OpenRouter: compensate for older typescript

OpenRouter: models list: prevent schema changes from breaking working models. Fixes #787
OpenRouter: models list: ignore missing fields on 'openrouter/auto'. Fixes #787
2026-05-10 21:50:14 -07:00 · 2025-04-05 10:30:15 -07:00 · 2025-04-05 10:26:44 -07:00 · 2025-04-05 10:25:35 -07:00 · 2025-03-29 12:35:29 -07:00 · 2025-03-29 12:35:05 -07:00
42 changed files with 1038 additions and 379 deletions
@@ -51,8 +51,7 @@ jobs:
        with:
          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
          tags: |
-            type=raw,value=development,enable=${{ github.ref == 'refs/heads/main' }}
-            type=raw,value=stable,enable=${{ github.ref == 'refs/heads/main-stable' }}
+            type=raw,value=stable,enable=${{ github.ref == 'refs/heads/v1-stable' }}
            type=ref,event=tag  # Use the tag name as a tag for tag builds
            type=semver,pattern={{version}}  # Generate semantic versioning tags for tag builds
            type=sha # Just in case none of the above applies
@@ -11,27 +11,42 @@ Stay ahead of the curve with big-AGI. 🚀 Pros & Devs love big-AGI. 🤖

 [![Official Website](https://img.shields.io/badge/BIG--AGI.com-%23096bde?style=for-the-badge&logo=vercel&label=launch)](https://big-agi.com)

+> 🚀 Big-AGI 2 is launching Q4 2024. Be the first to experience it before the public release.
+>
+> 👉 [Apply for Early Access](https://y2rjg0zillz.typeform.com/to/ZSADpr5u?utm_source=gh-stable&utm_medium=readme&utm_campaign=ea2)
+
 Or fork & run on Vercel

 [![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?repository-url=https%3A%2F%2Fgithub.com%2Fenricoros%2Fbig-AGI&env=OPENAI_API_KEY&envDescription=Backend%20API%20keys%2C%20optional%20and%20may%20be%20overridden%20by%20the%20UI.&envLink=https%3A%2F%2Fgithub.com%2Fenricoros%2Fbig-AGI%2Fblob%2Fmain%2Fdocs%2Fenvironment-variables.md&project-name=big-AGI)

-## 👉 [roadmap](https://github.com/users/enricoros/projects/4/views/2) 👉 [installation](docs/installation.md) 👉 [documentation](docs/README.md)
+### New Version

-> Note: bigger better features (incl. Beam-2) are being cooked outside of `main`.
+This repository contains two main versions:

-[//]: # (big-AGI is an open book; see the **[ready-to-ship and future ideas]&#40;https://github.com/users/enricoros/projects/4/views/2&#41;** in our open roadmap)
+- Big-AGI 2: next-generation, bringing the most advanced AI experience
+  - `v2-dev`: V2 development branch, the exciting one, future default
+- Big-AGI Stable: as deployed on big-agi.com
+  - `v1-stable`: Current stable version & Docker 'latest' tag

-### What's New in 1.16.2 · Jun 7, 2024 (minor release)
+Note: After the V2 release in Q4, `v2/dev` will become the default branch and `v1/dev` will reach EOL.

- Improve web downloads, as text, markdwon, or HTML
- Proper support for Gemini models
- Added the latest Mistral model
- Tokenizer support for gpt-4o
- Updates to Beam
+### Quick links: 👉 [roadmap](https://github.com/users/enricoros/projects/4/views/2) 👉 [installation](docs/installation.md) 👉 [documentation](docs/README.md)

-### What's New in 1.16.1 · May 13, 2024 (minor release)
+### What's New in 1.16.1...1.16.9 · Jan 21, 2025 (patch releases)

- Support for the new OpenAI GPT-4o 2024-05-13 model
+- 1.16.9: Docker Gemini fix (R1 models are supported in Big-AGI 2)
+- 1.16.8: OpenAI ChatGPT-4o Latest (o1 models are supported in Big-AGI 2)
+- 1.16.7: OpenAI support for GPT-4o 2024-08-06
+- 1.16.6: Groq support for Llama 3.1 models
+- 1.16.5: GPT-4o Mini support
+- 1.16.4: 8192 tokens support for Claude 3.5 Sonnet
+- 1.16.3: Anthropic Claude 3.5 Sonnet model support
+- 1.16.2: Improve web downloads, as text, markdwon, or HTML
+- 1.16.2: Proper support for Gemini models
+- 1.16.2: Added the latest Mistral model
+- 1.16.2: Tokenizer support for gpt-4o
+- 1.16.2: Updates to Beam
+- 1.16.1: Support for the new OpenAI GPT-4o 2024-05-13 model

 ### What's New in 1.16.0 · May 9, 2024 · Crystal Clear

@@ -10,17 +10,21 @@ by release.
 - milestone: [1.17.0](https://github.com/enricoros/big-agi/milestone/17)
 - work in progress: [big-AGI open roadmap](https://github.com/users/enricoros/projects/4/views/2), [help here](https://github.com/users/enricoros/projects/4/views/4)

-### What's New in 1.16.2 · Jun 7, 2024 (minor release)
+### What's New in 1.16.1...1.16.9 · Jan 21, 2025 (patch releases)

- Improve web downloads, as text, markdwon, or HTML
- Proper support for Gemini models
- Added the latest Mistral model
- Tokenizer support for gpt-4o
- Updates to Beam
-
-### What's New in 1.16.1 · May 13, 2024 (minor release)
-
- Support for the new OpenAI GPT-4o 2024-05-13 model
+- 1.16.9: Docker Gemini fix (R1 models are supported in Big-AGI 2)
+- 1.16.8: OpenAI ChatGPT-4o Latest (o1 models are supported in Big-AGI 2)
+- 1.16.7: OpenAI support for GPT-4o 2024-08-06
+- 1.16.6: Groq support for Llama 3.1 models
+- 1.16.5: GPT-4o Mini support
+- 1.16.4: 8192 tokens support for Claude 3.5 Sonnet
+- 1.16.3: Anthropic Claude 3.5 Sonnet model support
+- 1.16.2: Improve web downloads, as text, markdwon, or HTML
+- 1.16.2: Proper support for Gemini models
+- 1.16.2: Added the latest Mistral model
+- 1.16.2: Tokenizer support for gpt-4o
+- 1.16.2: Updates to Beam
+- 1.16.1: Support for the new OpenAI GPT-4o 2024-05-13 model

 ### What's New in 1.16.0 · May 9, 2024 · Crystal Clear

@@ -79,7 +79,7 @@
        "typescript": "^5.4.5"
      },
      "engines": {
-        "node": "^20.0.0 || ^18.0.0"
+        "node": "^22.0.0 || ^20.0.0 || ^18.0.0"
      }
    },
    "node_modules/@babel/code-frame": {
@@ -88,6 +88,6 @@
    "typescript": "^5.4.5"
  },
  "engines": {
-    "node": "^20.0.0 || ^18.0.0"
+    "node": "^22.0.0 || ^20.0.0 || ^18.0.0"
  }
 }
@@ -277,7 +277,7 @@ export function AppChat() {
    const conversation = getConversation(conversationId);
    if (!conversation)
      return;
-    const imaginedPrompt = await imaginePromptFromText(messageText) || 'An error sign.';
+    const imaginedPrompt = await imaginePromptFromText(messageText, conversationId) || 'An error sign.';
    await handleExecuteAndOutcome('generate-image', conversationId, [
      ...conversation.messages,
      createDMessage('user', imaginedPrompt),
@@ -280,6 +280,7 @@ export function ChatMessage(props: {
  const wasEdited = !!messageUpdated;

  const textSel = selText ? selText : messageText;
+  // WARNING: if you get an issue here, you're downgrading from the new Big-AGI 2 data format to 1.x.
  const isSpecialT2I = textSel.startsWith('https://images.prodia.xyz/') || textSel.startsWith('/draw ') || textSel.startsWith('/imagine ') || textSel.startsWith('/img ');
  const couldDiagram = textSel.length >= 100 && !isSpecialT2I;
  const couldImagine = textSel.length >= 3 && !isSpecialT2I;
@@ -2,7 +2,7 @@ import type { DLLMId } from '~/modules/llms/store-llms';
 import type { StreamingClientUpdate } from '~/modules/llms/vendors/unifiedStreamingClient';
 import { autoSuggestions } from '~/modules/aifn/autosuggestions/autoSuggestions';
 import { conversationAutoTitle } from '~/modules/aifn/autotitle/autoTitle';
-import { llmStreamingChatGenerate, VChatContextRef, VChatContextName, VChatMessageIn } from '~/modules/llms/llm.client';
+import { llmStreamingChatGenerate, VChatContextRef, VChatMessageIn, VChatStreamContextName } from '~/modules/llms/llm.client';
 import { speakText } from '~/modules/elevenlabs/elevenlabs.client';

 import type { DMessage } from '~/common/state/store-chats';
@@ -63,7 +63,7 @@ type StreamMessageStatus = { outcome: StreamMessageOutcome, errorMessage?: strin
 export async function streamAssistantMessage(
  llmId: DLLMId,
  messagesHistory: VChatMessageIn[],
-  contextName: VChatContextName,
+  contextName: VChatStreamContextName,
  contextRef: VChatContextRef,
  throttleUnits: number, // 0: disable, 1: default throttle (12Hz), 2+ reduce the message frequency with the square root
  autoSpeak: ChatAutoSpeakType,
@@ -14,6 +14,7 @@ import { capitalizeFirstLetter } from '~/common/util/textUtils';

 import { NewsItems } from './news.data';
 import { beamNewsCallout } from './beam.data';
+import { bigAgi2NewsCallout } from './bigAgi2.data';


 // number of news items to show by default, before the expander
@@ -110,6 +111,13 @@ export function AppNews() {
            const addPadding = false; //!firstCard; // || showExpander;
            return <React.Fragment key={idx}>

+              {/* Inject the Big-AGI 2.0 item here*/}
+              {idx === 0 && (
+                <Box sx={{ mb: 3 }}>
+                  {bigAgi2NewsCallout}
+                </Box>
+              )}
+
              {/* Inject the Beam item here*/}
              {idx === 2 && (
                <Box sx={{ mb: 3 }}>
@@ -2,7 +2,6 @@ import * as React from 'react';

 import { Button, Card, CardContent, Grid, Typography } from '@mui/joy';
 import LaunchIcon from '@mui/icons-material/Launch';
-import ThumbUpRoundedIcon from '@mui/icons-material/ThumbUpRounded';

 import { Link } from '~/common/components/Link';

@@ -0,0 +1,40 @@
+import * as React from 'react';
+
+import { Button, Card, CardContent, Grid, Typography } from '@mui/joy';
+import AccessTimeIcon from '@mui/icons-material/AccessTime';
+import LaunchIcon from '@mui/icons-material/Launch';
+
+import { Link } from '~/common/components/Link';
+
+
+const bigAgi2SurveyUrl = 'https://y2rjg0zillz.typeform.com/to/ZSADpr5u?utm_source=gh-stable&utm_medium=news&utm_campaign=ea2';
+
+export const bigAgi2NewsCallout =
+  <Card variant='solid' invertedColors>
+    <CardContent sx={{ gap: 2 }}>
+      <Typography level='title-lg'>
+        Big-AGI 2.0 - In Development
+      </Typography>
+      <Typography level='body-sm'>
+        We&apos;re building the next version of Big-AGI with your needs in mind. New features, better performance, enhanced AI interactions. Help us shape it.
+      </Typography>
+      <Grid container spacing={1}>
+        <Grid xs={12} sm={7}>
+          <Button
+            fullWidth variant='soft' color='primary' endDecorator={<LaunchIcon />}
+            component={Link} href={bigAgi2SurveyUrl} noLinkStyle target='_blank'
+          >
+            Apply for Early Access
+          </Button>
+        </Grid>
+        <Grid xs={12} sm={5} sx={{ display: 'flex', flexAlign: 'center', justifyContent: 'center' }}>
+          <Button
+            fullWidth variant='outlined' color='primary' startDecorator={<AccessTimeIcon />}
+            disabled
+          >
+            Coming Fall 2024
+          </Button>
+        </Grid>
+      </Grid>
+    </CardContent>
+  </Card>;
@@ -61,7 +61,7 @@ export const NewsItems: NewsItem[] = [
    ]
  }*/
  {
-    versionCode: '1.16.2',
+    versionCode: '1.16.9',
    versionName: 'Crystal Clear',
    versionDate: new Date('2024-06-07T05:00:00Z'),
    // versionDate: new Date('2024-05-13T19:00:00Z'),
@@ -77,8 +77,16 @@ export const NewsItems: NewsItem[] = [
      { text: <>More: <B issue={517}>code soft-wrap</B>, selection toolbar, <B issue={507}>3x faster</B> on Apple silicon</>, issue: 507 },
      { text: <>Updated <B>Anthropic</B>*, <B>Groq</B>, <B>Ollama</B>, <B>OpenAI</B>*, <B>OpenRouter</B>*, and <B>Perplexity</B></> },
      { text: <>Developers: update LLMs data structures</>, dev: true },
-      { text: <>1.16.1: Support for <B>OpenAI</B> <B href='https://openai.com/index/hello-gpt-4o/'>GPT-4o</B> (refresh your OpenAI models)</> },
+      { text: <>1.16.1: Support for <B>OpenAI</B> <B href='https://openai.com/index/hello-gpt-4o/'>GPT-4o</B></> },
      { text: <>1.16.2: Proper <B>Gemini</B> support, <B>HTML/Markdown</B> downloads, and latest <B>Mistral</B></> },
+      { text: <>1.16.3: Support for <B href='https://www.anthropic.com/news/claude-3-5-sonnet'>Claude 3.5 Sonnet</B> (refresh your <B>Anthropic</B> models)</> },
+      { text: <>1.16.4: <B>8192 tokens</B> support for Claude 3.5 Sonnet</> },
+      { text: <>1.16.5: OpenAI <B>GPT-4o Mini</B> support</> },
+      { text: <>1.16.6: Groq <B>Llama 3.1</B> support</> },
+      { text: <>1.16.7: Gpt-4o <B>2024-08-06</B></> },
+      { text: <>1.16.8: <B>ChatGPT-4o</B> latest</> },
+      { text: <>1.16.9: <B>Gemini</B> fixes</> },
+      { text: <>OpenAI <B>o1</B>, DeepSeek R1, and newer models require Big-AGI 2. <B href='https://y2rjg0zillz.typeform.com/to/ZSADpr5u?utm_source=gh-stable&utm_medium=news&utm_campaign=ea2'>Sign up here</B></> },
    ],
  },
  {
@@ -7,7 +7,7 @@ import { useAppStateStore } from '~/common/state/store-appstate';


