Merge 28b9cca413966040a7904d64e0c001606a823b96 into da1123d84baa8ced8a9427f866b068b4652e7d9c

feat(search): add optional systemInstructions to API request body

README.md

@@ -159,6 +159,7 @@ Perplexica runs on Next.js and handles all API requests. It works right away on
 
 [![Deploy to Sealos](https://raw.githubusercontent.com/labring-actions/templates/main/Deploy-on-Sealos.svg)](https://usw.sealos.io/?openapp=system-template%3FtemplateName%3Dperplexica)
 [![Deploy to RepoCloud](https://d16t0pc4846x52.cloudfront.net/deploylobe.svg)](https://repocloud.io/details/?app_id=267)
+[![Run on ClawCloud](https://raw.githubusercontent.com/ClawCloud/Run-Template/refs/heads/main/Run-on-ClawCloud.svg)](https://template.run.claw.cloud/?referralCode=U11MRQ8U9RM4&openapp=system-fastdeploy%3FtemplateName%3Dperplexica)
 
 ## Upcoming Features
 

data/.gitignore

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

data/models.json

@@ -1,157 +0,0 @@
-{
-  "_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

@@ -33,6 +33,7 @@ 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
 }
 ```
@@ -63,6 +64,8 @@ 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": "1.10.1",
+  "version": "1.10.2",
   "license": "MIT",
   "author": "ItzCrazyKns",
   "scripts": {

sample.config.toml

@@ -22,5 +22,8 @@ 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/config/route.ts

@@ -7,6 +7,7 @@ import {
   getGroqApiKey,
   getOllamaApiEndpoint,
   getOpenaiApiKey,
+  getDeepseekApiKey,
   updateConfig,
 } from '@/lib/config';
 import {
@@ -53,6 +54,7 @@ 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();
@@ -88,6 +90,9 @@ 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

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

src/app/settings/page.tsx

@@ -20,6 +20,7 @@ interface SettingsType {
   anthropicApiKey: string;
   geminiApiKey: string;
   ollamaApiUrl: string;
+  deepseekApiKey: string;
   customOpenaiApiKey: string;
   customOpenaiApiUrl: string;
   customOpenaiModelName: string;
@@ -838,6 +839,25 @@ 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>
           </div>

src/components/MessageBox.tsx

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

src/lib/config.ts

@@ -25,6 +25,9 @@ interface Config {
     OLLAMA: {
       API_URL: string;
     };
+    DEEPSEEK: {
+      API_KEY: string;
+    };
     CUSTOM_OPENAI: {
       API_URL: string;
       API_KEY: string;
@@ -63,6 +66,8 @@ 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/providers/anthropic.ts

@@ -1,22 +1,48 @@
 import { ChatAnthropic } from '@langchain/anthropic';
-import { ChatModel, getModelsList, RawModel } from '.';
+import { ChatModel } from '.';
 import { getAnthropicApiKey } from '../config';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 
-const loadModels = () => {
+const anthropicChatModels: Record<string, string>[] = [
-  return getModelsList()?.['chatModels']['anthropic']  as unknown as RawModel[]
+  {
-}
+    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',
+  },
+];
 
 export const loadAnthropicChatModels = async () => {
   const anthropicApiKey = getAnthropicApiKey();
-  if (!anthropicApiKey) return {};
 
-  const models = loadModels()
+  if (!anthropicApiKey) return {};
 
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    models.forEach((model) => {
+    anthropicChatModels.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatAnthropic({

src/lib/providers/deepseek.ts

@@ -0,0 +1,44 @@
+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

@@ -3,24 +3,61 @@ import {
   GoogleGenerativeAIEmbeddings,
 } from '@langchain/google-genai';
 import { getGeminiApiKey } from '../config';
-import { ChatModel, EmbeddingModel, getModelsList, RawModel } from '.';
+import { ChatModel, EmbeddingModel } from '.';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 import { Embeddings } from '@langchain/core/embeddings';
 
-const loadModels = (modelType: 'chat' | 'embedding') => {
+const geminiChatModels: Record<string, string>[] = [
-  return getModelsList()?.[modelType === 'chat' ? 'chatModels' : 'embeddingModels']['gemini']  as unknown as RawModel[]
+  {
-}
+    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',
+  },
+];
 
 export const loadGeminiChatModels = async () => {
   const geminiApiKey = getGeminiApiKey();
-  if (!geminiApiKey) return {};
 
-  const models = loadModels('chat');
+  if (!geminiApiKey) return {};
 
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    models.forEach((model) => {
+    geminiChatModels.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatGoogleGenerativeAI({
