feat(providers): add models.json for models list

data/.gitignore

@@ -1,2 +1,3 @@
 *
+!models.json
 !.gitignore

data/models.json

@@ -0,0 +1,157 @@
+{
+  "_comment": "Ollama models are fetched from the Ollama API, so they are not included here.",
+  "chatModels": {
+    "openai": [
+      {
+        "displayName": "GPT-3.5 Turbo",
+        "key": "gpt-3.5-turbo"
+      },
+      {
+        "displayName": "GPT-4",
+        "key": "gpt-4"
+      },
+      {
+        "displayName": "GPT-4 Turbo",
+        "key": "gpt-4-turbo"
+      },
+      {
+        "displayName": "GPT-4 Omni",
+        "key": "gpt-4o"
+      },
+      {
+        "displayName": "GPT-4 Omni Mini",
+        "key": "gpt-4o-mini"
+      }
+    ],
+    "groq": [
+      {
+        "displayName": "Gemma2 9B IT",
+        "key": "gemma2-9b-it"
+      },
+      {
+        "displayName": "Llama 3.3 70B Versatile",
+        "key": "llama-3.3-70b-versatile"
+      },
+      {
+        "displayName": "Llama 3.1 8B Instant",
+        "key": "llama-3.1-8b-instant"
+      },
+      {
+        "displayName": "Llama3 70B 8192",
+        "key": "llama3-70b-8192"
+      },
+      {
+        "displayName": "Llama3 8B 8192",
+        "key": "llama3-8b-8192"
+      },
+      {
+        "displayName": "Mixtral 8x7B 32768",
+        "key": "mixtral-8x7b-32768"
+      },
+      {
+        "displayName": "Qwen QWQ 32B (Preview)",
+        "key": "qwen-qwq-32b"
+      },
+      {
+        "displayName": "Mistral Saba 24B (Preview)",
+        "key": "mistral-saba-24b"
+      },
+      {
+        "displayName": "DeepSeek R1 Distill Llama 70B (Preview)",
+        "key": "deepseek-r1-distill-llama-70b"
+      }
+    ],
+    "gemini": [
+      {
+        "displayName": "Gemini 2.5 Pro Experimental",
+        "key": "gemini-2.5-pro-exp-03-25"
+      },
+      {
+        "displayName": "Gemini 2.0 Flash",
+        "key": "gemini-2.0-flash"
+      },
+      {
+        "displayName": "Gemini 2.0 Flash-Lite",
+        "key": "gemini-2.0-flash-lite"
+      },
+      {
+        "displayName": "Gemini 2.0 Flash Thinking Experimental",
+        "key": "gemini-2.0-flash-thinking-exp-01-21"
+      },
+      {
+        "displayName": "Gemini 1.5 Flash",
+        "key": "gemini-1.5-flash"
+      },
+      {
+        "displayName": "Gemini 1.5 Flash-8B",
+        "key": "gemini-1.5-flash-8b"
+      },
+      {
+        "displayName": "Gemini 1.5 Pro",
+        "key": "gemini-1.5-pro"
+      }
+    ],
+    "anthropic": [
+      {
+        "displayName": "Claude 3.7 Sonnet",
+        "key": "claude-3-7-sonnet-20250219"
+      },
+      {
+        "displayName": "Claude 3.5 Haiku",
+        "key": "claude-3-5-haiku-20241022"
+      },
+      {
+        "displayName": "Claude 3.5 Sonnet v2",
+        "key": "claude-3-5-sonnet-20241022"
+      },
+      {
+        "displayName": "Claude 3.5 Sonnet",
+        "key": "claude-3-5-sonnet-20240620"
+      },
+      {
+        "displayName": "Claude 3 Opus",
+        "key": "claude-3-opus-20240229"
+      },
+      {
+        "displayName": "Claude 3 Sonnet",
+        "key": "claude-3-sonnet-20240229"
+      },
+      {
+        "displayName": "Claude 3 Haiku",
+        "key": "claude-3-haiku-20240307"
+      }
+    ]
+  },
+  "embeddingModels": {
+    "openai": [
+      {
+        "displayName": "Text Embedding 3 Large",
+        "key": "text-embedding-3-large"
+      },
+      {
+        "displayName": "Text Embedding 3 Small",
+        "key": "text-embedding-3-small"
+      }
+    ],
+    "gemini": [
+      {
+        "displayName": "Gemini Embedding",
+        "key": "gemini-embedding-exp"
+      }
+    ],
+    "transformers": [
+      {
+        "displayName": "BGE Small",
+        "key": "xenova-bge-small-en-v1.5"
+      },
+      {
+        "displayName": "GTE Small",
+        "key": "xenova-gte-small"
+      },
+      {
+        "displayName": "Bert Multilingual",
+        "key": "xenova-bert-base-multilingual-uncased"
+      }
+    ]
+  }
+}

docs/API/SEARCH.md

@@ -33,7 +33,6 @@ The API accepts a JSON object in the request body, where you define the focus mo
     ["human", "Hi, how are you?"],
     ["assistant", "I am doing well, how can I help you today?"]
   ],
-  "systemInstructions": "Focus on providing technical details about Perplexica's architecture.",
   "stream": false
 }
 ```
@@ -64,8 +63,6 @@ The API accepts a JSON object in the request body, where you define the focus mo
 
 - **`query`** (string, required): The search query or question.
 
-- **`systemInstructions`** (string, optional): Custom instructions provided by the user to guide the AI's response. These instructions are treated as user preferences and have lower priority than the system's core instructions. For example, you can specify a particular writing style, format, or focus area.
-
 - **`history`** (array, optional): An array of message pairs representing the conversation history. Each pair consists of a role (either 'human' or 'assistant') and the message content. This allows the system to use the context of the conversation to refine results. Example:
 
   ```json