 // update this variable every time you want to broadcast a new version to clients
-export const incrementalNewsVersion: number = 16.1; // not notifying for 16.2
+export const incrementalNewsVersion: number = 16.1; // not notifying for 1.16.9


 interface NewsState {
@@ -8,8 +8,11 @@ export function prettyBaseModel(model: string | undefined): string {
  if (!model) return '';
  if (model.includes('gpt-4-vision-preview')) return 'GPT-4 Vision';
  if (model.includes('gpt-4-1106-preview')) return 'GPT-4 Turbo';
-  if (model.includes('gpt-4-32k')) return 'gpt-4-32k';
-  if (model.includes('gpt-4')) return 'gpt-4';
+  if (model.includes('gpt-4-32k')) return 'GPT-4-32k';
+  if (model.includes('gpt-4o-mini')) return 'GPT-4o Mini';
+  if (model.includes('gpt-4o')) return 'GPT-4o';
+  if (model.includes('gpt-4-turbo')) return 'GPT-4 Turbo';
+  if (model.includes('gpt-4')) return 'GPT-4';
  if (model.includes('gpt-3.5-turbo-instruct')) return '3.5 Turbo Instruct';
  if (model.includes('gpt-3.5-turbo-1106')) return '3.5 Turbo 16k';
  if (model.includes('gpt-3.5-turbo-16k')) return '3.5 Turbo 16k';
@@ -1,4 +1,4 @@
-import { llmChatGenerateOrThrow, VChatFunctionIn } from '~/modules/llms/llm.client';
+import { llmChatGenerateOrThrow, VChatFunctionIn, VChatMessageIn } from '~/modules/llms/llm.client';
 import { useModelsStore } from '~/modules/llms/store-llms';

 import { useChatStore } from '~/common/state/store-chats';
@@ -83,13 +83,18 @@ export function autoSuggestions(conversationId: string, assistantMessageId: stri

  // Follow-up: Auto-Diagrams
  if (suggestDiagrams) {
-    llmChatGenerateOrThrow(funcLLMId, [
-        { role: 'system', content: systemMessage.text },
-        { role: 'user', content: userMessage.text },
-        { role: 'assistant', content: assistantMessageText },
-      ], [suggestPlantUMLFn], 'draw_plantuml_diagram',
+    const instructions: VChatMessageIn[] = [
+      { role: 'system', content: systemMessage.text },
+      { role: 'user', content: userMessage.text },
+      { role: 'assistant', content: assistantMessageText },
+    ];
+    llmChatGenerateOrThrow(
+      funcLLMId,
+      instructions,
+      'chat-followup-diagram', conversationId,
+      [suggestPlantUMLFn], 'draw_plantuml_diagram',
    ).then(chatResponse => {
-
+      // cheap way to check if the function was supported
      if (!('function_arguments' in chatResponse))
        return;

@@ -1,5 +1,5 @@
 import { getFastLLMId } from '~/modules/llms/store-llms';
-import { llmChatGenerateOrThrow } from '~/modules/llms/llm.client';
+import { llmChatGenerateOrThrow, VChatMessageIn } from '~/modules/llms/llm.client';

 import { useChatStore } from '~/common/state/store-chats';

@@ -34,21 +34,23 @@ export async function conversationAutoTitle(conversationId: string, forceReplace

  try {
    // LLM chat-generate call
+    const instructions: VChatMessageIn[] = [
+      { role: 'system', content: `You are an AI conversation titles assistant who specializes in creating expressive yet few-words chat titles.` },
+      {
+        role: 'user', content:
+          'Analyze the given short conversation (every line is truncated) and extract a concise chat title that ' +
+          'summarizes the conversation in as little as a couple of words.\n' +
+          'Only respond with the lowercase short title and nothing else.\n' +
+          '\n' +
+          '```\n' +
+          historyLines.join('\n') +
+          '```\n',
+      },
+    ];
    const chatResponse = await llmChatGenerateOrThrow(
      fastLLMId,
-      [
-        { role: 'system', content: `You are an AI conversation titles assistant who specializes in creating expressive yet few-words chat titles.` },
-        {
-          role: 'user', content:
-            'Analyze the given short conversation (every line is truncated) and extract a concise chat title that ' +
-            'summarizes the conversation in as little as a couple of words.\n' +
-            'Only respond with the lowercase short title and nothing else.\n' +
-            '\n' +
-            '```\n' +
-            historyLines.join('\n') +
-            '```\n',
-        },
-      ],
+      instructions,
+      'chat-ai-title', conversationId,
      null, null,
    );

@@ -1,5 +1,5 @@
 import { getFastLLMId } from '~/modules/llms/store-llms';
-import { llmChatGenerateOrThrow } from '~/modules/llms/llm.client';
+import { llmChatGenerateOrThrow, VChatMessageIn } from '~/modules/llms/llm.client';


 const simpleImagineSystemPrompt =
@@ -10,14 +10,15 @@ Provide output as a lowercase prompt and nothing else.`;
 /**
 * Creates a caption for a drawing or photo given some description - used to elevate the quality of the imaging
 */
-export async function imaginePromptFromText(messageText: string): Promise<string | null> {
+export async function imaginePromptFromText(messageText: string, contextRef: string): Promise<string | null> {
  const fastLLMId = getFastLLMId();
  if (!fastLLMId) return null;
  try {
-    const chatResponse = await llmChatGenerateOrThrow(fastLLMId, [
+    const instructions: VChatMessageIn[] = [
      { role: 'system', content: simpleImagineSystemPrompt },
      { role: 'user', content: 'Write a prompt, based on the following input.\n\n```\n' + messageText.slice(0, 1000) + '\n```\n' },
-    ], null, null);
+    ];
+    const chatResponse = await llmChatGenerateOrThrow(fastLLMId, instructions, 'draw-expand-prompt', contextRef, null, null);
    return chatResponse.content?.trim() ?? null;
  } catch (error: any) {
    console.error('imaginePromptFromText: fetch request error:', error);
@@ -132,7 +132,7 @@ export class Agent {
    S.messages.push({ role: 'user', content: prompt });
    let content: string;
    try {
-      content = (await llmChatGenerateOrThrow(llmId, S.messages, null, null, 500)).content;
+      content = (await llmChatGenerateOrThrow(llmId, S.messages, 'chat-react-turn', null, null, null, 500)).content;
    } catch (error: any) {
      content = `Error in llmChatGenerateOrThrow: ${error}`;
    }
@@ -1,5 +1,5 @@
 import { DLLMId, findLLMOrThrow } from '~/modules/llms/store-llms';
-import { llmChatGenerateOrThrow } from '~/modules/llms/llm.client';
+import { llmChatGenerateOrThrow, VChatMessageIn } from '~/modules/llms/llm.client';


 // prompt to be tried when doing recursive summerization.
@@ -80,10 +80,11 @@ async function cleanUpContent(chunk: string, llmId: DLLMId, _ignored_was_targetW
  const autoResponseTokensSize = contextTokens ? Math.floor(contextTokens * outputTokenShare) : null;

  try {
-    const chatResponse = await llmChatGenerateOrThrow(llmId, [
+    const instructions: VChatMessageIn[] = [
      { role: 'system', content: cleanupPrompt },
      { role: 'user', content: chunk },
-    ], null, null, autoResponseTokensSize ?? undefined);
+    ];
+    const chatResponse = await llmChatGenerateOrThrow(llmId, instructions, 'chat-ai-summarize', null, null, null, autoResponseTokensSize ?? undefined);
    return chatResponse?.content ?? '';
  } catch (error: any) {
    return '';
@@ -1,7 +1,7 @@
 import * as React from 'react';

 import { DLLMId, findLLMOrThrow } from '~/modules/llms/store-llms';
-import { llmStreamingChatGenerate, VChatContextName, VChatContextRef, VChatMessageIn } from '~/modules/llms/llm.client';
+import { llmStreamingChatGenerate, VChatContextRef, VChatMessageIn, VChatStreamContextName } from '~/modules/llms/llm.client';


 // set to true to log to the console
@@ -20,7 +20,7 @@ export interface LLMChainStep {
 /**
 * React hook to manage a chain of LLM transformations.
 */
-export function useLLMChain(steps: LLMChainStep[], llmId: DLLMId | undefined, chainInput: string | undefined, onSuccess: (output: string, input: string) => void, contextName: VChatContextName, contextRef: VChatContextRef) {
+export function useLLMChain(steps: LLMChainStep[], llmId: DLLMId | undefined, chainInput: string | undefined, onSuccess: (output: string, input: string) => void, contextName: VChatStreamContextName, contextRef: VChatContextRef) {

  // state
  const [chain, setChain] = React.useState<ChainState | null>(null);
@@ -1,7 +1,7 @@
 import * as React from 'react';

 import type { DLLMId } from '~/modules/llms/store-llms';
-import { llmStreamingChatGenerate, VChatContextName, VChatContextRef, VChatMessageIn } from '~/modules/llms/llm.client';
+import { llmStreamingChatGenerate, VChatContextRef, VChatMessageIn, VChatStreamContextName } from '~/modules/llms/llm.client';


 export function useStreamChatText() {
@@ -13,7 +13,7 @@ export function useStreamChatText() {
  const abortControllerRef = React.useRef<AbortController | null>(null);


-  const startStreaming = React.useCallback(async (llmId: DLLMId, prompt: VChatMessageIn[], contextName: VChatContextName, contextRef: VChatContextRef) => {
+  const startStreaming = React.useCallback(async (llmId: DLLMId, prompt: VChatMessageIn[], contextName: VChatStreamContextName, contextRef: VChatContextRef) => {
    setStreamError(null);
    setPartialText(null);
    setText(null);
@@ -2,7 +2,7 @@ import { sendGAEvent } from '@next/third-parties/google';

 import { hasGoogleAnalytics } from '~/common/components/GoogleAnalytics';

-import type { ModelDescriptionSchema } from './server/llm.server.types';
+import type { GenerateContextNameSchema, ModelDescriptionSchema, StreamingContextNameSchema } from './server/llm.server.types';
 import type { OpenAIWire } from './server/openai/openai.wiretypes';
 import type { StreamingClientUpdate } from './vendors/unifiedStreamingClient';
 import { DLLM, DLLMId, DModelSource, DModelSourceId, LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, useModelsStore } from './store-llms';
@@ -21,14 +21,8 @@ export interface VChatMessageIn {

 export type VChatFunctionIn = OpenAIWire.ChatCompletion.RequestFunctionDef;

-export type VChatContextName =
-  | 'conversation'
-  | 'ai-diagram'
-  | 'ai-flattener'
-  | 'beam-scatter'
-  | 'beam-gather'
-  | 'call'
-  | 'persona-extract';
+export type VChatStreamContextName = StreamingContextNameSchema;
+export type VChatGenerateContextName = GenerateContextNameSchema;
 export type VChatContextRef = string;

 export interface VChatMessageOut {
@@ -122,7 +116,10 @@ function modelDescriptionToDLLMOpenAIOptions<TSourceSetup, TLLMOptions>(model: M
 export async function llmChatGenerateOrThrow<TSourceSetup = unknown, TAccess = unknown, TLLMOptions = unknown>(
  llmId: DLLMId,
  messages: VChatMessageIn[],
-  functions: VChatFunctionIn[] | null, forceFunctionName: string | null,
+  contextName: VChatGenerateContextName,
+  contextRef: VChatContextRef | null,
+  functions: VChatFunctionIn[] | null,
+  forceFunctionName: string | null,
  maxTokens?: number,
 ): Promise<VChatMessageOut | VChatMessageOrFunctionCallOut> {

@@ -146,14 +143,14 @@ export async function llmChatGenerateOrThrow<TSourceSetup = unknown, TAccess = u
    await new Promise(resolve => setTimeout(resolve, delay));

  // execute via the vendor
-  return await vendor.rpcChatGenerateOrThrow(access, options, messages, functions, forceFunctionName, maxTokens);
+  return await vendor.rpcChatGenerateOrThrow(access, options, messages, contextName, contextRef, functions, forceFunctionName, maxTokens);
 }


 export async function llmStreamingChatGenerate<TSourceSetup = unknown, TAccess = unknown, TLLMOptions = unknown>(
  llmId: DLLMId,
  messages: VChatMessageIn[],
-  contextName: VChatContextName,
+  contextName: VChatStreamContextName,
  contextRef: VChatContextRef,
  functions: VChatFunctionIn[] | null,
  forceFunctionName: string | null,
@@ -1,4 +1,5 @@
 import * as React from 'react';
+import TimeAgo from 'react-timeago';
 import { shallow } from 'zustand/shallow';

 import { Box, Button, ButtonGroup, Divider, FormControl, Input, Switch, Tooltip, Typography } from '@mui/joy';
@@ -132,10 +133,10 @@ export function LLMOptionsModal(props: { id: DLLMId, onClose: () => void }) {
            llm id: {llm.id}<br />
            context tokens: <b>{llm.contextTokens ? llm.contextTokens.toLocaleString() : 'not provided'}</b>{` · `}
            max output tokens: <b>{llm.maxOutputTokens ? llm.maxOutputTokens.toLocaleString() : 'not provided'}</b><br />
-            {!!llm.created && <>created: {(new Date(llm.created * 1000)).toLocaleString()}<br /></>}
+            {!!llm.created && <>created: <TimeAgo date={new Date(llm.created * 1000)} /><br /></>}
            {/*· tags: {llm.tags.join(', ')}*/}
            {!!llm.pricing && <>pricing: $<b>{llm.pricing.chatIn || '(unk) '}</b>/M in, $<b>{llm.pricing.chatOut || '(unk) '}</b>/M out<br /></>}
-            {!!llm.benchmark && <>benchmark: <b>{llm.benchmark.cbaElo?.toLocaleString() || '(unk) '}</b> CBA Elo<br /></>}
+            {/*{!!llm.benchmark && <>benchmark: <b>{llm.benchmark.cbaElo?.toLocaleString() || '(unk) '}</b> CBA Elo<br /></>}*/}
            config: {JSON.stringify(llm.options)}
          </Typography>
        </Box>}
@@ -4,14 +4,71 @@ import { LLM_IF_OAI_Chat, LLM_IF_OAI_Vision } from '../../store-llms';

 const roundTime = (date: string) => Math.round(new Date(date).getTime() / 1000);

