This website requires JavaScript.
bcc83f30d9
Properly count DS reasoning tokens and properly save them per-user
2025-01-25 11:06:38 +00:00
e5a26215e1
Add native Gemini model list endpoint
2025-01-24 08:34:12 +00:00
cd6cc76a46
Attempt to improve the o1 timeout hack
2025-01-22 15:23:17 +00:00
613bb789fb
fix error checking o1 deployments
2024-12-18 00:54:44 +00:00
f1c698388e
[Gemini] Support the new thinking config for 2.0 Flash Thinking
2025-01-22 09:59:11 +00:00
75605a2bfb
Add preliminary deepseek-reasoner support
2025-01-20 07:28:30 +00:00
58e67d40e2
Check logging for AWS keys (untested)
2025-01-20 11:28:53 +00:00
796b4eee47
Make the OpenAI checker properly clone orgs again, and fix the error with the liveness check
2025-01-06 06:55:41 +00:00
0f482e67d2
Fix OpenAI -> Google AI conversion
2025-01-05 14:02:17 +00:00
496ec09905
Add v1alpha support (needed for 2.0 flash thinking with the new 'thought' parameter), already used by ST
2025-01-04 19:11:02 +00:00
f522dba6a3
Fix errors with o1-preview and o1-mini
2025-01-03 04:41:29 +00:00
25ba8447d9
And add display names as well
2025-01-03 04:41:29 +00:00
91b8c01a9d
Do the same for the AWS endpoint
2025-01-03 04:41:29 +00:00
82b88764ba
Fix Anthropic model list to be actually compatible with Anthropic API (required by some frontends), remove old models
2025-01-03 04:50:55 +00:00
6ea9235ff8
Actually camelCase is canon for Gemini, oops
2024-12-31 08:23:45 +00:00
372ad85283
Support camelCase Gemini params and validate vision
2024-12-31 10:16:04 +00:00
c2f5d2fbf3
Add /v1/models to deepseek
2024-12-31 08:23:45 +00:00
c264413495
Leave a comment about concurrency for "special" user tokens
2024-12-31 08:23:45 +00:00
8d27082ad0
Fix formatting changes with upstream
2024-12-31 08:23:45 +00:00
e2b602fd52
Adjust chunked transfer to send 4KB (CF's buffer size) of data every 49 seconds
2024-12-31 08:23:45 +00:00
b00fb88cab
Don't overwrite the reasoning effort by default
2024-12-31 07:45:11 +00:00
1cc281f6fe
Add automatic prefill for Deepseek - works the same way as with Claude
2024-12-31 07:45:11 +00:00
8f4d00ed26
Init commit, some things: - 'Transfer-Encoding: chunked' for o1 requests to prevent CF's 100 second limit - Better tool/function call support - Deepseek support - Handling system as an array for AWS Claude - Image support for Gemini - Better o1 support (reasoning effort, developer role, context size)
2024-12-31 00:00:00 +00:00
75cb98c727
o1 "developer" role support and reasoning effort
2024-12-27 17:13:58 +00:00
14d11d4368
o1 200k context
2024-12-27 17:21:05 +00:00
2fbe0bff0d
Check for o1 instead of o1-preview for the o1 family
2024-12-27 17:17:39 +00:00
d5fd93f64a
Adjust release o1 context to 200K
2024-12-27 17:05:48 +00:00
36e2430a8f
adjusts gemini keychecker to trigger real generation for better rate limit detection
2024-12-07 01:31:10 -06:00
28447d0811
resolves server-side error when a Gemini prompt is blocked due to safety
2024-12-07 00:54:13 -06:00
6d54cbc785
maybe handles gemini daily key block idk
2024-12-05 15:06:29 -06:00
9d7a4f4b51
maybe fixes gemini's fucked error messages idk
2024-12-05 14:54:35 -06:00
3496a2a9bd
fixes incorrect 3.5 sonnet v2 model id in gcp model reassignment
2024-12-03 19:53:56 -06:00
5072638ec2
attempt at fixing persistent 'invalid csrf token' error on some browsers
2024-12-03 19:49:16 -06:00
8a325a1e0b
Add 2 million Google AI context (khanon/oai-reverse-proxy!82 )
2024-12-04 01:27:21 +00:00
5eeb2875b4
adds haiku 3.5
2024-11-04 18:20:19 +00:00
c67dad1617
fixes AWS claude v2
2024-10-25 18:10:26 +00:00
fe61745e24
fixes issue with AWS model assignment when requesting legacy claude (claude-2.1, etc)
2024-10-25 17:49:24 +00:00
251ea6d412
fixes typo in AWS ThrottlingException
2024-10-23 15:36:21 -05:00
55f7337ea4
adjusts AWS keychecker to treat rate limited models as available models
2024-10-23 15:35:32 -05:00
f3b876887e
fixes issue with AWS model name reassignment
2024-10-23 13:59:10 -05:00
49c578f4dc
adds Sonnet 3.5v2 AWS model ID and adjusts AWS model assignment to raise error on no match
2024-10-23 13:39:34 -05:00
4190d5fef6
fixes missing comma.......