package.json

@@ -1,6 +1,6 @@
 {
   "name": "perplexica-frontend",
-  "version": "1.10.2",
+  "version": "1.10.1",
   "license": "MIT",
   "author": "ItzCrazyKns",
   "scripts": {

sample.config.toml

@@ -22,8 +22,5 @@ MODEL_NAME = ""
 [MODELS.OLLAMA]
 API_URL = "" # Ollama API URL - http://host.docker.internal:11434
 
-[MODELS.DEEPSEEK]
-API_KEY = ""
-
 [API_ENDPOINTS]
 SEARXNG = "" # SearxNG API URL - http://localhost:32768

src/app/api/chat/route.ts

@@ -29,7 +29,6 @@ type Message = {
   messageId: string;
   chatId: string;
   content: string;
-  userSessionId: string;
 };
 
 type ChatModel = {
@@ -139,7 +138,6 @@ const handleHistorySave = async (
     where: eq(chats.id, message.chatId),
   });
 
-  let currentDate = new Date();
   if (!chat) {
     await db
       .insert(chats)
@@ -149,8 +147,6 @@ const handleHistorySave = async (
         createdAt: new Date().toString(),
         focusMode: focusMode,
         files: files.map(getFileDetails),
-        userSessionId: message.userSessionId,
-        timestamp: currentDate.toISOString(),
       })
       .execute();
   }

src/app/api/chats/route.ts

@@ -1,47 +1,10 @@
 import db from '@/lib/db';
-import { chats } from '@/lib/db/schema';
-import { eq, sql} from 'drizzle-orm';
 
 export const GET = async (req: Request) => {
   try {
-    // get header from request
-    const headers = await req.headers;
-    const userSessionId = headers.get('user-session-id')?.toString() ?? '';
-    const maxRecordLimit = parseInt(headers.get('max-record-limit') || '20', 10);
-
-    if (userSessionId == '') {
-      return Response.json({ chats: {} }, { status: 200 });
-    }
-    
-    let chatsRes = await db.query.chats.findMany({
-      where: eq(chats.userSessionId, userSessionId),
-    });
-    
-    chatsRes = chatsRes.reverse();
-    // Keep only the latest records in the database. Delete older records.
-    if (chatsRes.length > maxRecordLimit) {
-      const deleteChatsQuery = sql`DELETE FROM chats
-        WHERE userSessionId = ${userSessionId} AND (
-          timestamp IS NULL OR
-          timestamp NOT in (
-            SELECT timestamp FROM chats
-            WHERE userSessionId = ${userSessionId}
-            ORDER BY timestamp DESC
-            LIMIT ${maxRecordLimit}
-          )
-        )
-      `;
-      await db.run(deleteChatsQuery);
-      // Delete messages that no longer link with the chat from the database.
-      const deleteMessagesQuery = sql`DELETE FROM messages
-        WHERE chatId NOT IN (
-          SELECT id FROM chats
-        )
-      `;
-      await db.run(deleteMessagesQuery);
-    }
-
-    return Response.json({ chats: chatsRes }, { status: 200 });
+    let chats = await db.query.chats.findMany();
+    chats = chats.reverse();
+    return Response.json({ chats: chats }, { status: 200 });
   } catch (err) {
     console.error('Error in getting chats: ', err);
     return Response.json(

src/app/api/config/route.ts

@@ -7,7 +7,6 @@ import {
   getGroqApiKey,
   getOllamaApiEndpoint,
   getOpenaiApiKey,
-  getDeepseekApiKey,
   updateConfig,
 } from '@/lib/config';
 import {
@@ -54,7 +53,6 @@ export const GET = async (req: Request) => {
     config['anthropicApiKey'] = getAnthropicApiKey();
     config['groqApiKey'] = getGroqApiKey();
     config['geminiApiKey'] = getGeminiApiKey();
-    config['deepseekApiKey'] = getDeepseekApiKey();
     config['customOpenaiApiUrl'] = getCustomOpenaiApiUrl();
     config['customOpenaiApiKey'] = getCustomOpenaiApiKey();
     config['customOpenaiModelName'] = getCustomOpenaiModelName();
@@ -90,9 +88,6 @@ export const POST = async (req: Request) => {
         OLLAMA: {
           API_URL: config.ollamaApiUrl,
         },
-        DEEPSEEK: {
-          API_KEY: config.deepseekApiKey,
-        },
         CUSTOM_OPENAI: {
           API_URL: config.customOpenaiApiUrl,
           API_KEY: config.customOpenaiApiKey,

src/app/api/search/route.ts

@@ -34,7 +34,6 @@ interface ChatRequestBody {
   query: string;
   history: Array<[string, string]>;
   stream?: boolean;
-  systemInstructions?: string;
 }
 
 export const POST = async (req: Request) => {
@@ -126,7 +125,7 @@ export const POST = async (req: Request) => {
       embeddings,
       body.optimizationMode,
       [],
-      body.systemInstructions || '',
+      '',
     );
 
     if (!body.stream) {

src/app/library/page.tsx

@@ -1,6 +1,5 @@
 'use client';
 