@@ -40,14 +77,13 @@ export const loadGeminiChatModels = async () => {
 
 export const loadGeminiEmbeddingModels = async () => {
   const geminiApiKey = getGeminiApiKey();
-  if (!geminiApiKey) return {};
 
-  const models = loadModels('embedding');
+  if (!geminiApiKey) return {};
 
   try {
     const embeddingModels: Record<string, EmbeddingModel> = {};
 
-    models.forEach((model) => {
+    geminiEmbeddingModels.forEach((model) => {
       embeddingModels[model.key] = {
         displayName: model.displayName,
         model: new GoogleGenerativeAIEmbeddings({

src/lib/providers/groq.ts

@@ -1,22 +1,96 @@
 import { ChatOpenAI } from '@langchain/openai';
 import { getGroqApiKey } from '../config';
-import { ChatModel, getModelsList, RawModel } from '.';
+import { ChatModel } from '.';
 import { BaseChatModel } from '@langchain/core/language_models/chat_models';
 
-const loadModels = () => {
+const groqChatModels: Record<string, string>[] = [
-  return getModelsList()?.chatModels['groq'] as unknown as RawModel[]
+  {
-}
+    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',
+  },
+];
 
 export const loadGroqChatModels = async () => {
   const groqApiKey = getGroqApiKey();
-  if (!groqApiKey) return {};
 
-  const models = loadModels()
+  if (!groqApiKey) return {};
 
   try {
     const chatModels: Record<string, ChatModel> = {};
 
-    models.forEach((model) => {
+    groqChatModels.forEach((model) => {
       chatModels[model.key] = {
         displayName: model.displayName,
         model: new ChatOpenAI({

src/lib/providers/index.ts

@@ -1,39 +1,27 @@
-import { Embeddings } from '@langchain/core/embeddings'
+import { Embeddings } from '@langchain/core/embeddings';
-import { BaseChatModel } from '@langchain/core/language_models/chat_models'
+import { BaseChatModel } from '@langchain/core/language_models/chat_models';
-import { loadOpenAIChatModels, loadOpenAIEmbeddingModels } from './openai'
+import { loadOpenAIChatModels, loadOpenAIEmbeddingModels } from './openai';
 import {
   getCustomOpenaiApiKey,
   getCustomOpenaiApiUrl,
   getCustomOpenaiModelName,
-} from '../config'
+} from '../config';
-import { ChatOpenAI } from '@langchain/openai'
+import { ChatOpenAI } from '@langchain/openai';
-import { loadOllamaChatModels, loadOllamaEmbeddingModels } from './ollama'
+import { loadOllamaChatModels, loadOllamaEmbeddingModels } from './ollama';
-import { loadGroqChatModels } from './groq'
+import { loadGroqChatModels } from './groq';
-import { loadAnthropicChatModels } from './anthropic'
+import { loadAnthropicChatModels } from './anthropic';
-import { loadGeminiChatModels, loadGeminiEmbeddingModels } from './gemini'
+import { loadGeminiChatModels, loadGeminiEmbeddingModels } from './gemini';
-import { loadTransformersEmbeddingsModels } from './transformers'
+import { loadTransformersEmbeddingsModels } from './transformers';
-import path from 'path'
+import { loadDeepseekChatModels } from './deepseek';
-import fs from 'fs'
 
 export interface ChatModel {
-  displayName: string
+  displayName: string;
-  model: BaseChatModel
+  model: BaseChatModel;
 }
 
 export interface EmbeddingModel {
-  displayName: string
+  displayName: string;
-  model: Embeddings
+  model: Embeddings;
-}
-
-export type RawModel = {
-  displayName: string
-  key: string
-}
-
-type ModelsList = {
-  [key in "chatModels" | "embeddingModels"]: {
-    [key: string]: RawModel[]
-  }
 }
 
 export const chatModelProviders: Record<
@@ -45,7 +33,8 @@ export const chatModelProviders: Record<
   groq: loadGroqChatModels,
   anthropic: loadAnthropicChatModels,
   gemini: loadGeminiChatModels,
-}
+  deepseek: loadDeepseekChatModels,
+};
 
 export const embeddingModelProviders: Record<
   string,
@@ -55,43 +44,21 @@ 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 customOpenAiApiKey = getCustomOpenaiApiKey();
-  const customOpenAiApiUrl = getCustomOpenaiApiUrl()
+  const customOpenAiApiUrl = getCustomOpenaiApiUrl();
-  const customOpenAiModelName = getCustomOpenaiModelName()
+  const customOpenAiModelName = getCustomOpenaiModelName();
 
   models['custom_openai'] = {
     ...(customOpenAiApiKey && customOpenAiApiUrl && customOpenAiModelName
@@ -109,20 +76,20 @@ 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

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

src/lib/providers/openai.ts

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

src/lib/providers/transformers.ts

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