2024-10-22 16:06:39 +00:00
1644e82f25
adds Sonnet 20241022 snapshot IDs
2024-10-22 15:52:52 +00:00
0bbdc0b841
fixes google ai language filter and updates readme
2024-10-18 23:15:02 -05:00
c4a633a5d6
fixes gcp oauth2 token refresh not updating cloned key instance
2024-10-18 22:41:24 -05:00
0c6ec3254f
finally DOES something about broken GCP streaming, boebeitfully
2024-10-12 20:10:59 -05:00
13aa55cd3d
handles gemini ai test message from sillytavern
2024-10-12 09:01:08 -05:00
ba4532b38d
more fixes for annoying gemini API design that allows arrays or single objects for contents parts
2024-10-09 17:11:53 -05:00
b57627e69b
adds stripHeaders to global mutators in createQueuedProxyMiddleware
2024-10-09 16:59:12 -05:00
536803853a
uses removeHeader instead of setHeader to empty string
2024-10-09 16:44:53 -05:00
ad0a3c0936
removes cors/sec-fetch headers to fix venus chub with anthropic api
2024-10-09 16:32:48 -05:00
161f5aba3e
handles sillytavern using both camel and snake-cased parameters for gemini api
2024-10-06 11:03:48 -05:00
514d1b7e31
fixes azure
2024-10-01 16:15:04 -05:00
22d7f966c6
fixes for gemini api streaming
2024-09-29 12:44:18 -05:00
cfb6353c65
updates google ai safety settings schema
2024-09-25 21:19:17 -05:00
a7fed3136e
fixes google ai gemini
2024-09-25 15:58:52 -05:00
29638cf26e
minor cleanup to pow challenge
2024-09-22 11:28:33 -05:00
ee26e7be65
various improvements and fixes to PoW challenge UI and token refresh
2024-09-22 11:11:30 -05:00
ff0d3dfdcd
prevents overwriting anthropic-version header if it's already provided
2024-09-19 00:55:17 -05:00
81a3ae1746
maybe fixes missing anthropic version header in some cases
2024-09-19 00:50:17 -05:00
4dfd57fcb4
updates render dockerfile to correctly copy patches dir into build context
2024-09-16 23:39:43 -05:00
d21e274358
Add configurable network interface or SOCKS/HTTP proxy for outgoing requests (khanon/oai-reverse-proxy!80 )
2024-09-16 15:17:57 +00:00
6e97e036b2
fixes refreshed PoW tokens not actually being reactivated
2024-09-15 18:01:23 -05:00
7a4a16dd2f
fixes chatgpt-latest missing from models endpoint
2024-09-15 06:02:35 -05:00
f1cfa644c5
maybe fixes openai sk-svcacct keys
2024-09-12 23:05:43 -05:00
6a908b09cb
adds preliminary openai o1 support and some improvements to openai keychecker
2024-09-12 22:55:45 -05:00
86772ab32a
adds 503 as a 'successful' AWS keychecker response to deal with temporary outages
2024-09-10 16:07:13 -05:00
bd87ca60f7
Implement priority queue by tokens (khanon/oai-reverse-proxy!79 )
2024-09-09 16:48:46 +00:00
ac1897fd17
returns more clear proxy_note hint on AWS 503 error
2024-09-09 09:56:18 -05:00
2a6f85e2e2
Revert "handles AWS HTTP 503 ServiceUnavailableException similarly to 429s"
