OpenRouter: compensate for older typescript

(cherry picked from commit 22b32d571d)

.github/workflows/docker-image.yml

@@ -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

README.md

@@ -11,18 +11,35 @@ 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](https://github.com/users/enricoros/projects/4/views/2)** 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 1.16.1...1.16.3 · Jun 20, 2024 (patch releases)
+Note: After the V2 release in Q4, `v2/dev` will become the default branch and `v1/dev` will reach EOL.
 
+### 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...1.16.9 · Jan 21, 2025 (patch releases)
+
+- 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

docs/changelog.md

@@ -10,8 +10,14 @@ 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 1.16.1...1.16.3 · Jun 20, 2024 (patch releases)
+### What's New in 1.16.1...1.16.9 · Jan 21, 2025 (patch releases)
 
+- 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

package-lock.json

@@ -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": {

package.json

@@ -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"
   }
 }

src/apps/chat/components/message/ChatMessage.tsx

@@ -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;

src/apps/news/AppNews.tsx

@@ -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 }}>

src/apps/news/beam.data.tsx

@@ -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';
 

src/apps/news/bigAgi2.data.tsx

@@ -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>;

src/apps/news/news.data.tsx

@@ -61,7 +61,7 @@ export const NewsItems: NewsItem[] = [
     ]
   }*/
   {
-    versionCode: '1.16.3',
+    versionCode: '1.16.9',
     versionName: 'Crystal Clear',
     versionDate: new Date('2024-06-07T05:00:00Z'),
     // versionDate: new Date('2024-05-13T19:00:00Z'),
@@ -80,6 +80,13 @@ export const NewsItems: NewsItem[] = [
       { 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></> },
     ],
   },
   {

src/apps/news/news.version.tsx

@@ -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 16.3
+export const incrementalNewsVersion: number = 16.1; // not notifying for 1.16.9
 
 
 interface NewsState {

src/common/util/modelUtils.ts

@@ -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';

src/modules/llms/server/anthropic/anthropic.models.ts

@@ -22,22 +22,29 @@ export const hardcodedAnthropicModels: (ModelDescriptionSchema & { isLegacy?: bo
   //   },
   // },
   {
-    id: 'claude-3-5-sonnet-20240620',
+    id: 'claude-3-5-sonnet-20241022',
     label: 'Claude 3.5 Sonnet',
-    created: roundTime('2024-06-20 06:00'),
-    description: 'The most intelligent Claude model',
+    created: roundTime('2024-10-22 06:00'),
+    description: 'Most intelligent Claude model to date',
     contextWindow: 200000, // Characters
-    maxCompletionTokens: 4096,
+    maxCompletionTokens: 8192,
     trainingDataCutoff: 'Apr 2024',
     interfaces: [LLM_IF_OAI_Chat, LLM_IF_OAI_Vision],
     pricing: { chatIn: 3, chatOut: 15 },
-    benchmark: {
-      heCode: 92.0,
-      vqaMmmu: 68.3,
-      // TODO: Update with official benchmarks when available
-      cbaElo: 1256 - 1, // Placeholder
-      cbaMmlu: 86.8 - 1, // Placeholder
-    },
+    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', // ...
@@ -81,6 +88,7 @@ export const hardcodedAnthropicModels: (ModelDescriptionSchema & { isLegacy?: bo
     pricing: { chatIn: 3, chatOut: 15 },
     benchmark: { cbaElo: 1203, cbaMmlu: 79 },
     hidden: true,
+    isLegacy: true,
   },
   {
     id: 'claude-3-haiku-20240307',
@@ -131,24 +139,4 @@ export const hardcodedAnthropicModels: (ModelDescriptionSchema & { isLegacy?: bo
     pricing: { chatIn: 0.8, chatOut: 2.4 },
     hidden: true,
   },
-  // {
-  //   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,
-  // },
 ];

src/modules/llms/server/anthropic/anthropic.router.ts

@@ -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';

src/modules/llms/server/gemini/gemini.models.ts

@@ -1,6 +1,6 @@
 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
@@ -12,7 +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
@@ -20,175 +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);
+  // 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));
@@ -200,14 +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 {
@@ -218,11 +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,
   };
 }

src/modules/llms/server/gemini/gemini.router.ts

@@ -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, llmsGenerateContextSchema, llmsListModelsOutputSchema } from '../llm.server.types';
+import { llmsChatGenerateOutputSchema, llmsGenerateContextSchema, llmsListModelsOutputSchema, type ModelDescriptionSchema } from '../llm.server.types';
 
 import { OpenAIHistorySchema, openAIHistorySchema, OpenAIModelSchema, openAIModelSchema } from '../openai/openai.router';
 
@@ -148,9 +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 {

src/modules/llms/server/gemini/gemini.wiretypes.ts

@@ -9,28 +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([
-    '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',
-  ])),
+  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(),
@@ -172,7 +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'

src/modules/llms/server/openai/models.cba.ts

@@ -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': '',
+// };

src/modules/llms/server/openai/models.data.ts

@@ -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));

src/modules/llms/server/openai/openai.router.ts

@@ -256,7 +256,8 @@ export const llmOpenAIRouter = createTRPCRouter({
         case 'openrouter':
           models = openAIModels
             .sort(openRouterModelFamilySortFn)
-            .map(openRouterModelToModelDescription);
+            .map(openRouterModelToModelDescription)
+            .filter(desc => !!desc) as ModelDescriptionSchema[];
           break;
 
       }

src/modules/llms/server/openai/openrouter.wiretypes.ts

@@ -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({

vercel_PRODUCTION.json

@@ -1,7 +1,7 @@
 {
   "functions": {
-    "app/api/trpc-node/**/*": {
-      "maxDuration": 25
+    "api/trpc-node/**/*": {
+      "maxDuration": 30
     }
   }
 }