-export const hardcodedAnthropicModels: ModelDescriptionSchema[] = [
+export const hardcodedAnthropicModels: (ModelDescriptionSchema & { isLegacy?: boolean })[] = [
+  // Claude 3.5 models - https://docs.anthropic.com/en/docs/about-claude/models
+  // {
+  //   id: 'claude-3.5-opus', // ...
+  //   label: 'Claude 3.5 Opus',
+  //   created: roundTime(?),
+  //   description: ?,
+  //   contextWindow: 200000 ?, // Characters
+  //   maxCompletionTokens: 4096 ?,
+  //   trainingDataCutoff: ?,
+  //   interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+  //   pricing: { chatIn: 15, chatOut: 75 },
+  //   benchmark: {
+  //     cbaElo: 1256, // Placeholder
+  //     cbaMmlu: 86.8, // Placeholder
+  //   },
+  // },
+  {
+    id: 'claude-3-5-sonnet-20241022',
+    label: 'Claude 3.5 Sonnet',
+    created: roundTime('2024-10-22 06:00'),
+    description: 'Most intelligent Claude model to date',
+    contextWindow: 200000, // Characters
+    maxCompletionTokens: 8192,
+    trainingDataCutoff: 'Apr 2024',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 3, chatOut: 15 },
+    benchmark: { cbaElo: 1269, cbaMmlu: 88.7 }, // moved from 3.5 Sonnet (Previous Version), TO UPDATE!!
+  },
+  {
+    id: 'claude-3-5-sonnet-20240620',
+    label: 'Claude 3.5 Sonnet (Previous)',
+    created: roundTime('2024-06-20 06:00'),
+    description: 'The most intelligent Claude model',
+    contextWindow: 200000, // Characters
+    maxCompletionTokens: 8192,
+    trainingDataCutoff: 'Apr 2024',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 3, chatOut: 15 },
+    benchmark: { cbaElo: 1269 - 0.1, cbaMmlu: 88.7 - 0.1 },
+    hidden: true,
+  },
+  // {
+  //   id: 'claude-3.5-haiku', // ...
+  //   label: 'Claude 3.5 Haiku',
+  //   created: roundTime(?),
+  //   description: ?,
+  //   contextWindow: 200000 ?, // Characters
+  //   maxCompletionTokens: 4096 ?,
+  //   trainingDataCutoff: ?,
+  //   interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+  //   pricing: { chatIn: 0.25, chatOut: 1.25 },
+  //   benchmark: {
+  //     cbaElo: 1181, // Placeholder
+  //     cbaMmlu: 75.2, // Placeholder
+  //   },
+  // },

-  // Claude-3 models - https://docs.anthropic.com/claude/docs/models-overview#model-comparison
+
+  // Claude 3 models
  {
    id: 'claude-3-opus-20240229',
    label: 'Claude 3 Opus',
    created: roundTime('2024-02-29'),
-    description: 'Most powerful model for highly complex tasks',
+    description: 'Powerful model for complex tasks',
    contextWindow: 200000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Aug 2023',
@@ -23,19 +80,21 @@ export const hardcodedAnthropicModels: ModelDescriptionSchema[] = [
    id: 'claude-3-sonnet-20240229',
    label: 'Claude 3 Sonnet',
    created: roundTime('2024-02-29'),
-    description: 'Ideal balance of intelligence and speed for enterprise workloads',
+    description: 'Balance of speed, cost, and performance',
    contextWindow: 200000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Aug 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
    pricing: { chatIn: 3, chatOut: 15 },
    benchmark: { cbaElo: 1203, cbaMmlu: 79 },
+    hidden: true,
+    isLegacy: true,
  },
  {
    id: 'claude-3-haiku-20240307',
    label: 'Claude 3 Haiku',
    created: roundTime('2024-03-07'),
-    description: 'Fastest and most compact model for near-instant responsiveness',
+    description: 'Fastest, most cost-effective model',
    contextWindow: 200000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Aug 2023',
@@ -55,6 +114,7 @@ export const hardcodedAnthropicModels: ModelDescriptionSchema[] = [
    interfaces: [LLM_IF_OAI_Chat],
    pricing: { chatIn: 8, chatOut: 24 },
    benchmark: { cbaElo: 1119 },
+    hidden: true,
  },
  {
    id: 'claude-2.0',
@@ -77,25 +137,6 @@ export const hardcodedAnthropicModels: ModelDescriptionSchema[] = [
    maxCompletionTokens: 4096,
    interfaces: [LLM_IF_OAI_Chat],
    pricing: { chatIn: 0.8, chatOut: 2.4 },
-  },
-  {
-    id: 'claude-instant-1.1',
-    label: 'Claude Instant 1.1',
-    created: roundTime('2023-03-14'),
-    description: 'Precise and fast',
-    contextWindow: 100000,
-    maxCompletionTokens: 2048,
-    interfaces: [LLM_IF_OAI_Chat],
-    hidden: true,
-  },
-  {
-    id: 'claude-1.3',
-    label: 'Claude 1.3',
-    created: roundTime('2023-03-14'),
-    description: 'Claude 1.3 is the latest version of Claude v1',
-    contextWindow: 100000,
-    maxCompletionTokens: 4096,
-    interfaces: [LLM_IF_OAI_Chat],
    hidden: true,
  },
 ];
@@ -8,7 +8,7 @@ import { fetchJsonOrTRPCError } from '~/server/api/trpc.router.fetchers';
 import { fixupHost } from '~/common/util/urlUtils';

 import { OpenAIHistorySchema, openAIHistorySchema, OpenAIModelSchema, openAIModelSchema } from '../openai/openai.router';
-import { llmsChatGenerateOutputSchema, llmsListModelsOutputSchema } from '../llm.server.types';
+import { llmsChatGenerateOutputSchema, llmsGenerateContextSchema, llmsListModelsOutputSchema } from '../llm.server.types';

 import { AnthropicWireMessagesRequest, anthropicWireMessagesRequestSchema, AnthropicWireMessagesResponse, anthropicWireMessagesResponseSchema } from './anthropic.wiretypes';
 import { hardcodedAnthropicModels } from './anthropic.models';
@@ -17,7 +17,9 @@ import { hardcodedAnthropicModels } from './anthropic.models';
 // Default hosts
 const DEFAULT_API_VERSION_HEADERS = {
  'anthropic-version': '2023-06-01',
-  'anthropic-beta': 'messages-2023-12-15',
+  // Former Betas:
+  // - messages-2023-12-15: to use the Messages API
+  'anthropic-beta': 'max-tokens-3-5-sonnet-2024-07-15',
 };
 const DEFAULT_MAX_TOKENS = 2048;
 const DEFAULT_ANTHROPIC_HOST = 'api.anthropic.com';
@@ -158,7 +160,11 @@ const listModelsInputSchema = z.object({

 const chatGenerateInputSchema = z.object({
  access: anthropicAccessSchema,
-  model: openAIModelSchema, history: openAIHistorySchema,
+  model: openAIModelSchema,
+  history: openAIHistorySchema,
+  // functions: openAIFunctionsSchema.optional(),
+  // forceFunctionName: z.string().optional(),
+  context: llmsGenerateContextSchema.optional(),
 });


@@ -1,6 +1,10 @@
 import type { GeminiModelSchema } from './gemini.wiretypes';
 import type { ModelDescriptionSchema } from '../llm.server.types';
-import { LLM_IF_OAI_Chat, LLM_IF_OAI_Json, LLM_IF_OAI_Vision } from '../../store-llms';
+import { LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json, LLM_IF_OAI_Vision } from '../../store-llms';
+
+
+// dev options
+const DEV_DEBUG_GEMINI_MODELS = false;


 // supported interfaces
@@ -8,7 +12,21 @@ const geminiChatInterfaces: GeminiModelSchema['supportedGenerationMethods'] = ['

 // unsupported interfaces
 const filterUnallowedNames = ['Legacy'];
-const filterUnallowedInterfaces: GeminiModelSchema['supportedGenerationMethods'] = ['generateAnswer', 'embedContent', 'embedText'];
+const filterUnallowedInterfaces: GeminiModelSchema['supportedGenerationMethods'] = [
+  'generateAnswer',   // e.g. removes "models/aqa"
+  'embedContent',     // e.g. removes "models/embedding-001"
+  'embedText',        // e.g. removes "models/text-embedding-004"
+  'predict',          // e.g. removes "models/imagen-3.0-generate-002" (appeared on 2025-02-09)
+];
+const filterLyingModelNames: GeminiModelSchema['name'][] = [
+  // 2025-02-27: verified, old model is no more
+  'models/gemini-2.0-flash-exp', // verified, replaced by gemini-2.0-flash, which is non-free anymore
+
+  // 2025-02-09 update: as of now they cleared the list, so we restart
+  // 2024-12-10: name of models that are not what they say they are (e.g. 1114 is actually 1121 as of )
+  'models/gemini-1.5-flash-8b-exp-0924', // replaced by non-free
+  'models/gemini-1.5-flash-8b-exp-0827', // replaced by non-free
+];


 /* Manual models details
@@ -16,172 +34,405 @@ const filterUnallowedInterfaces: GeminiModelSchema['supportedGenerationMethods']
   - Latest version    gemini-1.0-pro-latest    <model>-<generation>-<variation>-latest
   - Latest stable     version  gemini-1.0-pro  <model>-<generation>-<variation>
   - Stable versions   gemini-1.0-pro-001       <model>-<generation>-<variation>-<version>
+
+   Gemini capabilities chart (updated 2024-10-01):
+   - [table stakes] System instructions
+   - JSON Mode, with optional JSON Schema [NOTE: JSON Schema is poorly supported?]
+   - Adjustable Safety Settings
+   - Caching
+   - Tuning
+   - [good] Function calling, with configuration
+   - [great] Code execution
 */
+
+// Experimental Gemini models are Free of charge
+const geminiExpPricingFree: ModelDescriptionSchema['pricing'] = {
+  // input: 'free', output: 'free',
+};
+
+const gemini20FlashPricing: ModelDescriptionSchema['pricing'] = {
+  chatIn: 0.10, // inputAudio: 0.70,
+  chatOut: 0.40,
+};
+
+const gemini20FlashLitePricing: ModelDescriptionSchema['pricing'] = {
+  chatIn: 0.075,
+  chatOut: 0.30,
+};
+
+const gemini15FlashPricing: ModelDescriptionSchema['pricing'] = {
+  chatIn: 0.075,
+  chatOut: 0.30,
+};
+
+const gemini15Flash8BPricing: ModelDescriptionSchema['pricing'] = {
+  chatIn: 0.0375,
+  chatOut: 0.15,
+};
+
+const gemini15ProPricing: ModelDescriptionSchema['pricing'] = {
+  chatIn: 1.25,
+  chatOut: 5.00,
+};
+
+
 const _knownGeminiModels: ({
  id: string,
+  labelOverride?: string,
  isNewest?: boolean,
-  isPreview?: boolean
-  symLink?: string
-} & Pick<ModelDescriptionSchema, 'interfaces' | 'pricing' | 'trainingDataCutoff' | 'hidden'>)[] = [
+  isPreview?: boolean,
+  symLink?: string,
+  deprecated?: string, // Gemini may provide deprecation dates
+  _delete?: boolean, // some gemini models are not acknowledged by Google Docs anymore, and leaving them in the list will confuse users
+} & Pick<ModelDescriptionSchema, 'interfaces' | 'pricing' | 'hidden' | 'benchmark'>)[] = [

-  // Generation 1.5
+  /// Generation 2.5
+
+  // 2.5 Pro Experimental
+  {
+    id: 'models/gemini-2.5-pro-exp-03-25',
+    isPreview: true,
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1443 },
+  },
+
+  /// Generation 2.0
+
+  // 2.0 Experimental - Pro
+  {
+    hidden: true, // showing the 2.5 instead
+    id: 'models/gemini-2.0-pro-exp-02-05', // Base model: Gemini 2.0 Pro
+    isPreview: true,
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1380 },
+  },
+  {
+    hidden: true, // only keeping the latest
+    id: 'models/gemini-2.0-pro-exp',
+    symLink: 'models/gemini-2.0-pro-exp-02-05',
+    // copied from symlink
+    isPreview: true,
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1380 },
+  },
+  {
+    _delete: true, // replaced by gemini-2.0-pro-exp-02-05, 2025-02-27: verified, old model is no more
+    id: 'models/gemini-exp-1206',
+    labelOverride: 'Gemini 2.0 Pro Experimental 1206',
+    isPreview: true,
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1373 },
+  },
+
+  // 2.0 Experimental - Flash Thinking
+  {
+    hidden: true, // only keeping the latest
+    id: 'models/gemini-2.0-flash-thinking-exp', // alias to the latest Flash Thinking model
+    labelOverride: 'Gemini 2.0 Flash Thinking Experimental',
+    symLink: 'models/gemini-2.0-flash-thinking-exp-01-21',
+    // copied from symlink
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    benchmark: { cbaElo: 1385 },
+    isPreview: true,
+  },
+  {
+    id: 'models/gemini-2.0-flash-thinking-exp-01-21',
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    benchmark: { cbaElo: 1385 },
+    isPreview: true,
+  },
+  {
+    hidden: true, // replaced by gemini-2.0-flash-thinking-exp-01-21 - 2025-02-27: seems still different on the API, hence no deletion yet
+    id: 'models/gemini-2.0-flash-thinking-exp-1219',
+    labelOverride: 'Gemini 2.0 Flash Thinking Experimental 12-19',
+    pricing: geminiExpPricingFree,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    benchmark: { cbaElo: 1363 },
+    isPreview: true,
+  },
+
+  // 2.0 Experimental - Flash Image Generation
+  {
+    id: 'models/gemini-2.0-flash-exp-image-generation',
+    // labelOverride: 'Gemini 2.0 Flash Native Image Generation',
+    pricing: geminiExpPricingFree,
+    interfaces: [
+      LLM_IF_OAI_Chat, LLM_IF_OAI_Vision,
+      // LLM_IF_HOTFIX_StripSys0, // This first Gemini Image Generation model does not support the developer instruction
+    ],
+    isPreview: true,
+  },
+
+  // 2.0 Flash
+  {
+    id: 'models/gemini-2.0-flash-001',
+    pricing: gemini20FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1358 },
+  },
+  {
+    id: 'models/gemini-2.0-flash',
+    symLink: 'models/gemini-2.0-flash-001',
+    // copied from symlink
+    pricing: gemini20FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1358 },
+  },
+
+  // 2.0 Flash Lite
+  {
+    id: 'models/gemini-2.0-flash-lite',
+    pricing: gemini20FlashLitePricing,
+    symLink: 'models/gemini-2.0-flash-lite-001',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1309 },
+  },
+  {
+    id: 'models/gemini-2.0-flash-lite-001',
+    pricing: gemini20FlashLitePricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1309 },
+  },
+  {
+    hidden: true, // discouraged, as the official is out
+    id: 'models/gemini-2.0-flash-lite-preview-02-05',
+    isPreview: true,
+    pricing: gemini20FlashLitePricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1309 },
+  },
+  {
+    id: 'models/gemini-2.0-flash-lite-preview',
+    symLink: 'models/gemini-2.0-flash-lite-preview-02-05',
+    // coped from symlink
+    isPreview: true,
+    pricing: gemini20FlashLitePricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1309 },
+  },
+
+
+  /// Generation 1.5
+
+  // Gemini 1.5 Flash Models
  {
    id: 'models/gemini-1.5-flash-latest', // updated regularly and might be a preview version
-    isNewest: true,
    isPreview: true,
-    pricing: {
-      chatIn: 0.70,   // 0.35 up to 128k tokens, 0.70 prompts > 128k tokens
-      chatOut: 2.10,  // 1.05 up to 128k tokens, 2.10 prompts > 128k tokens
-    },
-    trainingDataCutoff: 'May 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json], // input: audio, images and text
+    pricing: gemini15FlashPricing,
+    // symLink: '-002 or newer',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
  },
  {
    id: 'models/gemini-1.5-flash',
-    // copied from above
-    pricing: {
-      chatIn: 0.70,   // 0.35 up to 128k tokens, 0.70 prompts > 128k tokens
-      chatOut: 2.10,  // 1.05 up to 128k tokens, 2.10 prompts > 128k tokens
-    },
-    trainingDataCutoff: 'Apr 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json], // input: audio, images and text
+    // Defaults to version 002 on Oct 8, 2024
+    symLink: 'models/gemini-1.5-flash-002',
+    pricing: gemini15FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1271 },
+  },
+  {
+    id: 'models/gemini-1.5-flash-002', // new stable version
+    pricing: gemini15FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1271 },
    hidden: true,
  },
  {
-    id: 'models/gemini-1.5-flash-001',
-    // copied from above
-    pricing: {
-      chatIn: 0.70,   // 0.35 up to 128k tokens, 0.70 prompts > 128k tokens
-      chatOut: 2.10,  // 1.05 up to 128k tokens, 2.10 prompts > 128k tokens
-    },
-    trainingDataCutoff: 'Apr 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json], // input: audio, images and text
+    id: 'models/gemini-1.5-flash-001', // previous stable version
+    pricing: gemini15FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1227 },
+    hidden: true,
+  },
+  {
+    id: 'models/gemini-1.5-flash-001-tuning', // supports model tuning
+    pricing: gemini15FlashPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn /* Tuning ... */],
    hidden: true,
  },

