fix: update LMStudio endpoint from /health to /v1/models

- Replace health check endpoint with proper model listing endpoint
- Remove workaround that returned 200 for unexpected GET /v1/health

docker-compose.yaml

@@ -37,6 +37,7 @@ services:
       args:
         - NEXT_PUBLIC_API_URL=http://127.0.0.1:3001/api
         - NEXT_PUBLIC_WS_URL=ws://127.0.0.1:3001
+      network: host
     image: itzcrazykns1337/perplexica-frontend:main
     depends_on:
       - perplexica-backend

docs/installation/UPDATING.md

@@ -10,23 +10,25 @@ To update Perplexica to the latest version, follow these steps:
    git clone https://github.com/ItzCrazyKns/Perplexica.git
    ```
 
-2. Navigate to the Project Directory.
+2. Navigate to the project directory.
 
-3. Pull latest images from registry.
+3. Check for changes in the configuration files. If the `sample.config.toml` file contains new fields, delete your existing `config.toml` file, rename `sample.config.toml` to `config.toml`, and update the configuration accordingly.
+
+4. Pull the latest images from the registry.
 
    ```bash
    docker compose pull
    ```
 
-4. Update and Recreate containers.
+5. Update and recreate the containers.
 
    ```bash
    docker compose up -d
    ```
 
-5. Once the command completes running go to http://localhost:3000 and verify the latest changes.
+6. Once the command completes, go to http://localhost:3000 and verify the latest changes.
 
-## For non Docker users
+## For non-Docker users
 
 1. Clone the latest version of Perplexica from GitHub:
 
@@ -34,7 +36,14 @@ To update Perplexica to the latest version, follow these steps:
    git clone https://github.com/ItzCrazyKns/Perplexica.git
    ```
 
-2. Navigate to the Project Directory
+2. Navigate to the project directory.
-3. Execute `npm i` in both the `ui` folder and the root directory.
+
-4. Once packages are updated, execute `npm run build` in both the `ui` folder and the root directory.
+3. Check for changes in the configuration files. If the `sample.config.toml` file contains new fields, delete your existing `config.toml` file, rename `sample.config.toml` to `config.toml`, and update the configuration accordingly.
-5. Finally, start both the frontend and the backend by running `npm run start` in both the `ui` folder and the root directory.
+
+4. Execute `npm i` in both the `ui` folder and the root directory.
+
+5. Once the packages are updated, execute `npm run build` in both the `ui` folder and the root directory.
+
+6. Finally, start both the frontend and the backend by running `npm run start` in both the `ui` folder and the root directory.
+
+---

sample.config.toml

@@ -15,12 +15,16 @@ API_KEY = ""
 [MODELS.GEMINI]
 API_KEY = ""
 
-[MODELS.CUSTOM_OPENAI]
-API_KEY = ""
-API_URL = ""
-
 [MODELS.OLLAMA]
 API_URL = "" # Ollama API URL - http://host.docker.internal:11434
 
+[MODELS.LMSTUDIO]
+API_URL = "" # LM STUDIO API URL - http://host.docker.internal:1234
+
+[MODELS.CUSTOM_OPENAI]
+API_KEY = ""
+API_URL = ""
+MODEL_NAME = ""
+
 [API_ENDPOINTS]
 SEARXNG = "http://localhost:32768" # SearxNG API URL

src/config.ts

@@ -26,6 +26,9 @@ interface Config {
     OLLAMA: {
       API_URL: string;
     };
+    LMSTUDIO: { 
+      API_URL: string;
+    };
     CUSTOM_OPENAI: {
       API_URL: string;
       API_KEY: string;
@@ -66,6 +69,8 @@ export const getSearxngApiEndpoint = () =>
 
 export const getOllamaApiEndpoint = () => loadConfig().MODELS.OLLAMA.API_URL;
 
+export const getLMStudioApiEndpoint = () => loadConfig().MODELS.LMSTUDIO.API_URL;
+
 export const getCustomOpenaiApiKey = () =>
   loadConfig().MODELS.CUSTOM_OPENAI.API_KEY;
 