-import crypto from 'crypto';
 import DeleteChat from '@/components/DeleteChat';
 import { cn, formatTimeDifference } from '@/lib/utils';
 import { BookOpenText, ClockIcon, Delete, ScanEye } from 'lucide-react';
@@ -22,34 +21,10 @@ const Page = () => {
     const fetchChats = async () => {
       setLoading(true);
 
-      let userSessionId = localStorage.getItem('userSessionId');
-      if (!userSessionId) {
-        userSessionId = crypto.randomBytes(20).toString('hex');
-        localStorage.setItem('userSessionId', userSessionId)
-      }
-
-      // Get maxRecordLimit from localStorage or set default
-      let maxRecordLimit = localStorage.getItem('maxRecordLimit');
-      if (!maxRecordLimit) {
-        maxRecordLimit = '20';
-        localStorage.setItem('maxRecordLimit', maxRecordLimit);
-      } else {
-        let valueInt = parseInt(maxRecordLimit, 10) || 20;
-        if (valueInt < 1) {
-          valueInt = 1;
-        } else if (valueInt > 100) {
-          valueInt = 100;
-        }
-        maxRecordLimit = valueInt.toString();
-        localStorage.setItem('maxRecordLimit', maxRecordLimit);
-      }
-
       const res = await fetch(`/api/chats`, {
         method: 'GET',
         headers: {
           'Content-Type': 'application/json',
-          'user-session-id': userSessionId!,
-          'max-record-limit': maxRecordLimit,
         },
       });
 

src/app/settings/page.tsx

@@ -20,11 +20,9 @@ interface SettingsType {
   anthropicApiKey: string;
   geminiApiKey: string;
   ollamaApiUrl: string;
-  deepseekApiKey: string;
   customOpenaiApiKey: string;
   customOpenaiApiUrl: string;
   customOpenaiModelName: string;
-  maxRecordLimit: string;
 }
 
 interface InputProps extends React.InputHTMLAttributes<HTMLInputElement> {
@@ -147,7 +145,6 @@ const Page = () => {
   const [automaticVideoSearch, setAutomaticVideoSearch] = useState(false);
   const [systemInstructions, setSystemInstructions] = useState<string>('');
   const [savingStates, setSavingStates] = useState<Record<string, boolean>>({});
-  const [maxRecordLimit, setMaxRecordLimit] = useState<string>('20');
 
   useEffect(() => {
     const fetchConfig = async () => {
@@ -210,8 +207,6 @@ const Page = () => {
 
       setSystemInstructions(localStorage.getItem('systemInstructions')!);
 
-      setMaxRecordLimit(localStorage.getItem('maxRecordLimit') || data.maxRecordLimit || '20');
-
       setIsLoading(false);
     };
 
@@ -370,15 +365,6 @@ const Page = () => {
         localStorage.setItem('embeddingModel', value);
       } else if (key === 'systemInstructions') {
         localStorage.setItem('systemInstructions', value);
-      } else if (key === 'maxRecordLimit') {
-        let valueInt = parseInt(value, 10) || 20;
-        if (valueInt < 1) {
-          valueInt = 1;
-        } else if (valueInt > 100) {
-          valueInt = 100;
-        }
-        setMaxRecordLimit(valueInt.toString());
-        localStorage.setItem('maxRecordLimit', valueInt.toString());
       }
     } catch (err) {
       console.error('Failed to save:', err);
@@ -852,56 +838,6 @@ const Page = () => {
                     onSave={(value) => saveConfig('geminiApiKey', value)}
                   />
                 </div>
-
-                <div className="flex flex-col space-y-1">
-                  <p className="text-black/70 dark:text-white/70 text-sm">
-                    Deepseek API Key
-                  </p>
-                  <Input
-                    type="text"
-                    placeholder="Deepseek API Key"
-                    value={config.deepseekApiKey}
-                    isSaving={savingStates['deepseekApiKey']}
-                    onChange={(e) => {
-                      setConfig((prev) => ({
-                        ...prev!,
-                        deepseekApiKey: e.target.value,
-                      }));
-                    }}
-                    onSave={(value) => saveConfig('deepseekApiKey', value)}
-                  />
-                </div>
-              </div>
-            </SettingsSection>
-
-            <SettingsSection title="Chat History">
-              <div className="flex flex-col space-y-4">
-                <div className="flex flex-col space-y-1">
-                  <p className="text-black/70 dark:text-white/70 text-sm">
-                    Maximum Chat History Records
-                  </p>
-                  <div className="flex items-center space-x-2">
-                    <Input
-                      type="number"
-                      min="1"
-                      max="100"
-                      pattern="[0-9]*"
-                      inputMode="numeric"
-                      value={maxRecordLimit}
-                      isSaving={savingStates['maxRecordLimit']}
-                      onChange={(e) => {
-                        setMaxRecordLimit(e.target.value);
-                      }}
-                      onSave={(value) => saveConfig('maxRecordLimit', value)}
-                    />
-                    <span className="text-black/60 dark:text-white/60 text-sm">
-                      records
-                    </span>
-                  </div>
-                  <p className="text-xs text-black/60 dark:text-white/60 mt-1">
-                    Maximum number of chat records to keep in history. Older records will be automatically deleted.
-                  </p>
-                </div>
               </div>
             </SettingsSection>
           </div>

src/components/ChatWindow.tsx

@@ -95,18 +95,6 @@ const checkConfig = async (
       if (!embeddingModel || !embeddingModelProvider) {
         const embeddingModelProviders = providers.embeddingModelProviders;
 