+  // Gemini 1.5 Flash-8B Models
  {
-    id: 'models/gemini-1.5-pro-latest', // updated regularly and might be a preview version
-    isNewest: true,
-    isPreview: true,
-    pricing: {
-      chatIn: 7.00,   // $3.50 / 1 million tokens (for prompts up to 128K tokens), $7.00 / 1 million tokens (for prompts longer than 128K)
-      chatOut: 21.00, // $10.50 / 1 million tokens (128K or less), $21.00 / 1 million tokens (128K+)
-    },
-    trainingDataCutoff: 'May 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json], // input: audio, images and text
+    id: 'models/gemini-1.5-flash-8b-latest',
+    isPreview: false,
+    pricing: gemini15Flash8BPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
  },
  {
-    id: 'models/gemini-1.5-pro', // latest stable -> 001
-    // copied from above
-    pricing: {
-      chatIn: 7.00,   // $3.50 / 1 million tokens (for prompts up to 128K tokens), $7.00 / 1 million tokens (for prompts longer than 128K)
-      chatOut: 21.00, // $10.50 / 1 million tokens (128K or less), $21.00 / 1 million tokens (128K+)
-    },
-    trainingDataCutoff: 'Apr 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json],
+    id: 'models/gemini-1.5-flash-8b',
+    symLink: 'models/gemini-1.5-flash-8b-001',
+    pricing: gemini15Flash8BPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1213 },
+  },
+  {
+    id: 'models/gemini-1.5-flash-8b-001',
+    pricing: gemini15Flash8BPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1213 },
+    hidden: true,
+  },
+
+  // Gemini 1.5 Pro Models
+  {
+    id: 'models/gemini-1.5-pro-latest', // updated to latest stable version
+    pricing: gemini15ProPricing,
+    // symLink: '-002 or newer',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+  },
+  {
+    id: 'models/gemini-1.5-pro',
+    symLink: 'models/gemini-1.5-pro-002',
+    pricing: gemini15ProPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1302 },
+  },
+  {
+    id: 'models/gemini-1.5-pro-002',
+    pricing: gemini15ProPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1302 },
    hidden: true,
  },
  {
-    id: 'models/gemini-1.5-pro-001', // stable snapshot
-    // copied from above
-    pricing: {
-      chatIn: 7.00,   // $3.50 / 1 million tokens (for prompts up to 128K tokens), $7.00 / 1 million tokens (for prompts longer than 128K)
-      chatOut: 21.00, // $10.50 / 1 million tokens (128K or less), $21.00 / 1 million tokens (128K+)
-    },
-    trainingDataCutoff: 'Apr 2024',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json],
+    id: 'models/gemini-1.5-pro-001',
+    pricing: gemini15ProPricing,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Json, LLM_IF_OAI_Fn],
+    benchmark: { cbaElo: 1260 },
    hidden: true,
  },


-  // Generation 1.0
-  {
-    id: 'models/gemini-1.0-pro-latest',
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    id: 'models/gemini-1.0-pro',
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
-    interfaces: [LLM_IF_OAI_Chat],
-    hidden: true,
-  },
-  {
-    id: 'models/gemini-1.0-pro-001',
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
-    interfaces: [LLM_IF_OAI_Chat],
-    hidden: true,
-  },
+  /// Generation 1.0

-  // Generation 1.0 + Vision
+  // Gemini 1.0 Pro Vision Model
  {
    id: 'models/gemini-1.0-pro-vision-latest',
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision], // Text and Images
-    hidden: true,
-  },
-
-  // Older symlinks
-  {
-    id: 'models/gemini-pro',
-    symLink: 'models/gemini-1.0-pro',
-    // copied from symlinked
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
-    interfaces: [LLM_IF_OAI_Chat],
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
    hidden: true,
+    _delete: true, // confusing
  },
  {
    id: 'models/gemini-pro-vision',
-    // copied from symlinked
    symLink: 'models/gemini-1.0-pro-vision',
-    pricing: {
-      chatIn: 0.50,
-      chatOut: 1.50,
-    },
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision], // Text and Images
-    hidden: true,
+    _delete: true, // confusing
  },
+
+
+  /// Experimental
+
+  // LearnLM Experimental Model
+  {
+    id: 'models/learnlm-1.5-pro-experimental',
+    isPreview: true,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: geminiExpPricingFree,
+    // hidden: true,
+    // _delete: true,
+  },
+
+  {
+    id: 'models/gemma-3-27b-it',
+    isPreview: true,
+    interfaces: [
+      LLM_IF_OAI_Chat,
+      // LLM_IF_HOTFIX_StripImages, /* "Image input modality is not enabled for models/gemma-3-27b-it" */
+      // LLM_IF_HOTFIX_Sys0ToUsr0, /* "Developer instruction is not enabled for models/gemma-3-27b-it" */
+    ],
+    // pricing: geminiExpPricingFree,
+    // hidden: true,
+    // _delete: true,
+  },
+
 ];


 export function geminiFilterModels(geminiModel: GeminiModelSchema): boolean {
  const isAllowed = !filterUnallowedNames.some(name => geminiModel.displayName.includes(name));
  const isSupported = !filterUnallowedInterfaces.some(iface => geminiModel.supportedGenerationMethods.includes(iface));
-  return isAllowed && isSupported;
+  const isWhatItSaysItIs = !filterLyingModelNames.includes(geminiModel.name);
+  return isAllowed && isSupported && isWhatItSaysItIs;
 }

+
+const _sortOderIdPrefix: string[] = [
+  'models/gemini-exp',
+  'models/gemini-2.5-pro',
+  'models/gemini-2.0-pro',
+  'models/gemini-2.0-flash-exp-image-generation',
+  'models/gemini-2.0-flash-thinking',
+  'models/gemini-2.0-flash-0',
+  'models/gemini-2.0-flash',
+  'models/gemini-2.0-flash-lite',
+  'models/gemini-1.5-pro',
+  'models/gemini-1.5-flash',
+  'models/gemini-1.5-flash-8b',
+  'models/gemini-1.0-pro',
+  'models/gemini-pro',
+  'models/gemma',
+  'models/learnlm',
+] as const;
+
 export function geminiSortModels(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
+  // links to the bottom
+  const aIsLink = a.label.startsWith('🔗');
+  const bIsLink = b.label.startsWith('🔗');
+  if (aIsLink && !bIsLink) return 1;
+  if (!aIsLink && bIsLink) return -1;
+
  // hidden to the bottom, then names descending
-  if (a.hidden && !b.hidden) return 1;
-  if (!a.hidden && b.hidden) return -1;
+  // if (a.hidden && !b.hidden) return 1;
+  // if (!a.hidden && b.hidden) return -1;
+
+  // models beginning with 'gemini-' to the top
+  // const aGemini = a.label.startsWith('Gemini');
+  // const bGemini = b.label.startsWith('Gemini');
+  // if (aGemini && !bGemini) return -1;
+  // if (!aGemini && bGemini) return 1;
+
+  // sort by sortOrderIdPrefix
+  const aSortIdx = _sortOderIdPrefix.findIndex(p => a.id.startsWith(p));
+  const bSortIdx = _sortOderIdPrefix.findIndex(p => b.id.startsWith(p));
+  if (aSortIdx !== -1 && bSortIdx !== -1) {
+    if (aSortIdx < bSortIdx) return -1;
+    if (aSortIdx > bSortIdx) return 1;
+  }
+
+  // sort by label descending
  return b.label.localeCompare(a.label);
 }

-export function geminiModelToModelDescription(geminiModel: GeminiModelSchema): ModelDescriptionSchema {
+
+export function geminiModelToModelDescription(geminiModel: GeminiModelSchema): ModelDescriptionSchema | null {
  const { description, displayName, name: modelId, supportedGenerationMethods } = geminiModel;

+  // if (DEV_DEBUG_GEMINI_MODELS)
+  //   console.log('geminiModelToModelDescription', geminiModel);
+
  // find known manual mapping
  const knownModel = _knownGeminiModels.find(m => m.id === modelId);
+  if (!knownModel && DEV_DEBUG_GEMINI_MODELS)
+    console.warn('geminiModelToModelDescription: unknown model', modelId, geminiModel);
+
+  // handle _delete
+  if (knownModel?._delete)
+    return null;

  // handle symlinks
-  const label = knownModel?.symLink
-    ? `🔗 ${displayName.replace('1.0', '')} → ${knownModel.symLink}`
-    : displayName;
+  let label = knownModel?.symLink
+    ? `🔗 ${knownModel?.labelOverride || displayName} → ${knownModel.symLink}`
+    : knownModel?.labelOverride || displayName;
+
+  // FIX: the Gemini 1114 model now returns 1121 as the version.. highlight the issue
+  // if (geminiModel.name.endsWith('1114') && label.endsWith('1121'))
+  //   label += ' (really: 1114)';

  // handle hidden models
  const hasChatInterfaces = supportedGenerationMethods.some(iface => geminiChatInterfaces.includes(iface));
@@ -193,14 +444,13 @@ export function geminiModelToModelDescription(geminiModel: GeminiModelSchema): M

  // description
  const { version, topK, topP, temperature } = geminiModel;
-  const descriptionLong = description + ` (Version: ${version}, Defaults: temperature=${temperature}, topP=${topP}, topK=${topK}, interfaces=[${supportedGenerationMethods.join(',')}])`;
+  const descriptionLong = (description || 'No description.') + ` (Version: ${version}, Defaults: temperature=${temperature}, topP=${topP}, topK=${topK}, interfaces=[${supportedGenerationMethods.join(',')}])`;

  // use known interfaces, or add chat if this is a generateContent model
  const interfaces: ModelDescriptionSchema['interfaces'] = knownModel?.interfaces || [];
  if (!interfaces.length && hasChatInterfaces) {
-    interfaces.push(LLM_IF_OAI_Chat);
-    // if (geminiVisionNames.some(name => modelId.includes(name)))
-    //   interfaces.push(LLM_IF_OAI_Vision);
+    // newer models get good capabilities by default
+    interfaces.push(LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision, LLM_IF_OAI_Json);
  }

  return {
@@ -211,11 +461,13 @@ export function geminiModelToModelDescription(geminiModel: GeminiModelSchema): M
    description: descriptionLong,
    contextWindow: contextWindow,
    maxCompletionTokens: outputTokenLimit,
-    trainingDataCutoff: knownModel?.trainingDataCutoff,
+    // trainingDataCutoff: knownModel?.trainingDataCutoff, // disabled as we don't get this from Gemini
    interfaces,
+    // parameterSpecs: knownModel?.parameterSpecs,
    // rateLimits: isGeminiPro ? { reqPerMinute: 60 } : undefined,
-    // benchmarks: ...
-    pricing: knownModel?.pricing, // TODO: needs <>128k, and per-character and per-image pricing
+    benchmark: knownModel?.benchmark,
+    pricing: knownModel?.pricing,
    hidden,
+    // deprecated: knownModel?.deprecated,
  };
 }
@@ -8,7 +8,7 @@ import { createTRPCRouter, publicProcedure } from '~/server/api/trpc.server';
 import { fetchJsonOrTRPCError } from '~/server/api/trpc.router.fetchers';

 import { fixupHost } from '~/common/util/urlUtils';
-import { llmsChatGenerateOutputSchema, llmsListModelsOutputSchema } from '../llm.server.types';
+import { llmsChatGenerateOutputSchema, llmsGenerateContextSchema, llmsListModelsOutputSchema, type ModelDescriptionSchema } from '../llm.server.types';

 import { OpenAIHistorySchema, openAIHistorySchema, OpenAIModelSchema, openAIModelSchema } from '../openai/openai.router';

@@ -120,8 +120,11 @@ const accessOnlySchema = z.object({

 const chatGenerateInputSchema = z.object({
  access: geminiAccessSchema,
-  model: openAIModelSchema, history: openAIHistorySchema,
-  // functions: openAIFunctionsSchema.optional(), forceFunctionName: z.string().optional(),
+  model: openAIModelSchema,
+  history: openAIHistorySchema,
+  // functions: openAIFunctionsSchema.optional(),
+  // forceFunctionName: z.string().optional(),
+  context: llmsGenerateContextSchema.optional(),
 });


