistry[paramId];
if (regDef && regDef.type === 'enum' && 'values' in regDef) {
const currentValue = result.userParameters[paramId];
if (currentValue && typeof currentValue === 'string' && !(regDef.values as readonly string[]).includes(currentValue))
delete result.userParameters[paramId]; // reset to default (undefined)
}
}
}
return result;
});
// Always preserve custom models
// - NOTE: shall we check for the undelying ref to still be in the service, to auto-clean-up older models?
const customModels = previousServiceLLMs.filter(llm => llm.isUserClone === true && !consumedPreviousIds.has(llm.id));
const missingModels = !keepMissingLLMs ? [] : previousServiceLLMs.filter(llm => !llm.isUserClone && !consumedPreviousIds.has(llm.id));
// Build the final list in priority order
const newLlms = [...customModels, ...missingModels, ...mergedServiceLLMs, ...otherServiceLLMs];
return {
llms: newLlms,
modelAssignments: llmsHeuristicUpdateAssignments(newLlms, modelAssignments),
@@ -125,6 +164,15 @@ export const useModelsStore = create<LlmsStore>()(persist(
};
}),
removeCustomModels: (serviceId: DModelsServiceId) =>
set(state => {
const newLlms = state.llms.filter(llm => !(llm.sId === serviceId && llm.isUserClone === true));
return {
llms: newLlms,
modelAssignments: llmsHeuristicUpdateAssignments(newLlms, state.modelAssignments),
};
}),
rerankLLMsByServices: (serviceIdOrder: DModelsServiceId[]) =>
set(state => {
// Create a mapping of service IDs to their index in the provided order
@@ -195,8 +243,40 @@ export const useModelsStore = create<LlmsStore>()(persist(
}),
})),
resetServiceUserParameters: (serviceId: DModelsServiceId) =>
set(({ llms }) => ({
llms: llms.map((llm: DLLM): DLLM => {
if (llm.sId !== serviceId) return llm;
// strip away just the user parameters
const { userParameters /*, userContextTokens, userMaxOutputTokens, userPricing, ...*/, ...rest } = llm;
return rest;
}),
})),
userCloneLLM: (sourceId: DLLMId, cloneLabel: string, cloneVariant: string): DLLMId | null => {
const { llms } = get();
const sourceLlm = llms.find(llm => llm.id === sourceId);
if (!sourceLlm) return null;
// check uniqueness
const cloneId = getDLLMCloneId(sourceId, cloneVariant);
if (llms.some(llm => llm.id === cloneId)) return null;
// create clone
const cloneLlm = createDLLMUserClone(sourceLlm, cloneLabel, cloneVariant);
// IMPORTANT: we have to have this LLM be part of the same group (or the UI will break on multiple-grouping)
const serviceStartIndex = llms.findIndex(llm => llm.sId === sourceLlm.sId);
const newLlms = [...llms];
newLlms.splice(serviceStartIndex, 0, cloneLlm);
set({ llms: newLlms });
return cloneId;
},
createModelsService: (vendor: IModelVendor): DModelsService => {
// e.g. 'openai', 'openai-1', 'openai-2' - finds the first available slot
function _locallyUniqueServiceId(vendorId: ModelVendorId, existingServices: DModelsService[]): DModelsServiceId {
let serviceId: DModelsServiceId = vendorId;
let serviceIdx = 0;
@@ -207,32 +287,35 @@ export const useModelsStore = create<LlmsStore>()(persist(
return serviceId;
}
function _relabelServicesFromSameVendor(vendorId: ModelVendorId, services: DModelsService[]): DModelsService[] {
let n = 0;
return services.map((s: DModelsService): DModelsService =>
(s.vId !== vendorId) ? s
: { ...s, label: s.label.replace(/ #\d+$/, '') + (++n > 1 ? ` #${n}` : '') },
);
// e.g. 'OpenAI', 'OpenAI #2', 'OpenAI #3' - uses max index + 1, never relabels existing
function _nextAutoLabelForVendor(vendorId: ModelVendorId, vendorName: string, existingServices: DModelsService[]): string {
const sameVendorServices = existingServices.filter(s => s.vId === vendorId);
if (sameVendorServices.length === 0)
return vendorName;
let maxIndex = 1;
for (const s of sameVendorServices) {
const match = s.label.match(/ #(\d+)$/);
if (match)
maxIndex = Math.max(maxIndex, parseInt(match[1], 10));
}
return `${vendorName} #${maxIndex + 1}`;
}
const { sources: existingServices, confServiceId } = get();
// create the service
const newService: DModelsService = {
id: _locallyUniqueServiceId(vendor.id, existingServices),
label: vendor.name,
label: _nextAutoLabelForVendor(vendor.id, vendor.name, existingServices),
vId: vendor.id,
setup: vendor.initializeSetup?.() || {},
};
const newServices = _relabelServicesFromSameVendor(vendor.id, [...existingServices, newService]);
set({
sources: newServices,
sources: [...existingServices, newService],
confServiceId: confServiceId ?? newService.id,
});
return newServices[newServices.length - 1];
return newService;
},
removeService: (id: DModelsServiceId) =>
@@ -245,6 +328,26 @@ export const useModelsStore = create<LlmsStore>()(persist(
};
}),
updateServiceLabel: (id: DModelsServiceId, label: string, allowEmpty: boolean = false) =>
set(state => {
// fallback label to vendor name if empty
if (!allowEmpty && !label.trim()) {
const service = state.sources.find(s => s.id === id);
const vendor = service ? findModelVendor(service.vId) : null;
label = vendor?.name || label;
}
// allow max of 32 chars for the name
if (label.length > 32)
label = label.substring(0, 32);
return {
sources: state.sources.map((s: DModelsService): DModelsService =>
s.id === id
? { ...s, label: label }
: s,
),
};
}),
updateServiceSettings: <TServiceSettings>(id: DModelsServiceId, partialSettings: Partial<TServiceSettings>) =>
set(state => ({
sources: state.sources.map((s: DModelsService): DModelsService =>

Some files were not shown because too many files have changed in this diff Show More