-        let userSessionId = localStorage.getItem('userSessionId');
-        if (!userSessionId) {
-          userSessionId = crypto.randomBytes(20).toString('hex');
-          localStorage.setItem('userSessionId', userSessionId!)
-        }
-
-        let maxRecordLimit = localStorage.getItem('maxRecordLimit');
-        if (!maxRecordLimit) {
-          maxRecordLimit = '20';
-          localStorage.setItem('maxRecordLimit', maxRecordLimit);
-        }
-
         if (
           !embeddingModelProviders ||
           Object.keys(embeddingModelProviders).length === 0
@@ -354,7 +342,6 @@ const ChatWindow = ({ id }: { id?: string }) => {
     let added = false;
 
     messageId = messageId ?? crypto.randomBytes(7).toString('hex');
-    let userSessionId = localStorage.getItem('userSessionId');
 
     setMessages((prevMessages) => [
       ...prevMessages,
@@ -479,7 +466,6 @@ const ChatWindow = ({ id }: { id?: string }) => {
           messageId: messageId,
           chatId: chatId!,
           content: message,
-          userSessionId: userSessionId,
         },
         chatId: chatId!,
         files: fileIds,

src/components/MessageBox.tsx

@@ -48,7 +48,6 @@ const MessageBox = ({
   const [speechMessage, setSpeechMessage] = useState(message.content);
 
   useEffect(() => {
-    const citationRegex = /\[([^\]]+)\]/g;
     const regex = /\[(\d+)\]/g;
     let processedMessage = message.content;
 
@@ -68,33 +67,11 @@ const MessageBox = ({
     ) {
       setParsedMessage(
         processedMessage.replace(
-          citationRegex,
-          (_, capturedContent: string) => {
-            const numbers = capturedContent
-              .split(',')
-              .map((numStr) => numStr.trim());
-
-            const linksHtml = numbers
-              .map((numStr) => {
-                const number = parseInt(numStr);
-
-                if (isNaN(number) || number <= 0) {
-                  return `[${numStr}]`;
-                }
-
-                const source = message.sources?.[number - 1];
-                const url = source?.metadata?.url;
-
-                if (url) {
-                  return `<a href="${url}" target="_blank" className="bg-light-secondary dark:bg-dark-secondary px-1 rounded ml-1 no-underline text-xs text-black/70 dark:text-white/70 relative">${numStr}</a>`;
-                } else {
-                  return `[${numStr}]`;
-                }
-              })
-              .join('');
-
-            return linksHtml;
-          },
+          regex,
+          (_, number) =>
+            `<a href="${
+              message.sources?.[number - 1]?.metadata?.url
+            }" target="_blank" className="bg-light-secondary dark:bg-dark-secondary px-1 rounded ml-1 no-underline text-xs text-black/70 dark:text-white/70 relative">${number}</a>`,
         ),
       );
       return;

src/lib/config.ts

@@ -25,9 +25,6 @@ interface Config {
     OLLAMA: {
       API_URL: string;
     };
-    DEEPSEEK: {
-      API_KEY: string;
-    };
     CUSTOM_OPENAI: {
       API_URL: string;
       API_KEY: string;
@@ -66,8 +63,6 @@ export const getSearxngApiEndpoint = () =>
 
 export const getOllamaApiEndpoint = () => loadConfig().MODELS.OLLAMA.API_URL;
 
-export const getDeepseekApiKey = () => loadConfig().MODELS.DEEPSEEK.API_KEY;
-
 export const getCustomOpenaiApiKey = () =>
   loadConfig().MODELS.CUSTOM_OPENAI.API_KEY;
 

src/lib/db/schema.ts

@@ -25,6 +25,4 @@ export const chats = sqliteTable('chats', {
   files: text('files', { mode: 'json' })
     .$type<File[]>()
     .default(sql`'[]'`),
-  userSessionId: text('userSessionId'),
-  timestamp: text('timestamp'),
 });

src/lib/providers/anthropic.ts

@@ -1,48 +1,22 @@
 import { ChatAnthropic } from '@langchain/anthropic';
-import { ChatModel } from '.';
+import { ChatModel, getModelsList, RawModel } from '.';
 import { getAnthropicApiKey } from '../config';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 
-const anthropicChatModels: Record<string, string>[] = [
-  {
-    displayName: 'Claude 3.7 Sonnet',
-    key: 'claude-3-7-sonnet-20250219',
-  },
-  {
-    displayName: 'Claude 3.5 Haiku',
-    key: 'claude-3-5-haiku-20241022',
-  },
-  {
-    displayName: 'Claude 3.5 Sonnet v2',
-    key: 'claude-3-5-sonnet-20241022',
-  },
-  {
-    displayName: 'Claude 3.5 Sonnet',
-    key: 'claude-3-5-sonnet-20240620',
-  },
-  {
-    displayName: 'Claude 3 Opus',
-    key: 'claude-3-opus-20240229',
-  },
-  {
-    displayName: 'Claude 3 Sonnet',
-    key: 'claude-3-sonnet-20240229',
-  },
-  {
-    displayName: 'Claude 3 Haiku',
-    key: 'claude-3-haiku-20240307',
-  },
-];
+const loadModels = () => {
+  return getModelsList()?.['chatModels']['anthropic']  as unknown as RawModel[]
+}
 
 export const loadAnthropicChatModels = async () => {
   const anthropicApiKey = getAnthropicApiKey();
-
   if (!anthropicApiKey) return {};
 