@@ -145,9 +148,10 @@ export const llmGeminiRouter = createTRPCRouter({
      //       as the List API already all the info on all the models

      // map to our output schema
-      const models = detailedModels
+      const models = (detailedModels
        .filter(geminiFilterModels)
        .map(geminiModel => geminiModelToModelDescription(geminiModel))
+        .filter(model => !!model) as ModelDescriptionSchema[])
        .sort(geminiSortModels);

      return {
@@ -9,27 +9,30 @@ export const geminiModelsStreamGenerateContentPath = '/v1beta/{model=models/*}:s


 // models.list = /v1beta/models
+const Methods_enum = z.enum([
+  'bidiGenerateContent', // appeared on 2024-12, see https://github.com/enricoros/big-AGI/issues/700
+  'createCachedContent', // appeared on 2024-06-10, see https://github.com/enricoros/big-AGI/issues/565
+  'countMessageTokens',
+  'countTextTokens',
+  'countTokens',
+  'createTunedModel',
+  'createTunedTextModel',
+  'embedContent',
+  'embedText',
+  'generateAnswer',
+  'generateContent',
+  'generateMessage',
+  'generateText',
+]);

 const geminiModelSchema = z.object({
  name: z.string(),
  version: z.string(),
  displayName: z.string(),
-  description: z.string(),
+  description: z.string().optional(),
  inputTokenLimit: z.number().int().min(1),
  outputTokenLimit: z.number().int().min(1),
-  supportedGenerationMethods: z.array(z.enum([
-    'countMessageTokens',
-    'countTextTokens',
-    'countTokens',
-    'createTunedModel',
-    'createTunedTextModel',
-    'embedContent',
-    'embedText',
-    'generateAnswer',
-    'generateContent',
-    'generateMessage',
-    'generateText',
-  ])),
+  supportedGenerationMethods: z.array(z.union([Methods_enum, z.string()])), // relaxed with z.union to not break on expansion
  temperature: z.number().optional(),
  topP: z.number().optional(),
  topK: z.number().optional(),
@@ -171,7 +174,7 @@ export const geminiGeneratedContentResponseSchema = z.object({
  // either all requested candidates are returned or no candidates at all
  // no candidates are returned only if there was something wrong with the prompt (see promptFeedback)
  candidates: z.array(z.object({
-    index: z.number(),
+    index: z.number().optional(),
    content: geminiContentSchema.optional(), // this can be missing if the finishReason is not 'MAX_TOKENS'
    finishReason: geminiFinishReasonSchema.optional(),
    safetyRatings: z.array(geminiSafetyRatingSchema).optional(), // undefined when finishReason is 'RECITATION'
@@ -22,6 +22,9 @@ import type { OpenAIWire } from './openai/openai.wiretypes';
 import { openAIAccess, openAIAccessSchema, openAIChatCompletionPayload, openAIHistorySchema, openAIModelSchema } from './openai/openai.router';


+import { llmsStreamingContextSchema } from './llm.server.types';
+
+
 // configuration
 const USER_SYMBOL_MAX_TOKENS = '🧱';
 const USER_SYMBOL_PROMPT_BLOCKED = '🚫';
@@ -46,17 +49,14 @@ type MuxingFormat = 'sse' | 'json-nl';
 */
 type AIStreamParser = (data: string, eventType?: string) => { text: string, close: boolean };

-const streamingContextSchema = z.object({
-  method: z.literal('chat-stream'),
-  name: z.enum(['conversation', 'ai-diagram', 'ai-flattener', 'call', 'beam-scatter', 'beam-gather', 'persona-extract']),
-  ref: z.string(),
-});

 const chatStreamingInputSchema = z.object({
  access: z.union([anthropicAccessSchema, geminiAccessSchema, ollamaAccessSchema, openAIAccessSchema]),
  model: openAIModelSchema,
  history: openAIHistorySchema,
-  context: streamingContextSchema,
+  // NOTE: made it optional for now as we have some old requests without it
+  // 2024-07-07: remove .optional()
+  context: llmsStreamingContextSchema.optional(),
 });
 export type ChatStreamingInputSchema = z.infer<typeof chatStreamingInputSchema>;

@@ -12,6 +12,8 @@ const pricingSchema = z.object({
 const benchmarkSchema = z.object({
  cbaElo: z.number().optional(),
  cbaMmlu: z.number().optional(),
+  heCode: z.number().optional(), // HumanEval, code, 0-shot
+  vqaMmmu: z.number().optional(), // Visual Question Answering, MMMU, 0-shot
 });

 // const rateLimitsSchema = z.object({
@@ -46,6 +48,25 @@ export const llmsListModelsOutputSchema = z.object({
 });


+// Chat Generation Input (some parts of)
+
+const generateContextNameSchema = z.enum(['chat-ai-title', 'chat-ai-summarize', 'chat-followup-diagram', 'chat-react-turn', 'draw-expand-prompt']);
+export type GenerateContextNameSchema = z.infer<typeof generateContextNameSchema>;
+export const llmsGenerateContextSchema = z.object({
+  method: z.literal('chat-generate'),
+  name: generateContextNameSchema,
+  ref: z.string(),
+});
+
+const streamingContextNameSchema = z.enum(['conversation', 'ai-diagram', 'ai-flattener', 'call', 'beam-scatter', 'beam-gather', 'persona-extract']);
+export type StreamingContextNameSchema = z.infer<typeof streamingContextNameSchema>;
+export const llmsStreamingContextSchema = z.object({
+  method: z.literal('chat-stream'),
+  name: streamingContextNameSchema,
+  ref: z.string(),
+});
+
+
 // (non-streaming) Chat Generation Output

 export const llmsChatGenerateOutputSchema = z.object({
@@ -11,7 +11,7 @@ import { capitalizeFirstLetter } from '~/common/util/textUtils';
 import { fixupHost } from '~/common/util/urlUtils';

 import { OpenAIHistorySchema, openAIHistorySchema, OpenAIModelSchema, openAIModelSchema } from '../openai/openai.router';
-import { llmsChatGenerateOutputSchema, llmsListModelsOutputSchema, ModelDescriptionSchema } from '../llm.server.types';
+import { llmsChatGenerateOutputSchema, llmsGenerateContextSchema, llmsListModelsOutputSchema, ModelDescriptionSchema } from '../llm.server.types';

 import { OLLAMA_BASE_MODELS, OLLAMA_PREV_UPDATE } from './ollama.models';
 import { WireOllamaChatCompletionInput, wireOllamaChunkedOutputSchema, wireOllamaListModelsSchema, wireOllamaModelInfoSchema } from './ollama.wiretypes';
@@ -117,8 +117,11 @@ const adminPullModelSchema = z.object({

 const chatGenerateInputSchema = z.object({
  access: ollamaAccessSchema,
-  model: openAIModelSchema, history: openAIHistorySchema,
-  // functions: openAIFunctionsSchema.optional(), forceFunctionName: z.string().optional(),
+  model: openAIModelSchema,
+  history: openAIHistorySchema,
+  // functions: openAIFunctionsSchema.optional(),
+  // forceFunctionName: z.string().optional(),
+  context: llmsGenerateContextSchema.optional(),
 });

 const listPullableOutputSchema = z.object({
@@ -0,0 +1,84 @@
+// here for reference only - for future mapping of CBA scores to the model IDs
+// const modelIdToPrefixMap: { [key: string]: string } = {
+//   // Anthropic models
+//   'Claude 3.5 Sonnet': 'claude-3-5-sonnet-20240620',
+//   'Claude 3 Opus': 'claude-3-opus-20240229',
+//   'Claude 3 Sonnet': 'claude-3-sonnet-20240229',
+//   'Claude 3 Haiku': 'claude-3-haiku-20240307',
+//   'Claude-2.1': 'claude-2.1',
+//   'Claude-2.0': 'claude-2.0',
+//   'Claude-1': '', // No exact match
+//   'Claude-Instant-1': 'claude-instant-1.2', // Closest match
+//
+//   // Gemini models
+//   'Gemini-1.5-Pro-Exp-0801': 'models/gemini-1.5-pro-latest', // Closest match
+//   'Gemini Advanced App (2024-05-14)': '', // No exact match
+//   'Gemini-1.5-Pro-001': 'models/gemini-1.5-pro-001',
+//   'Gemini-1.5-Pro-Preview-0409': 'models/gemini-1.5-pro-latest', // Closest match
+//   'Gemini-1.5-Flash-001': 'models/gemini-1.5-flash-001',
+//   'Gemini App (2024-01-24)': '', // No exact match
+//   'Gemini-1.0-Pro-001': 'models/gemini-1.0-pro-001',
+//   'Gemini Pro': 'models/gemini-pro',
+//
+//   // OpenAI models (from the previous file)
+//   'GPT-4o-2024-05-13': 'gpt-4o-2024-05-13',
+//   'GPT-4o-mini-2024-07-18': 'gpt-4o-mini-2024-07-18',
+//   'GPT-4-Turbo-2024-04-09': 'gpt-4-turbo-2024-04-09',
+//   'GPT-4-1106-preview': 'gpt-4-1106-preview',
+//   'GPT-4-0125-preview': 'gpt-4-0125-preview',
+//   'GPT-4-0314': 'gpt-4-0314',
+//   'GPT-4-0613': 'gpt-4-0613',
+//   'GPT-3.5-Turbo-0613': 'gpt-3.5-turbo-0613',
+//   'GPT-3.5-Turbo-0314': 'gpt-3.5-turbo-0314',
+//   'GPT-3.5-Turbo-0125': 'gpt-3.5-turbo-0125',
+//
+//   // Mistral models (from the previous file)
+//   'Mistral-Large-2402': 'mistral-large-2402',
+//   'Mixtral-8x7b-Instruct-v0.1': 'mistralai/Mixtral-8x7B-Instruct-v0.1',
+//
+//   // Other models without matches
+//   'Gemini-1.5-Pro-Exp-0801': '',
+//   'Meta-Llama-3.1-405b-Instruct': '',
+//   'Gemini-1.5-Pro-001': '',
+//   'Meta-Llama-3.1-70b-Instruct': '',
+//   'Yi-Large-preview': '',
+//   'Deepseek-v2-API-0628': '',
+//   'Gemma-2-27b-it': '',
+//   'Yi-Large': '',
+//   'Nemotron-4-340B-Instruct': '',
+//   'GLM-4-0520': '',
+//   'Llama-3-70b-Instruct': '',
+//   'Reka-Core-20240501': '',
+//   'Command R+': '',
+//   'Gemma-2-9b-it': '',
+//   'Qwen2-72B-Instruct': '',
+//   'GLM-4-0116': '',
+//   'Qwen-Max-0428': '',
+//   'DeepSeek-Coder-V2-Instruct': '',
+//   'Reka-Flash-Preview-20240611': '',
+//   'Meta-Llama-3.1-8b-Instruct': '',
+//   'Qwen1.5-110B-Chat': '',
+//   'Yi-1.5-34B-Chat': '',
+//   'Reka-Flash-21B-online': '',
+//   'Llama-3-8b-Instruct': '',
+//   'Command R': '',
+//   'Reka-Flash-21B': '',
+//   'Qwen1.5-72B-Chat': '',
+//   'Mixtral-8x22b-Instruct-v0.1': '',
+//   'Zephyr-ORPO-141b-A35b-v0.1': '',
+//   'Qwen1.5-32B-Chat': '',
+//   'Mistral-Next': '',
+//   'Phi-3-Medium-4k-Instruct': '',
+//   'Starling-LM-7B-beta': '',
+//   'Yi-34B-Chat': '',
+//   'Qwen1.5-14B-Chat': '',
+//   'WizardLM-70B-v1.0': '',
+//   'Tulu-2-DPO-70B': '',
+//   'DBRX-Instruct-Preview': '',
+//   'Phi-3-Small-8k-Instruct': '',
+//   'Llama-2-70b-chat': '',
+//   'OpenChat-3.5-0106': '',
+//   'Vicuna-33B': '',
+//   'Snowflake Arctic Instruct': '',
+//   'Starling-LM-7B-alpha': '',
+// };
@@ -9,34 +9,139 @@ import { wireTogetherAIListOutputSchema } from './togetherai.wiretypes';


 // [Azure] / [OpenAI]
+// https://platform.openai.com/docs/models
 const _knownOpenAIChatModels: ManualMappings = [