2024-09-09 09:43:59 -05:00
ffcaa23511
handles AWS HTTP 503 ServiceUnavailableException similarly to 429s
2024-09-09 08:07:08 -05:00
1d5b8efa23
reduces key lockout period to more quickly drain queue after AWS rate limit resolves
2024-09-08 17:17:22 -05:00
905273abf2
fixes aws mistral token cost estimation
2024-09-08 17:15:59 -05:00
9e6fd7c24c
Implement tools (function calling) for Claude
2024-08-16 15:45:49 +03:00
ac92a19946
improves reliability of inference profile detection for AWS keychecker
2024-09-07 17:36:29 -05:00
96fe974ad0
Use AWS Inference Profiles for higher rate limits (khanon/oai-reverse-proxy!78 )
2024-09-01 22:55:07 +00:00
578615fbd2
fixes typo in new Claude system prompt schema
2024-08-30 10:23:57 -05:00
5dc4050e52
disable periodic GCP key rechecks to workaround keychecker bug
2024-08-29 15:25:37 -05:00
cf615ee62c
applies prettier to GCP checker
2024-08-29 15:15:56 -05:00
ee61f9be2b
removes unnecessary log from last commit
2024-08-27 23:58:32 -05:00
0c448cb59d
fixes azure dalle using wrong rate limit and out-of-spec Retry-After header
2024-08-27 23:43:11 -05:00
51a9ccceb2
supports alternate claude system prompt format
2024-08-27 23:27:20 -05:00
ce490efd7d
minor adjustments to HMAC signing
2024-08-22 19:53:53 -05:00
5000e59a61
fix for google makersuite prompt validation/transformation
2024-08-22 14:18:18 -05:00
27cbb7aaaf
Make deprecated function_call OpenAI field check less strict
2024-08-16 16:09:35 +03:00
d54acad6ad
adds support for sonnet 8192 output tokens on anthropic api
2024-08-15 11:55:13 -05:00
5e1fffe07d
adds chatgpt-4o-latest
2024-08-15 11:54:42 -05:00
f7fd5f00f2
fixes
2024-08-14 14:41:47 -05:00
6d323f6ea1
do not transform mistral chat prompts to text when using la plateforme
2024-08-14 12:26:27 -05:00
2959ed3f7f
fixes aws keychecker not detecting claude 2.1
2024-08-14 10:44:55 -05:00
b58e7cb830
always applies Mistral prompt fixes on messages input
2024-08-14 10:44:22 -05:00
cfc1290f83
fixes aws keychecker not detecting claude 2.1
2024-08-14 10:44:55 -05:00
14f228f666
always applies Mistral prompt fixes on messages input
2024-08-14 10:44:22 -05:00
f531272b00
Refactor AWS service code and add AWS Mistral support (khanon/oai-reverse-proxy!75 )
2024-08-14 04:40:41 +00:00
d264fdd573
adds mistral chat-to-text transformation, for better prefix compatibility
2024-08-13 23:16:31 -05:00
9c3e345720
update deps
2024-08-13 20:31:19 -05:00
37c421bb45
fixes token counting for streaming Mistral Text prompts
2024-08-13 20:27:31 -05:00
6c5fed90e2
rename function
2024-08-13 20:15:14 -05:00
9479fa4ab0
serviceinfo tweak
2024-08-13 20:13:46 -05:00
e145f5757e
implements aws mistral streaming
2024-08-13 20:04:02 -05:00
user