+  const models = loadModels()
+
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    anthropicChatModels.forEach((model) => {
+    models.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatAnthropic({

src/lib/providers/deepseek.ts

@@ -1,44 +0,0 @@
-import { ChatOpenAI } from '@langchain/openai';
-import { getDeepseekApiKey } from '../config';
-import { ChatModel } from '.';
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-
-const deepseekChatModels: Record<string, string>[] = [
-  {
-    displayName: 'Deepseek Chat (Deepseek V3)',
-    key: 'deepseek-chat',
-  },
-  {
-    displayName: 'Deepseek Reasoner (Deepseek R1)',
-    key: 'deepseek-reasoner',
-  },
-];
-
-export const loadDeepseekChatModels = async () => {
-  const deepseekApiKey = getDeepseekApiKey();
-
-  if (!deepseekApiKey) return {};
-
-  try {
-    const chatModels: Record<string, ChatModel> = {};
-
-    deepseekChatModels.forEach((model) => {
-      chatModels[model.key] = {
-        displayName: model.displayName,
-        model: new ChatOpenAI({
-          openAIApiKey: deepseekApiKey,
-          modelName: model.key,
-          temperature: 0.7,
-          configuration: {
-            baseURL: 'https://api.deepseek.com',
-          },
-        }) as unknown as BaseChatModel,
-      };
-    });
-
-    return chatModels;
-  } catch (err) {
-    console.error(`Error loading Deepseek models: ${err}`);
-    return {};
-  }
-};

src/lib/providers/gemini.ts

@@ -3,61 +3,24 @@ import {
   GoogleGenerativeAIEmbeddings,
 } from '@langchain/google-genai';
 import { getGeminiApiKey } from '../config';
-import { ChatModel, EmbeddingModel } from '.';
+import { ChatModel, EmbeddingModel, getModelsList, RawModel } from '.';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { Embeddings } from '@langchain/core/embeddings';
 
-const geminiChatModels: Record<string, string>[] = [
-  {
-    displayName: 'Gemini 2.5 Pro Experimental',
-    key: 'gemini-2.5-pro-exp-03-25',
-  },
-  {
-    displayName: 'Gemini 2.0 Flash',
-    key: 'gemini-2.0-flash',
-  },
-  {
-    displayName: 'Gemini 2.0 Flash-Lite',
-    key: 'gemini-2.0-flash-lite',
-  },
-  {
-    displayName: 'Gemini 2.0 Flash Thinking Experimental',
-    key: 'gemini-2.0-flash-thinking-exp-01-21',
-  },
-  {
-    displayName: 'Gemini 1.5 Flash',
-    key: 'gemini-1.5-flash',
-  },
-  {
-    displayName: 'Gemini 1.5 Flash-8B',
-    key: 'gemini-1.5-flash-8b',
-  },
-  {
-    displayName: 'Gemini 1.5 Pro',
-    key: 'gemini-1.5-pro',
-  },
-];
-
-const geminiEmbeddingModels: Record<string, string>[] = [
-  {
-    displayName: 'Text Embedding 004',
-    key: 'models/text-embedding-004',
-  },
-  {
-    displayName: 'Embedding 001',
-    key: 'models/embedding-001',
-  },
-];
+const loadModels = (modelType: 'chat' | 'embedding') => {
+  return getModelsList()?.[modelType === 'chat' ? 'chatModels' : 'embeddingModels']['gemini']  as unknown as RawModel[]
+}
 
 export const loadGeminiChatModels = async () => {
   const geminiApiKey = getGeminiApiKey();
-
   if (!geminiApiKey) return {};
 
+  const models = loadModels('chat');
+
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    geminiChatModels.forEach((model) => {
+    models.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatGoogleGenerativeAI({
@@ -77,13 +40,14 @@ export const loadGeminiChatModels = async () => {
 
 export const loadGeminiEmbeddingModels = async () => {
   const geminiApiKey = getGeminiApiKey();
-
   if (!geminiApiKey) return {};
 
+  const models = loadModels('embedding');
+
   try {
     const embeddingModels: Record<string, EmbeddingModel> = {};
 
-    geminiEmbeddingModels.forEach((model) => {
+    models.forEach((model) => {
       embeddingModels[model.key] = {
         displayName: model.displayName,
         model: new GoogleGenerativeAIEmbeddings({