-  // GPT-4o -> 2024-05-13
+  // GPT-4o -> 2024-05-13 (Starting October 2nd, 2024, gpt-4o will point to the gpt-4o-2024-08-06 snapshot)
  {
    idPrefix: 'gpt-4o',
    label: 'GPT-4o',
-    description: 'Currently points to gpt-4o-2024-05-13.',
-    symLink: 'gpt-4o-2024-05-13',
+    description: 'Points to gpt-4o-2024-08-06 starting on Oct 2, 2024.',
+    symLink: 'gpt-4o-2024-08-06',
    hidden: true,
    // copied from symlinked
    contextWindow: 128000,
-    maxCompletionTokens: 4096,
+    maxCompletionTokens: 16384,
    trainingDataCutoff: 'Oct 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
-    pricing: { chatIn: 5, chatOut: 15 },
-    benchmark: { cbaElo: 1310 },
+    pricing: { chatIn: 2.5, chatOut: 10 },
+    benchmark: { cbaElo: 1286 + 1 },
  },
  {
    isLatest: true,
+    idPrefix: 'gpt-4o-2024-08-06',
+    label: 'GPT-4o (2024-08-06)',
+    description: 'Latest snapshot that supports Structured Outputs',
+    contextWindow: 128000,
+    maxCompletionTokens: 16384,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json], // + Structured Outputs?
+    pricing: { chatIn: 2.5, chatOut: 10 },
+    benchmark: { cbaElo: 1286 + 1 },
+  },
+  {
    idPrefix: 'gpt-4o-2024-05-13',
    label: 'GPT-4o (2024-05-13)',
-    description: 'Advanced, multimodal flagship model that’s cheaper and faster than GPT-4 Turbo.',
+    description: 'Advanced, multimodal flagship model that\'s cheaper and faster than GPT-4 Turbo.',
    contextWindow: 128000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Oct 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 5, chatOut: 15 },
-    benchmark: { cbaElo: 1310 },
+    benchmark: { cbaElo: 1286 },
+    hidden: true,
+  },
+  {
+    idPrefix: 'chatgpt-4o-latest',
+    label: 'ChatGPT-4o Latest',
+    description: 'Intended for research and evaluation. Dynamic model continuously updated to the current version of GPT-4o in ChatGPT.',
+    contextWindow: 128000,
+    maxCompletionTokens: 16384,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
+    pricing: { chatIn: 5, chatOut: 15 },
+  },
+
+  // GPT-4o mini
+  {
+    idPrefix: 'gpt-4o-mini',
+    label: 'GPT-4o mini',
+    description: 'Currently points to gpt-4o-mini-2024-07-18.',
+    symLink: 'gpt-4o-mini-2024-07-18',
+    hidden: true,
+    // copied from symlinked
+    contextWindow: 128000,
+    maxCompletionTokens: 16384,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
+    pricing: { chatIn: 0.15, chatOut: 0.60 },
+    benchmark: { cbaElo: 1277, cbaMmlu: 82.0 },
+  },
+  {
+    idPrefix: 'gpt-4o-mini-2024-07-18',
+    label: 'GPT-4o Mini (2024-07-18)',
+    description: 'Affordable model for fast, lightweight tasks. GPT-4o mini is cheaper and more capable than GPT-3.5 Turbo.',
+    contextWindow: 128000,
+    maxCompletionTokens: 16384,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
+    pricing: { chatIn: 0.15, chatOut: 0.60 },
+  },
+
+  // o1-preview
+  {
+    idPrefix: 'o1-preview',
+    label: 'o1 Preview',
+    description: 'Supported in Big-AGI 2. Points to the most recent snapshot of the o1 model: o1-preview-2024-09-12',
+    symLink: 'o1-preview-2024-09-12',
+    hidden: true,
+    // copied from symlinked
+    contextWindow: 128000,
+    maxCompletionTokens: 32768,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 15, chatOut: 60 },
+    isPreview: true,
+  },
+  {
+    hidden: true, // we can't support it in Big-AGI 1
+    idPrefix: 'o1-preview-2024-09-12',
+    label: 'o1 Preview (2024-09-12)',
+    description: 'Supported in Big-AGI 2. New reasoning model for complex tasks that require broad general knowledge.',
+    contextWindow: 128000,
+    maxCompletionTokens: 32768,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 15, chatOut: 60 },
+    isPreview: true,
+  },
+
+  // o1-mini
+  {
+    idPrefix: 'o1-mini',
+    label: 'o1 Mini',
+    description: 'Supported in Big-AGI 2. Points to the most recent o1-mini snapshot: o1-mini-2024-09-12',
+    symLink: 'o1-mini-2024-09-12',
+    hidden: true,
+    // copied from symlinked
+    contextWindow: 128000,
+    maxCompletionTokens: 65536,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 3, chatOut: 12 },
+    isPreview: true,
+  },
+  {
+    hidden: true, // we can't support it in Big-AGI 1
+    idPrefix: 'o1-mini-2024-09-12',
+    label: 'o1 Mini (2024-09-12)',
+    description: 'Supported in Big-AGI 2. Fast, cost-efficient reasoning model tailored to coding, math, and science use cases.',
+    contextWindow: 128000,
+    maxCompletionTokens: 65536,
+    trainingDataCutoff: 'Oct 2023',
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
+    pricing: { chatIn: 3, chatOut: 12 },
+    isPreview: true,
  },

  // GPT4 Turbo with Vision -> 2024-04-09
@@ -52,7 +157,7 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Dec 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 10, chatOut: 30 },
-    benchmark: { cbaElo: 1261 },
+    benchmark: { cbaElo: 1257 },
  },
  {
    idPrefix: 'gpt-4-turbo-2024-04-09',
@@ -63,12 +168,12 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Dec 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 10, chatOut: 30 },
-    benchmark: { cbaElo: 1261 },
+    benchmark: { cbaElo: 1257 },
  },

  // GPT4 Turbo Previews
  {
-    idPrefix: 'gpt-4-turbo-preview', // GPT-4 Turbo preview model -> 0125
+    idPrefix: 'gpt-4-turbo-preview',
    label: 'GPT-4 Preview Turbo',
    description: 'GPT-4 Turbo preview model. Currently points to gpt-4-0125-preview.',
    symLink: 'gpt-4-0125-preview',
@@ -80,63 +185,33 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Dec 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 10, chatOut: 30 },
-    benchmark: { cbaElo: 1251 },
+    benchmark: { cbaElo: 1245 },
  },
  {
-    idPrefix: 'gpt-4-0125-preview', // GPT-4 Turbo preview model
+    idPrefix: 'gpt-4-0125-preview',
    label: 'GPT-4 Turbo (0125)',
-    description: 'GPT-4 Turbo preview model intended to reduce cases of "laziness" where the model doesn\'t complete a task. Returns a maximum of 4,096 output tokens.',
-    isPreview: true,
+    description: 'GPT-4 Turbo preview model intended to reduce cases of "laziness" where the model doesn\'t complete a task.',
    contextWindow: 128000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Dec 2023',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 10, chatOut: 30 },
-    benchmark: { cbaElo: 1251 },
+    benchmark: { cbaElo: 1245 },
    hidden: true,
  },
  {
    idPrefix: 'gpt-4-1106-preview', // GPT-4 Turbo preview model
    label: 'GPT-4 Turbo (1106)',
-    description: 'GPT-4 Turbo preview model featuring improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more. Returns a maximum of 4,096 output tokens.',
-    isPreview: true,
+    description: 'GPT-4 Turbo preview model featuring improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.',
    contextWindow: 128000,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Apr 2023',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Json],
    pricing: { chatIn: 10, chatOut: 30 },
-    benchmark: { cbaElo: 1255 },
+    benchmark: { cbaElo: 1251 },
    hidden: true,
  },

-  // GPT4 Vision Previews
-  {
-    idPrefix: 'gpt-4-vision-preview', // GPT-4 Turbo vision preview
-    label: 'GPT-4 Preview Vision',
-    description: 'GPT-4 model with the ability to understand images, in addition to all other GPT-4 Turbo capabilities. This is a preview model, we recommend developers to now use gpt-4-turbo which includes vision capabilities. Currently points to gpt-4-1106-vision-preview.',
-    symLink: 'gpt-4-1106-vision-preview',
-    // copied from symlinked
-    isPreview: true,
-    contextWindow: 128000,
-    maxCompletionTokens: 4096,
-    trainingDataCutoff: 'Apr 2023',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
-    pricing: { chatIn: 10, chatOut: 30 },
-    hidden: true, // Deprecated in favor of gpt-4-turbo
-  },
-  {
-    idPrefix: 'gpt-4-1106-vision-preview',
-    label: 'GPT-4 Preview Vision (1106)',
-    description: 'GPT-4 model with the ability to understand images, in addition to all other GPT-4 Turbo capabilities. This is a preview model, we recommend developers to now use gpt-4-turbo which includes vision capabilities. Returns a maximum of 4,096 output tokens.',
-    isPreview: true,
-    contextWindow: 128000,
-    maxCompletionTokens: 4096,
-    trainingDataCutoff: 'Apr 2023',
-    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision, LLM_IF_OAI_Fn],
-    pricing: { chatIn: 10, chatOut: 30 },
-    hidden: true, // Deprecated in favor of gpt-4-turbo
-  },
-

  // GPT4-32k's
  {
@@ -182,7 +257,7 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Sep 2021',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    pricing: { chatIn: 30, chatOut: 60 },
-    benchmark: { cbaElo: 1164 },
+    benchmark: { cbaElo: 1161 },
  },
  {
    idPrefix: 'gpt-4-0314',
@@ -192,7 +267,7 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Sep 2021',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    pricing: { chatIn: 30, chatOut: 60 },
-    benchmark: { cbaElo: 1189 },
+    benchmark: { cbaElo: 1186 },
    hidden: true,
  },
  {
@@ -206,39 +281,27 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Sep 2021',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    pricing: { chatIn: 30, chatOut: 60 },
-    benchmark: { cbaElo: 1164 },
+    benchmark: { cbaElo: 1161 },
+    isLegacy: true,
  },

-
-  // 3.5-Turbo-Instruct (Not for Chat)
-  {
-    idPrefix: 'gpt-3.5-turbo-instruct',
-    label: '3.5-Turbo Instruct',
-    description: 'Similar capabilities as GPT-3 era models. Compatible with legacy Completions endpoint and not Chat Completions.',
-    contextWindow: 4097,
-    trainingDataCutoff: 'Sep 2021',
-    interfaces: [/* NO: LLM_IF_OAI_Chat,*/ LLM_IF_OAI_Complete],
-    pricing: { chatIn: 1.5, chatOut: 2 },
-    hidden: true,
-  },
-
-
-  // 3.5-Turbo-16k's
+  // 3.5-Turbo
+  // As of July 2024, gpt-4o-mini should be used in place of gpt-3.5-turbo, as it is cheaper, more capable, multimodal, and just as fast.
  {
    idPrefix: 'gpt-3.5-turbo-0125',
    label: '3.5-Turbo (0125)',
-    description: 'The latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and a fix for a bug which caused a text encoding issue for non-English language function calls. Returns a maximum of 4,096 output tokens.',
+    description: 'The latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and a fix for a bug which caused a text encoding issue for non-English language function calls.',
    contextWindow: 16385,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Sep 2021',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    pricing: { chatIn: 0.5, chatOut: 1.5 },
-    benchmark: { cbaElo: 1104 },
+    benchmark: { cbaElo: 1105 },
  },
  {
    idPrefix: 'gpt-3.5-turbo-1106',
    label: '3.5-Turbo (1106)',
-    description: 'The latest GPT-3.5 Turbo model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.',
+    description: 'GPT-3.5 Turbo model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more.',
    contextWindow: 16385,
    maxCompletionTokens: 4096,
    trainingDataCutoff: 'Sep 2021',
@@ -250,7 +313,7 @@ const _knownOpenAIChatModels: ManualMappings = [
  {
    idPrefix: 'gpt-3.5-turbo',
    label: '3.5-Turbo',
-    description: 'Currently points to gpt-3.5-turbo-0125.',
+    description: 'Currently points to gpt-3.5-turbo-0125. As of July 2024, gpt-4o-mini should be used in place of gpt-3.5-turbo, as it is cheaper, more capable, multimodal, and just as fast.',
    symLink: 'gpt-3.5-turbo-0125',
    hidden: true,
    // copied
@@ -259,7 +322,19 @@ const _knownOpenAIChatModels: ManualMappings = [
    trainingDataCutoff: 'Sep 2021',
    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
    pricing: { chatIn: 0.5, chatOut: 1.5 },
-    benchmark: { cbaElo: 1104 },
+    benchmark: { cbaElo: 1105 },
+  },
+
+  // 3.5-Turbo-Instruct (Not for Chat)
+  {
+    idPrefix: 'gpt-3.5-turbo-instruct',
+    label: '3.5-Turbo Instruct',
+    description: 'Similar capabilities as GPT-3 era models. Compatible with legacy Completions endpoint and not Chat Completions.',
+    contextWindow: 4097,
+    trainingDataCutoff: 'Sep 2021',
+    interfaces: [/* NO: LLM_IF_OAI_Chat,*/ LLM_IF_OAI_Complete],
+    pricing: { chatIn: 1.5, chatOut: 2 },
+    hidden: true,
  },


@@ -667,10 +742,14 @@ export function openRouterModelFamilySortFn(a: { id: string }, b: { id: string }
  return aPrefixIndex !== -1 ? -1 : 1;
 }

-export function openRouterModelToModelDescription(wireModel: object): ModelDescriptionSchema {
+export function openRouterModelToModelDescription(wireModel: object): ModelDescriptionSchema | null {

  // parse the model
-  const model = wireOpenrouterModelsListOutputSchema.parse(wireModel);
+  const { data: model, error } = wireOpenrouterModelsListOutputSchema.safeParse(wireModel);
+  if (error) {
+    console.warn(`openrouterModelToModelDescription: Failed to parse model: ${error}`);
+    return null;
+  }

  // parse pricing
  const pricing: ModelDescriptionSchema['pricing'] = {
@@ -859,41 +938,84 @@ export function perplexityAIModelSort(a: ModelDescriptionSchema, b: ModelDescrip
 const _knownGroqModels: ManualMappings = [
  {
    isLatest: true,
+    idPrefix: 'llama-3.1-405b-reasoning',
+    label: 'Llama 3.1 · 405B',
+    description: 'LLaMA 3.1 405B developed by Meta with a context window of 131,072 tokens. Supports tool use.',
+    contextWindow: 131072,
+    maxCompletionTokens: 8000,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
+    isLatest: true,
+    idPrefix: 'llama-3.1-70b-versatile',
+    label: 'Llama 3.1 · 70B',
+    description: 'LLaMA 3.1 70B developed by Meta with a context window of 131,072 tokens. Supports tool use.',
+    contextWindow: 131072,
+    maxCompletionTokens: 8000,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
+    isLatest: true,
+    idPrefix: 'llama-3.1-8b-instant',
+    label: 'Llama 3.1 · 8B',
+    description: 'LLaMA 3.1 8B developed by Meta with a context window of 131,072 tokens. Supports tool use.',
+    contextWindow: 131072,
+    maxCompletionTokens: 8000,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
+    idPrefix: 'llama3-groq-70b-8192-tool-use-preview',
+    label: 'Llama 3 Groq · 70B Tool Use',
+    description: 'LLaMA 3 70B Tool Use developed by Groq with a context window of 8,192 tokens. Optimized for tool use.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
+    idPrefix: 'llama3-groq-8b-8192-tool-use-preview',
+    label: 'Llama 3 Groq · 8B Tool Use',
+    description: 'LLaMA 3 8B Tool Use developed by Groq with a context window of 8,192 tokens. Optimized for tool use.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
    idPrefix: 'llama3-70b-8192',
    label: 'Llama 3 · 70B',
-    description: 'LLaMA3 70b developed by Meta with a context window of 8,192 tokens.',
+    description: 'LLaMA3 70B developed by Meta with a context window of 8,192 tokens. Supports tool use.',
    contextWindow: 8192,
-    interfaces: [LLM_IF_OAI_Chat],
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+    // isLegacy: true,
+    hidden: true,
  },
  {
-    // isLatest: true,
    idPrefix: 'llama3-8b-8192',
    label: 'Llama 3 · 8B',
-    description: 'LLaMA3 8b developed by Meta with a context window of 8,192 tokens.',
+    description: 'LLaMA3 8B developed by Meta with a context window of 8,192 tokens. Supports tool use.',
    contextWindow: 8192,
-    interfaces: [LLM_IF_OAI_Chat],
-  },
-  {
-    idPrefix: 'llama2-70b-4096',
-    label: 'Llama 2 · 70B',
-    description: 'LLaMA2 70b developed by Meta with a context window of 4,096 tokens.',
-    contextWindow: 4096,
-    interfaces: [LLM_IF_OAI_Chat],
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+    // isLegacy: true,
    hidden: true,
  },
  {
    idPrefix: 'mixtral-8x7b-32768',
    label: 'Mixtral 8x7B',
-    description: 'Mixtral 8x7b developed by Mistral with a context window of 32,768 tokens.',
+    description: 'Mixtral 8x7B developed by Mistral with a context window of 32,768 tokens. Supports tool use.',
    contextWindow: 32768,
-    interfaces: [LLM_IF_OAI_Chat],
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
+  },
+  {
+    idPrefix: 'gemma2-9b-it',
+    label: 'Gemma 2 · 9B Instruct',
+    description: 'Gemma 2 9B developed by Google with a context window of 8,192 tokens. Supports tool use.',
+    contextWindow: 8192,
+    interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Fn],
  },
  {
    idPrefix: 'gemma-7b-it',
    label: 'Gemma 1.1 · 7B Instruct',
-    description: 'Gemma 7b developed by Google with a context window of 8,192 tokens.',
+    description: 'Gemma 7B developed by Google with a context window of 8,192 tokens. Supports tool use.',
    contextWindow: 8192,
    interfaces: [LLM_IF_OAI_Chat],
+    hidden: true,
  },
 ];