src/lib/providers/groq.ts

@@ -1,96 +1,22 @@
 import { ChatOpenAI } from '@langchain/openai';
 import { getGroqApiKey } from '../config';
-import { ChatModel } from '.';
+import { ChatModel, getModelsList, RawModel } from '.';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 
-const groqChatModels: Record<string, string>[] = [
-  {
-    displayName: 'Gemma2 9B IT',
-    key: 'gemma2-9b-it',
-  },
-  {
-    displayName: 'Llama 3.3 70B Versatile',
-    key: 'llama-3.3-70b-versatile',
-  },
-  {
-    displayName: 'Llama 3.1 8B Instant',
-    key: 'llama-3.1-8b-instant',
-  },
-  {
-    displayName: 'Llama3 70B 8192',
-    key: 'llama3-70b-8192',
-  },
-  {
-    displayName: 'Llama3 8B 8192',
-    key: 'llama3-8b-8192',
-  },
-  {
-    displayName: 'Mixtral 8x7B 32768',
-    key: 'mixtral-8x7b-32768',
-  },
-  {
-    displayName: 'Qwen QWQ 32B (Preview)',
-    key: 'qwen-qwq-32b',
-  },
-  {
-    displayName: 'Mistral Saba 24B (Preview)',
-    key: 'mistral-saba-24b',
-  },
-  {
-    displayName: 'Qwen 2.5 Coder 32B (Preview)',
-    key: 'qwen-2.5-coder-32b',
-  },
-  {
-    displayName: 'Qwen 2.5 32B (Preview)',
-    key: 'qwen-2.5-32b',
-  },
-  {
-    displayName: 'DeepSeek R1 Distill Qwen 32B (Preview)',
-    key: 'deepseek-r1-distill-qwen-32b',
-  },
-  {
-    displayName: 'DeepSeek R1 Distill Llama 70B (Preview)',
-    key: 'deepseek-r1-distill-llama-70b',
-  },
-  {
-    displayName: 'Llama 3.3 70B SpecDec (Preview)',
-    key: 'llama-3.3-70b-specdec',
-  },
-  {
-    displayName: 'Llama 3.2 1B Preview (Preview)',
-    key: 'llama-3.2-1b-preview',
-  },
-  {
-    displayName: 'Llama 3.2 3B Preview (Preview)',
-    key: 'llama-3.2-3b-preview',
-  },
-  {
-    displayName: 'Llama 3.2 11B Vision Preview (Preview)',
-    key: 'llama-3.2-11b-vision-preview',
-  },
-  {
-    displayName: 'Llama 3.2 90B Vision Preview (Preview)',
-    key: 'llama-3.2-90b-vision-preview',
-  },
-  /* {
-    displayName: 'Llama 4 Maverick 17B 128E Instruct (Preview)',
-    key: 'meta-llama/llama-4-maverick-17b-128e-instruct',
-  }, */
-  {
-    displayName: 'Llama 4 Scout 17B 16E Instruct (Preview)',
-    key: 'meta-llama/llama-4-scout-17b-16e-instruct',
-  },
-];
+const loadModels = () => {
+  return getModelsList()?.chatModels['groq'] as unknown as RawModel[]
+}
 
 export const loadGroqChatModels = async () => {
   const groqApiKey = getGroqApiKey();
-
   if (!groqApiKey) return {};
 
+  const models = loadModels()
+
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    groqChatModels.forEach((model) => {
+    models.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatOpenAI({

src/lib/providers/index.ts

@@ -1,27 +1,39 @@
-import { Embeddings } from '@langchain/core/embeddings';
-import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { loadOpenAIChatModels, loadOpenAIEmbeddingModels } from './openai';
+import { Embeddings } from '@langchain/core/embeddings'
+import { BaseChatModel } from '@langchain/core/language_models/chat_models'
+import { loadOpenAIChatModels, loadOpenAIEmbeddingModels } from './openai'
 import {
   getCustomOpenaiApiKey,
   getCustomOpenaiApiUrl,
   getCustomOpenaiModelName,
-} from '../config';
-import { ChatOpenAI } from '@langchain/openai';
-import { loadOllamaChatModels, loadOllamaEmbeddingModels } from './ollama';
-import { loadGroqChatModels } from './groq';
-import { loadAnthropicChatModels } from './anthropic';
-import { loadGeminiChatModels, loadGeminiEmbeddingModels } from './gemini';
-import { loadTransformersEmbeddingsModels } from './transformers';
-import { loadDeepseekChatModels } from './deepseek';
+} from '../config'
+import { ChatOpenAI } from '@langchain/openai'
+import { loadOllamaChatModels, loadOllamaEmbeddingModels } from './ollama'
+import { loadGroqChatModels } from './groq'
+import { loadAnthropicChatModels } from './anthropic'
+import { loadGeminiChatModels, loadGeminiEmbeddingModels } from './gemini'
+import { loadTransformersEmbeddingsModels } from './transformers'
+import path from 'path'
+import fs from 'fs'
 
 export interface ChatModel {
-  displayName: string;
-  model: BaseChatModel;
+  displayName: string
+  model: BaseChatModel
 }
 
 export interface EmbeddingModel {
-  displayName: string;
-  model: Embeddings;
+  displayName: string
+  model: Embeddings
+}
+
+export type RawModel = {
+  displayName: string
+  key: string
+}
+
+type ModelsList = {
+  [key in "chatModels" | "embeddingModels"]: {
+    [key: string]: RawModel[]
+  }
 }
 
 export const chatModelProviders: Record<
@@ -33,8 +45,7 @@ export const chatModelProviders: Record<
   groq: loadGroqChatModels,
   anthropic: loadAnthropicChatModels,
   gemini: loadGeminiChatModels,
-  deepseek: loadDeepseekChatModels,
-};
+}
 
 export const embeddingModelProviders: Record<
   string,
@@ -44,21 +55,43 @@ export const embeddingModelProviders: Record<
   ollama: loadOllamaEmbeddingModels,
   gemini: loadGeminiEmbeddingModels,
   transformers: loadTransformersEmbeddingsModels,
-};
+}
+
+export const getModelsList = (): ModelsList | null => {
+  const modelFile = path.join(process.cwd(), 'data/models.json')
+  try {
+    const content = fs.readFileSync(modelFile, 'utf-8')
+    return JSON.parse(content) as ModelsList
+  } catch (err) {
+    console.error(`Error reading models file: ${err}`)
+    return null
+  }
+}
+
+export const updateModelsList = (models: ModelsList) => {
+  try {
+    const modelFile = path.join(process.cwd(), 'data/models.json')
+    const content = JSON.stringify(models, null, 2)
+
+    fs.writeFileSync(modelFile, content, 'utf-8')
+  } catch(err) {
+    console.error(`Error updating models file: ${err}`)
+  }
+}
 
 export const getAvailableChatModelProviders = async () => {
-  const models: Record<string, Record<string, ChatModel>> = {};
+  const models: Record<string, Record<string, ChatModel>> = {}
 
   for (const provider in chatModelProviders) {
-    const providerModels = await chatModelProviders[provider]();
+    const providerModels = await chatModelProviders[provider]()
     if (Object.keys(providerModels).length > 0) {
-      models[provider] = providerModels;
+      models[provider] = providerModels
     }
   }
 
-  const customOpenAiApiKey = getCustomOpenaiApiKey();
-  const customOpenAiApiUrl = getCustomOpenaiApiUrl();
-  const customOpenAiModelName = getCustomOpenaiModelName();
+  const customOpenAiApiKey = getCustomOpenaiApiKey()
+  const customOpenAiApiUrl = getCustomOpenaiApiUrl()
+  const customOpenAiModelName = getCustomOpenaiModelName()
 
   models['custom_openai'] = {
     ...(customOpenAiApiKey && customOpenAiApiUrl && customOpenAiModelName
@@ -76,20 +109,20 @@ export const getAvailableChatModelProviders = async () => {
           },
         }
       : {}),
-  };
+  }
 
-  return models;
-};
+  return models
+}
 
 export const getAvailableEmbeddingModelProviders = async () => {
-  const models: Record<string, Record<string, EmbeddingModel>> = {};
+  const models: Record<string, Record<string, EmbeddingModel>> = {}
 
   for (const provider in embeddingModelProviders) {
-    const providerModels = await embeddingModelProviders[provider]();
+    const providerModels = await embeddingModelProviders[provider]()
     if (Object.keys(providerModels).length > 0) {
-      models[provider] = providerModels;
+      models[provider] = providerModels
     }
   }
 
-  return models;
-};
+  return models
+}

src/lib/providers/ollama.ts

@@ -1,24 +1,39 @@
-import axios from 'axios';
-import { getKeepAlive, getOllamaApiEndpoint } from '../config';
-import { ChatModel, EmbeddingModel } from '.';
-import { ChatOllama } from '@langchain/community/chat_models/ollama';
-import { OllamaEmbeddings } from '@langchain/community/embeddings/ollama';
-
-export const loadOllamaChatModels = async () => {
-  const ollamaApiEndpoint = getOllamaApiEndpoint();
-
-  if (!ollamaApiEndpoint) return {};
+import axios from 'axios'
+import { getKeepAlive, getOllamaApiEndpoint } from '../config'
+import { ChatModel, EmbeddingModel } from '.'
+import { ChatOllama } from '@langchain/community/chat_models/ollama'
+import { OllamaEmbeddings } from '@langchain/community/embeddings/ollama'
 
+const loadModels = async (apiURL: string) => {
   try {
-    const res = await axios.get(`${ollamaApiEndpoint}/api/tags`, {
+    const res = await axios.get(`${apiURL}/api/tags`, {
       headers: {
         'Content-Type': 'application/json',
       },
-    });
+    })
 
-    const { models } = res.data;
+    if (res.status !== 200) {
+      console.error(`Failed to load Ollama models: ${res.data}`)
+      return []
+    }
 
-    const chatModels: Record<string, ChatModel> = {};
+    const { models } = res.data
+
+    return models
+  } catch (err) {
+    console.error(`Error loading Ollama models: ${err}`)
+    return []
+  }
+}
+
+export const loadOllamaChatModels = async () => {
+  const ollamaApiEndpoint = getOllamaApiEndpoint()
+  if (!ollamaApiEndpoint) return {}
+
+  const models = await loadModels(ollamaApiEndpoint)
+
+  try {
+    const chatModels: Record<string, ChatModel> = {}
 
     models.forEach((model: any) => {
       chatModels[model.model] = {
@@ -29,31 +44,24 @@ export const loadOllamaChatModels = async () => {
           temperature: 0.7,
           keepAlive: getKeepAlive(),
         }),
-      };
-    });
+      }
+    })
 
-    return chatModels;
+    return chatModels
   } catch (err) {
-    console.error(`Error loading Ollama models: ${err}`);
-    return {};
+    console.error(`Error loading Ollama models: ${err}`)
+    return {}
   }
-};
+}
 
 export const loadOllamaEmbeddingModels = async () => {
-  const ollamaApiEndpoint = getOllamaApiEndpoint();
+  const ollamaApiEndpoint = getOllamaApiEndpoint()
+  if (!ollamaApiEndpoint) return {}
 