@@ -910,6 +1032,11 @@ export function groqModelToModelDescription(_model: unknown): ModelDescriptionSc
 }

 export function groqModelSortFn(a: ModelDescriptionSchema, b: ModelDescriptionSchema): number {
+  // sort hidden at the end
+  if (a.hidden && !b.hidden)
+    return 1;
+  if (!a.hidden && b.hidden)
+    return -1;
  // sort as per their order in the known models
  const aIndex = _knownGroqModels.findIndex(base => a.id.startsWith(base.idPrefix));
  const bIndex = _knownGroqModels.findIndex(base => b.id.startsWith(base.idPrefix));
@@ -12,7 +12,7 @@ import { fixupHost } from '~/common/util/urlUtils';

 import { OpenAIWire, WireOpenAICreateImageOutput, wireOpenAICreateImageOutputSchema, WireOpenAICreateImageRequest } from './openai.wiretypes';
 import { azureModelToModelDescription, groqModelSortFn, groqModelToModelDescription, lmStudioModelToModelDescription, localAIModelToModelDescription, mistralModelsSort, mistralModelToModelDescription, oobaboogaModelToModelDescription, openAIModelFilter, openAIModelToModelDescription, openRouterModelFamilySortFn, openRouterModelToModelDescription, perplexityAIModelDescriptions, perplexityAIModelSort, togetherAIModelsToModelDescriptions } from './models.data';
-import { llmsChatGenerateWithFunctionsOutputSchema, llmsListModelsOutputSchema, ModelDescriptionSchema } from '../llm.server.types';
+import { llmsChatGenerateWithFunctionsOutputSchema, llmsGenerateContextSchema, llmsListModelsOutputSchema, ModelDescriptionSchema } from '../llm.server.types';
 import { wilreLocalAIModelsApplyOutputSchema, wireLocalAIModelsAvailableOutputSchema, wireLocalAIModelsListOutputSchema } from './localai.wiretypes';


@@ -72,8 +72,11 @@ const listModelsInputSchema = z.object({

 const chatGenerateWithFunctionsInputSchema = z.object({
  access: openAIAccessSchema,
-  model: openAIModelSchema, history: openAIHistorySchema,
-  functions: openAIFunctionsSchema.optional(), forceFunctionName: z.string().optional(),
+  model: openAIModelSchema,
+  history: openAIHistorySchema,
+  functions: openAIFunctionsSchema.optional(),
+  forceFunctionName: z.string().optional(),
+  context: llmsGenerateContextSchema.optional(),
 });

 const createImagesInputSchema = z.object({
@@ -108,7 +111,7 @@ export const llmOpenAIRouter = createTRPCRouter({

      // [Azure]: use an older 'deployments' API to enumerate the models, and a modified OpenAI id to description mapping
      if (access.dialect === 'azure') {
-        const azureModels = await openaiGET(access, `/openai/deployments?api-version=2023-03-15-preview`);
+        const azureModels = await openaiGETOrThrow(access, `/openai/deployments?api-version=2023-03-15-preview`);

        const wireAzureListDeploymentsSchema = z.object({
          data: z.array(z.object({
@@ -146,7 +149,7 @@ export const llmOpenAIRouter = createTRPCRouter({


      // [non-Azure]: fetch openAI-style for all but Azure (will be then used in each dialect)
-      const openAIWireModelsResponse = await openaiGET<OpenAIWire.Models.Response>(access, '/v1/models');
+      const openAIWireModelsResponse = await openaiGETOrThrow<OpenAIWire.Models.Response>(access, '/v1/models');

      // [Together] missing the .data property
      if (access.dialect === 'togetherai')
@@ -253,7 +256,8 @@ export const llmOpenAIRouter = createTRPCRouter({
        case 'openrouter':
          models = openAIModels
            .sort(openRouterModelFamilySortFn)
-            .map(openRouterModelToModelDescription);
+            .map(openRouterModelToModelDescription)
+            .filter(desc => !!desc) as ModelDescriptionSchema[];
          break;

      }
@@ -267,17 +271,22 @@ export const llmOpenAIRouter = createTRPCRouter({
    .output(llmsChatGenerateWithFunctionsOutputSchema)
    .mutation(async ({ input }) => {

-      const { access, model, history, functions, forceFunctionName } = input;
+      const { access, model, history, functions, forceFunctionName, context } = input;
      const isFunctionsCall = !!functions && functions.length > 0;

      const completionsBody = openAIChatCompletionPayload(access.dialect, model, history, isFunctionsCall ? functions : null, forceFunctionName ?? null, 1, false);
-      const wireCompletions = await openaiPOST<OpenAIWire.ChatCompletion.Response, OpenAIWire.ChatCompletion.Request>(
+      const wireCompletions = await openaiPOSTOrThrow<OpenAIWire.ChatCompletion.Response, OpenAIWire.ChatCompletion.Request>(
        access, model.id, completionsBody, '/v1/chat/completions',
      );

      // expect a single output
-      if (wireCompletions?.choices?.length !== 1)
-        throw new TRPCError({ code: 'INTERNAL_SERVER_ERROR', message: `[OpenAI Issue] Expected 1 completion, got ${wireCompletions?.choices?.length}` });
+      if (wireCompletions?.choices?.length !== 1) {
+        console.error(`[POST] llmOpenAI.chatGenerateWithFunctions: ${access.dialect}: ${context?.name || 'no context'}: unexpected output${forceFunctionName ? ` (fn: ${forceFunctionName})` : ''}:`, model.id, wireCompletions?.choices);
+        throw new TRPCError({
+          code: 'UNPROCESSABLE_CONTENT',
+          message: `[OpenAI Issue] Expected 1 completion, got ${wireCompletions?.choices?.length}`,
+        });
+      }
      let { message, finish_reason } = wireCompletions.choices[0];

      // LocalAI hack/workaround, until https://github.com/go-skynet/LocalAI/issues/788 is fixed
@@ -318,7 +327,7 @@ export const llmOpenAIRouter = createTRPCRouter({
        delete requestBody.response_format;

      // create 1 image (dall-e-3 won't support more than 1, so better transfer the burden to the client)
-      const wireOpenAICreateImageOutput = await openaiPOST<WireOpenAICreateImageOutput, WireOpenAICreateImageRequest>(
+      const wireOpenAICreateImageOutput = await openaiPOSTOrThrow<WireOpenAICreateImageOutput, WireOpenAICreateImageRequest>(
        access, null, requestBody, '/v1/images/generations',
      );

@@ -340,7 +349,7 @@ export const llmOpenAIRouter = createTRPCRouter({
    .mutation(async ({ input: { access, text } }): Promise<OpenAIWire.Moderation.Response> => {
      try {

-        return await openaiPOST<OpenAIWire.Moderation.Response, OpenAIWire.Moderation.Request>(access, null, {
+        return await openaiPOSTOrThrow<OpenAIWire.Moderation.Response, OpenAIWire.Moderation.Request>(access, null, {
          input: text,
          model: 'text-moderation-latest',
        }, '/v1/moderations');
@@ -361,7 +370,7 @@ export const llmOpenAIRouter = createTRPCRouter({
  dialectLocalAI_galleryModelsAvailable: publicProcedure
    .input(listModelsInputSchema)
    .query(async ({ input: { access } }) => {
-      const wireLocalAIModelsAvailable = await openaiGET(access, '/models/available');
+      const wireLocalAIModelsAvailable = await openaiGETOrThrow(access, '/models/available');
      return wireLocalAIModelsAvailableOutputSchema.parse(wireLocalAIModelsAvailable);
    }),

@@ -374,7 +383,7 @@ export const llmOpenAIRouter = createTRPCRouter({
    }))
    .mutation(async ({ input: { access, galleryName, modelName } }) => {
      const galleryModelId = `${galleryName}@${modelName}`;
-      const wireLocalAIModelApply = await openaiPOST(access, null, { id: galleryModelId }, '/models/apply');
+      const wireLocalAIModelApply = await openaiPOSTOrThrow(access, null, { id: galleryModelId }, '/models/apply');
      return wilreLocalAIModelsApplyOutputSchema.parse(wireLocalAIModelApply);
    }),

@@ -385,7 +394,7 @@ export const llmOpenAIRouter = createTRPCRouter({
      jobId: z.string(),
    }))
    .query(async ({ input: { access, jobId } }) => {
-      const wireLocalAIModelsJobs = await openaiGET(access, `/models/jobs/${jobId}`);
+      const wireLocalAIModelsJobs = await openaiGETOrThrow(access, `/models/jobs/${jobId}`);
      return wireLocalAIModelsListOutputSchema.parse(wireLocalAIModelsJobs);
    }),

@@ -623,12 +632,12 @@ export function openAIChatCompletionPayload(dialect: OpenAIDialects, model: Open
  };
 }

-async function openaiGET<TOut extends object>(access: OpenAIAccessSchema, apiPath: string /*, signal?: AbortSignal*/): Promise<TOut> {
+async function openaiGETOrThrow<TOut extends object>(access: OpenAIAccessSchema, apiPath: string /*, signal?: AbortSignal*/): Promise<TOut> {
  const { headers, url } = openAIAccess(access, null, apiPath);
  return await fetchJsonOrTRPCError<TOut>(url, 'GET', headers, undefined, `OpenAI/${access.dialect}`);
 }

-async function openaiPOST<TOut extends object, TPostBody extends object>(access: OpenAIAccessSchema, modelRefId: string | null, body: TPostBody, apiPath: string /*, signal?: AbortSignal*/): Promise<TOut> {
+async function openaiPOSTOrThrow<TOut extends object, TPostBody extends object>(access: OpenAIAccessSchema, modelRefId: string | null, body: TPostBody, apiPath: string /*, signal?: AbortSignal*/): Promise<TOut> {
  const { headers, url } = openAIAccess(access, modelRefId, apiPath);
  return await fetchJsonOrTRPCError<TOut, TPostBody>(url, 'POST', headers, body, `OpenAI/${access.dialect}`);
 }
@@ -5,16 +5,20 @@ export const wireOpenrouterModelsListOutputSchema = z.object({
  id: z.string(),
  name: z.string(),
  description: z.string(),
+  // NOTE: for 'openrouter/auto', this is:  {
+  //   "prompt": "-1",
+  //   "completion": "-1"
+  // }
  pricing: z.object({
    prompt: z.string(),
    completion: z.string(),
-    image: z.string(),
-    request: z.string(),
+    image: z.string().optional(),
+    request: z.string().optional(),
  }),
  context_length: z.number(),
  architecture: z.object({
-    modality: z.string(), // z.enum(['text', 'multimodal']),
-    tokenizer: z.string(), // e.g. 'Mistral'
+    modality: z.string(), // z.enum(['text', 'multimodal', 'text+image->text]),
+    tokenizer: z.string(), // e.g. 'Mistral', 'Claude'
    instruct_type: z.string().nullable(),
  }),
  top_provider: z.object({
@@ -8,7 +8,7 @@ import type { DLLM, DLLMId, DModelSourceId } from '../store-llms';
 import type { ModelDescriptionSchema } from '../server/llm.server.types';
 import type { ModelVendorId } from './vendors.registry';
 import type { StreamingClientUpdate } from './unifiedStreamingClient';
-import type { VChatContextName, VChatContextRef, VChatFunctionIn, VChatMessageIn, VChatMessageOrFunctionCallOut, VChatMessageOut } from '../llm.client';
+import type { VChatContextRef, VChatFunctionIn, VChatGenerateContextName, VChatMessageIn, VChatMessageOrFunctionCallOut, VChatMessageOut, VChatStreamContextName } from '../llm.client';


 export interface IModelVendor<TSourceSetup = unknown, TAccess = unknown, TLLMOptions = unknown, TDLLM = DLLM<TSourceSetup, TLLMOptions>> {
@@ -44,6 +44,7 @@ export interface IModelVendor<TSourceSetup = unknown, TAccess = unknown, TLLMOpt
    access: TAccess,
    llmOptions: TLLMOptions,
    messages: VChatMessageIn[],
+    contextName: VChatGenerateContextName, contextRef: VChatContextRef | null,
    functions: VChatFunctionIn[] | null, forceFunctionName: string | null,
    maxTokens?: number,
  ) => Promise<VChatMessageOut | VChatMessageOrFunctionCallOut>;
@@ -53,7 +54,7 @@ export interface IModelVendor<TSourceSetup = unknown, TAccess = unknown, TLLMOpt
    llmId: DLLMId,
    llmOptions: TLLMOptions,
    messages: VChatMessageIn[],
-    contextName: VChatContextName, contexRef: VChatContextRef,
+    contextName: VChatStreamContextName, contextRef: VChatContextRef,
    functions: VChatFunctionIn[] | null, forceFunctionName: string | null,
    abortSignal: AbortSignal,
    onUpdate: (update: StreamingClientUpdate, done: boolean) => void,
@@ -3,7 +3,7 @@ import { apiAsync } from '~/common/util/trpc.client';

 import type { AnthropicAccessSchema } from '../../server/anthropic/anthropic.router';
 import type { IModelVendor } from '../IModelVendor';
-import type { VChatMessageOut } from '../../llm.client';
+import type { VChatContextRef, VChatGenerateContextName, VChatMessageOut } from '../../llm.client';
 import { unifiedStreamingClient } from '../unifiedStreamingClient';

 import { FALLBACK_LLM_RESPONSE_TOKENS, FALLBACK_LLM_TEMPERATURE, LLMOptionsOpenAI } from '../openai/openai.vendor';
@@ -47,7 +47,7 @@ export const ModelVendorAnthropic: IModelVendor<SourceSetupAnthropic, AnthropicA
  rpcUpdateModelsOrThrow: async (access) => await apiAsync.llmAnthropic.listModels.query({ access }),