-  if (!ollamaApiEndpoint) return {};
+  const models = await loadModels(ollamaApiEndpoint)
 
   try {
-    const res = await axios.get(`${ollamaApiEndpoint}/api/tags`, {
-      headers: {
-        'Content-Type': 'application/json',
-      },
-    });
-
-    const { models } = res.data;
-
-    const embeddingModels: Record<string, EmbeddingModel> = {};
+    const embeddingModels: Record<string, EmbeddingModel> = {}
 
     models.forEach((model: any) => {
       embeddingModels[model.model] = {
@@ -62,12 +70,12 @@ export const loadOllamaEmbeddingModels = async () => {
           baseUrl: ollamaApiEndpoint,
           model: model.model,
         }),
-      };
-    });
+      }
+    })
 
-    return embeddingModels;
+    return embeddingModels
   } catch (err) {
-    console.error(`Error loading Ollama embeddings models: ${err}`);
-    return {};
+    console.error(`Error loading Ollama embeddings models: ${err}`)
+    return {}
   }
-};
+}

src/lib/providers/openai.ts

@@ -1,52 +1,23 @@
 import { ChatOpenAI, OpenAIEmbeddings } from '@langchain/openai';
 import { getOpenaiApiKey } from '../config';
-import { ChatModel, EmbeddingModel } from '.';
+import { ChatModel, EmbeddingModel, getModelsList, RawModel } from '.';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { Embeddings } from '@langchain/core/embeddings';
 
-const openaiChatModels: Record<string, string>[] = [
-  {
-    displayName: 'GPT-3.5 Turbo',
-    key: 'gpt-3.5-turbo',
-  },
-  {
-    displayName: 'GPT-4',
-    key: 'gpt-4',
-  },
-  {
-    displayName: 'GPT-4 turbo',
-    key: 'gpt-4-turbo',
-  },
-  {
-    displayName: 'GPT-4 omni',
-    key: 'gpt-4o',
-  },
-  {
-    displayName: 'GPT-4 omni mini',
-    key: 'gpt-4o-mini',
-  },
-];
-
-const openaiEmbeddingModels: Record<string, string>[] = [
-  {
-    displayName: 'Text Embedding 3 Small',
-    key: 'text-embedding-3-small',
-  },
-  {
-    displayName: 'Text Embedding 3 Large',
-    key: 'text-embedding-3-large',
-  },
-];
+const loadModels = (modelType: 'chat' | 'embedding') => {
+  return getModelsList()?.[modelType === 'chat' ? 'chatModels' : 'embeddingModels']['openai']  as unknown as RawModel[]
+}
 
 export const loadOpenAIChatModels = async () => {
   const openaiApiKey = getOpenaiApiKey();
+  const models = loadModels('chat');
 
-  if (!openaiApiKey) return {};
+  if (!openaiApiKey || !models) return {};
 
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    openaiChatModels.forEach((model) => {
+    models.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatOpenAI({
@@ -66,13 +37,14 @@ export const loadOpenAIChatModels = async () => {
 
 export const loadOpenAIEmbeddingModels = async () => {
   const openaiApiKey = getOpenaiApiKey();
+  const models = loadModels('embedding');
 
-  if (!openaiApiKey) return {};
+  if (!openaiApiKey || !models) return {};
 
   try {
     const embeddingModels: Record<string, EmbeddingModel> = {};
 
-    openaiEmbeddingModels.forEach((model) => {
+    models.forEach((model) => {
       embeddingModels[model.key] = {
         displayName: model.displayName,
         model: new OpenAIEmbeddings({

src/lib/providers/transformers.ts

@@ -1,31 +1,30 @@
-import { HuggingFaceTransformersEmbeddings } from '../huggingfaceTransformer';
+import { EmbeddingModel, getModelsList, RawModel } from '.'
+import { HuggingFaceTransformersEmbeddings } from '../huggingfaceTransformer'
+
+const loadModels = () => {
+  return getModelsList()?.embeddingModels[
+    'transformers'
+  ] as unknown as RawModel[]
+}
 
 export const loadTransformersEmbeddingsModels = async () => {
   try {
-    const embeddingModels = {
-      'xenova-bge-small-en-v1.5': {
-        displayName: 'BGE Small',
-        model: new HuggingFaceTransformersEmbeddings({
-          modelName: 'Xenova/bge-small-en-v1.5',
-        }),
-      },
-      'xenova-gte-small': {
-        displayName: 'GTE Small',
-        model: new HuggingFaceTransformersEmbeddings({
-          modelName: 'Xenova/gte-small',
-        }),
-      },
-      'xenova-bert-base-multilingual-uncased': {
-        displayName: 'Bert Multilingual',
-        model: new HuggingFaceTransformersEmbeddings({
-          modelName: 'Xenova/bert-base-multilingual-uncased',
-        }),
-      },
-    };
+    const models = loadModels()
 
-    return embeddingModels;
+    const embeddingModels: Record<string, EmbeddingModel> = {}
+
+    models.forEach(model => {
+      embeddingModels[model.key] = {
+        displayName: model.displayName,
+        model: new HuggingFaceTransformersEmbeddings({
+          modelName: model.key,
+        }),
+      }
+    })
+
+    return embeddingModels
   } catch (err) {
-    console.error(`Error loading Transformers embeddings model: ${err}`);
-    return {};
+    console.error(`Error loading Transformers embeddings model: ${err}`)
+    return {}
   }
-};
+}