  // Chat Generate (non-streaming) with Functions
-  rpcChatGenerateOrThrow: async (access, llmOptions, messages, functions, forceFunctionName, maxTokens) => {
+  rpcChatGenerateOrThrow: async (access, llmOptions, messages, contextName: VChatGenerateContextName, contextRef: VChatContextRef | null, functions, forceFunctionName, maxTokens) => {
    if (functions?.length || forceFunctionName)
      throw new Error('Anthropic does not support functions');

@@ -61,6 +61,11 @@ export const ModelVendorAnthropic: IModelVendor<SourceSetupAnthropic, AnthropicA
          maxTokens: maxTokens || llmResponseTokens || FALLBACK_LLM_RESPONSE_TOKENS,
        },
        history: messages,
+        context: contextRef ? {
+          method: 'chat-generate',
+          name: contextName,
+          ref: contextRef,
+        } : undefined,
      }) as VChatMessageOut;
    } catch (error: any) {
      const errorMessage = error?.message || error?.toString() || 'Anthropic Chat Generate Error';
@@ -1,10 +1,10 @@
 import { GeminiIcon } from '~/common/components/icons/vendors/GeminiIcon';
- import { apiAsync } from '~/common/util/trpc.client';
+import { apiAsync } from '~/common/util/trpc.client';

 import type { GeminiAccessSchema } from '../../server/gemini/gemini.router';
 import type { GeminiBlockSafetyLevel } from '../../server/gemini/gemini.wiretypes';
 import type { IModelVendor } from '../IModelVendor';
-import type { VChatMessageOut } from '../../llm.client';
+import type { VChatContextRef, VChatGenerateContextName, VChatMessageOut } from '../../llm.client';
 import { unifiedStreamingClient } from '../unifiedStreamingClient';

 import { FALLBACK_LLM_RESPONSE_TOKENS, FALLBACK_LLM_TEMPERATURE } from '../openai/openai.vendor';
@@ -60,7 +60,7 @@ export const ModelVendorGemini: IModelVendor<SourceSetupGemini, GeminiAccessSche
  rpcUpdateModelsOrThrow: async (access) => await apiAsync.llmGemini.listModels.query({ access }),

  // Chat Generate (non-streaming) with Functions
-  rpcChatGenerateOrThrow: async (access, llmOptions, messages, functions, forceFunctionName, maxTokens) => {
+  rpcChatGenerateOrThrow: async (access, llmOptions, messages, contextName: VChatGenerateContextName, contextRef: VChatContextRef | null, functions, forceFunctionName, maxTokens) => {
    if (functions?.length || forceFunctionName)
      throw new Error('Gemini does not support functions');

@@ -74,6 +74,11 @@ export const ModelVendorGemini: IModelVendor<SourceSetupGemini, GeminiAccessSche
          maxTokens: maxTokens || maxOutputTokens || FALLBACK_LLM_RESPONSE_TOKENS,
        },
        history: messages,
+        context: contextRef ? {
+          method: 'chat-generate',
+          name: contextName,
+          ref: contextRef,
+        } : undefined,
      }) as VChatMessageOut;
    } catch (error: any) {
      const errorMessage = error?.message || error?.toString() || 'Gemini Chat Generate Error';
@@ -3,7 +3,7 @@ import { apiAsync } from '~/common/util/trpc.client';

 import type { IModelVendor } from '../IModelVendor';
 import type { OllamaAccessSchema } from '../../server/ollama/ollama.router';
-import type { VChatMessageOut } from '../../llm.client';
+import type { VChatContextRef, VChatGenerateContextName, VChatMessageOut } from '../../llm.client';
 import { unifiedStreamingClient } from '../unifiedStreamingClient';

 import { FALLBACK_LLM_RESPONSE_TOKENS, FALLBACK_LLM_TEMPERATURE, LLMOptionsOpenAI } from '../openai/openai.vendor';
@@ -42,7 +42,7 @@ export const ModelVendorOllama: IModelVendor<SourceSetupOllama, OllamaAccessSche
  rpcUpdateModelsOrThrow: async (access) => await apiAsync.llmOllama.listModels.query({ access }),

  // Chat Generate (non-streaming) with Functions
-  rpcChatGenerateOrThrow: async (access, llmOptions, messages, functions, forceFunctionName, maxTokens) => {
+  rpcChatGenerateOrThrow: async (access, llmOptions, messages, contextName: VChatGenerateContextName, contextRef: VChatContextRef | null, functions, forceFunctionName, maxTokens) => {
    if (functions?.length || forceFunctionName)
      throw new Error('Ollama does not support functions');

@@ -56,6 +56,11 @@ export const ModelVendorOllama: IModelVendor<SourceSetupOllama, OllamaAccessSche
          maxTokens: maxTokens || llmResponseTokens || FALLBACK_LLM_RESPONSE_TOKENS,
        },
        history: messages,
+        context: contextRef ? {
+          method: 'chat-generate',
+          name: contextName,
+          ref: contextRef,
+        } : undefined,
      }) as VChatMessageOut;
    } catch (error: any) {
      const errorMessage = error?.message || error?.toString() || 'Ollama Chat Generate Error';
@@ -3,7 +3,7 @@ import { apiAsync } from '~/common/util/trpc.client';

 import type { IModelVendor } from '../IModelVendor';
 import type { OpenAIAccessSchema } from '../../server/openai/openai.router';
-import type { VChatMessageOrFunctionCallOut } from '../../llm.client';
+import type { VChatContextRef, VChatGenerateContextName, VChatMessageOrFunctionCallOut } from '../../llm.client';
 import { unifiedStreamingClient } from '../unifiedStreamingClient';

 import { OpenAILLMOptions } from './OpenAILLMOptions';
@@ -60,7 +60,7 @@ export const ModelVendorOpenAI: IModelVendor<SourceSetupOpenAI, OpenAIAccessSche
  rpcUpdateModelsOrThrow: async (access) => await apiAsync.llmOpenAI.listModels.query({ access }),

  // Chat Generate (non-streaming) with Functions
-  rpcChatGenerateOrThrow: async (access, llmOptions, messages, functions, forceFunctionName, maxTokens) => {
+  rpcChatGenerateOrThrow: async (access, llmOptions, messages, contextName: VChatGenerateContextName, contextRef: VChatContextRef | null, functions, forceFunctionName, maxTokens) => {
    const { llmRef, llmTemperature, llmResponseTokens } = llmOptions;
    try {
      return await apiAsync.llmOpenAI.chatGenerateWithFunctions.mutate({
@@ -73,6 +73,11 @@ export const ModelVendorOpenAI: IModelVendor<SourceSetupOpenAI, OpenAIAccessSche
        functions: functions ?? undefined,
        forceFunctionName: forceFunctionName ?? undefined,
        history: messages,
+        context: contextRef ? {
+          method: 'chat-generate',
+          name: contextName,
+          ref: contextRef,
+        } : undefined,
      }) as VChatMessageOrFunctionCallOut;
    } catch (error: any) {
      const errorMessage = error?.message || error?.toString() || 'OpenAI Chat Generate Error';
@@ -3,7 +3,7 @@ import { frontendSideFetch } from '~/common/util/clientFetchers';

 import type { ChatStreamingInputSchema, ChatStreamingPreambleModelSchema, ChatStreamingPreambleStartSchema } from '../server/llm.server.streaming';
 import type { DLLMId } from '../store-llms';
-import type { VChatContextName, VChatContextRef, VChatFunctionIn, VChatMessageIn } from '../llm.client';
+import type { VChatContextRef, VChatFunctionIn, VChatMessageIn, VChatStreamContextName } from '../llm.client';

 import type { OpenAIAccessSchema } from '../server/openai/openai.router';
 import type { OpenAIWire } from '../server/openai/openai.wiretypes';
@@ -29,7 +29,7 @@ export async function unifiedStreamingClient<TSourceSetup = unknown, TLLMOptions
  llmId: DLLMId,
  llmOptions: TLLMOptions,
  messages: VChatMessageIn[],
-  contextName: VChatContextName, contextRef: VChatContextRef,
+  contextName: VChatStreamContextName, contextRef: VChatContextRef,
  functions: VChatFunctionIn[] | null, forceFunctionName: string | null,
  abortSignal: AbortSignal,
  onUpdate: (update: StreamingClientUpdate, done: boolean) => void,
@@ -1,7 +1,7 @@
 {
  "functions": {
-    "app/api/trpc-node/**/*": {
-      "maxDuration": 25
+    "api/trpc-node/**/*": {
+      "maxDuration": 30
    }
  }
 }
Author	SHA1	Message	Date
Enrico Ros	14724a864c	OpenRouter: compensate for older typescript	2025-04-05 10:30:15 -07:00
Enrico Ros	5e2b196c4d	OpenRouter: models list: prevent schema changes from breaking working models. Fixes #787	2025-04-05 10:26:44 -07:00
Enrico Ros	e7686f60b1	OpenRouter: models list: ignore missing fields on 'openrouter/auto'. Fixes #787	2025-04-05 10:25:35 -07:00
Enrico Ros	380f666d35	Roll Gemini descriptions. Fixes #783	2025-03-29 12:35:29 -07:00
Enrico Ros	3e277b1a35	Optional desc. #783	2025-03-29 12:35:05 -07:00
Enrico Ros	9bac46ea75	1.16.9 Release	2025-01-21 18:09:37 -08:00
Enrico Ros	2af4ee7dbe	Remove v1-dev, fully absorbed into v2-dev.	2025-01-21 18:05:18 -08:00
Enrico Ros	590fc0d021	Gemini: relax parser - Fixes #700	2024-12-19 01:09:40 -08:00
Enrico Ros	746b0dad40	Update Node to 22	2024-12-19 01:08:41 -08:00
Enrico Ros	b327da3ded	Fix #675 (pre-v2)	2024-11-06 16:37:18 -08:00
Enrico Ros	7a818bdcd0	Update branch names	2024-10-28 20:09:53 -07:00
Enrico Ros	c92ee2e22a	v1: document branch names	2024-10-28 20:02:18 -07:00
Enrico Ros	632a4a565f	[stable] OpenAI: update models	2024-10-25 10:13:13 -07:00
Enrico Ros	d712c275a0	[stable] Anthropic: update models	2024-10-25 10:06:42 -07:00
Enrico Ros	1adff7481b	Dev survey for Big-AGI 2.	2024-10-11 21:55:46 -07:00
Enrico Ros	393e19dda9	Vercel: fix timeout	2024-10-03 12:37:21 -07:00
Enrico Ros	39c5c7c9ba	Call out to Big-AGI 2	2024-09-13 14:06:12 -07:00
Enrico Ros	e64a5e59ef	1.16.8 Release	2024-09-13 13:50:00 -07:00
Enrico Ros	574c2cf0e3	Call out to Big-AGI 2	2024-09-13 13:49:11 -07:00
Enrico Ros	1d3321b336	OpenAI: o1 support label	2024-09-13 11:02:33 -07:00
Enrico Ros	de25e5822d	OpenAI: o1 relabel	2024-09-13 10:59:40 -07:00
Enrico Ros	6a904c9f37	OpenAI: 3.5 non legacy	2024-09-13 10:59:32 -07:00
Enrico Ros	30c3283572	OpenAI: add o1	2024-09-13 10:53:42 -07:00
Enrico Ros	10bba19079	OpenAI: add ChatGPT-4o-latest	2024-09-13 10:53:32 -07:00
Enrico Ros	713079f2f2	OpenAI: bits	2024-09-13 10:53:20 -07:00
Enrico Ros	6e16e989ac	OpenAI: move 4o-mini	2024-09-13 10:53:09 -07:00
Enrico Ros	4e89e0b1e4	OpenAI: clean IDs	2024-09-13 10:52:19 -07:00
Enrico Ros	6067c289ab	OpenAI: remove vision previews	2024-09-13 10:52:00 -07:00
Enrico Ros	32ebfea9cb	OpenAI: reorder	2024-09-13 10:20:52 -07:00
Enrico Ros	dec280d54d	1.16.7 Release (cherry picked from commit `22b32d571d`)	2024-08-07 02:51:59 -07:00
Enrico Ros	4823e97783	Mapping doc, for the future. (cherry picked from commit `a416cafc4e`)	2024-08-07 02:51:59 -07:00
Enrico Ros	6a5685995f	OpenAI: update models (cherry picked from commit `5f5efe6133`)	2024-08-07 02:51:59 -07:00
Enrico Ros	3b4d5691d7	1.16.6: Release. Fixes #604	2024-07-24 21:31:57 -07:00
Enrico Ros	45c09d021a	Groq: update output tokens (max 8,000 for 3.1)	2024-07-24 21:27:20 -07:00
Enrico Ros	8ef759fe0f	Groq: update Models	2024-07-24 21:27:12 -07:00
Enrico Ros	c06735fdd2	1.16.5: Release	2024-07-18 16:15:53 -07:00
Enrico Ros	cf4297a1af	OpenAI: support 4o Mini (16384 token output)	2024-07-18 16:15:37 -07:00
Enrico Ros	5d458d68bd	Warn devs.	2024-07-18 16:12:17 -07:00
Enrico Ros	c3db077ae8	1.16.4: release	2024-07-15 14:13:36 -07:00
Enrico Ros	779b265b20	Anthropic: 8192 tokens	2024-07-15 14:08:02 -07:00
Enrico Ros	7d6d7e619b	Anthropic: hardcode date	2024-06-20 12:42:10 -07:00
Enrico Ros	34caa16e39	1.16.3: release	2024-06-20 12:27:42 -07:00
Enrico Ros	976426dbd3	Anthropic: support Claude 3.5 Sonnet	2024-06-20 12:27:26 -07:00
Enrico Ros	b4d8e39d56	Gemini: acknowledge the new capability to `createCachedContent`. Fixes #565	2024-06-10 23:56:02 -07:00
Enrico Ros	11c41e7381	Function call: increase debug verbosity	2024-06-07 14:18:01 -07:00
Enrico Ros	358d8a54ff	Increase llms alignment before function calling.	2024-06-07 14:11:36 -07:00
Enrico Ros	3c8fedce68	Highlight issues with chatGenerateWithFunctions	2024-06-07 12:38:21 -07:00