@@ -76,10 +81,6 @@ export const getCustomOpenaiModelName = () =>
   loadConfig().MODELS.CUSTOM_OPENAI.MODEL_NAME;
 
 const mergeConfigs = (current: any, update: any): any => {
-  if (update === null || update === undefined) {
-    return current;
-  }
-
   if (typeof current !== 'object' || current === null) {
     return update;
   }

src/lib/chat/service.ts

@@ -0,0 +1 @@
+ 

src/lib/providers/index.ts

@@ -4,6 +4,7 @@ import { loadOpenAIChatModels, loadOpenAIEmbeddingsModels } from './openai';
 import { loadAnthropicChatModels } from './anthropic';
 import { loadTransformersEmbeddingsModels } from './transformers';
 import { loadGeminiChatModels, loadGeminiEmbeddingsModels } from './gemini';
+import { loadLMStudioChatModels, loadLMStudioEmbeddingsModels } from './lmstudio';
 import {
   getCustomOpenaiApiKey,
   getCustomOpenaiApiUrl,
@@ -17,6 +18,7 @@ const chatModelProviders = {
   ollama: loadOllamaChatModels,
   anthropic: loadAnthropicChatModels,
   gemini: loadGeminiChatModels,
+  lm_studio: loadLMStudioChatModels,
 };
 
 const embeddingModelProviders = {
@@ -24,6 +26,7 @@ const embeddingModelProviders = {
   local: loadTransformersEmbeddingsModels,
   ollama: loadOllamaEmbeddingsModels,
   gemini: loadGeminiEmbeddingsModels,
+  lm_studio: loadLMStudioEmbeddingsModels,
 };
 
 export const getAvailableChatModelProviders = async () => {

src/lib/providers/lmstudio.ts

@@ -0,0 +1,125 @@
+import { OpenAIEmbeddings } from '@langchain/openai';
+import { ChatOpenAI } from '@langchain/openai';
+import { getKeepAlive, getLMStudioApiEndpoint } from '../../config';
+import logger from '../../utils/logger';
+import axios from 'axios';
+
+const ensureV1Endpoint = (endpoint: string): string => {
+  return endpoint.endsWith('/v1') ? endpoint : `${endpoint}/v1`;
+};
+
+interface LMStudioModel {
+  id: string;
+  // add other properties if LM Studio API provides them
+}
+
+interface ChatModelConfig {
+  displayName: string;
+  model: ChatOpenAI;
+}
+
+const checkLMStudioAvailability = async (endpoint: string): Promise<boolean> => {
+  const v1Endpoint = ensureV1Endpoint(endpoint);
+  try {
+    await axios.get(`${v1Endpoint}/models`, {
+      timeout: 1000, // 1 second timeout
+      headers: {
+        'Content-Type': 'application/json',
+      },
+    });
+    return true;
+  } catch (err) {
+    logger.debug(`LM Studio server not available at ${endpoint}`);
+    return false;
+  }
+};
+
+export const loadLMStudioChatModels = async (): Promise<Record<string, ChatModelConfig>> => {
+  const lmStudioEndpoint = getLMStudioApiEndpoint();
+
+  if (!lmStudioEndpoint) {
+    logger.debug('LM Studio endpoint not configured, skipping');
+    return {};
+  }
+
+  // Check if server is available before attempting to load models
+  const isAvailable = await checkLMStudioAvailability(lmStudioEndpoint);
+  if (!isAvailable) {
+    return {};
+  }
+
+  try {
+    const v1Endpoint = ensureV1Endpoint(lmStudioEndpoint);
+    const response = await axios.get<{ data: LMStudioModel[] }>(`${v1Endpoint}/models`, {
+      timeout: 5000, // 5 second timeout for model loading
+      headers: {
+        'Content-Type': 'application/json',
+      },
+    });
+
+    const lmStudioModels = response.data.data;
+
+    const chatModels = lmStudioModels.reduce<Record<string, ChatModelConfig>>((acc, model) => {
+      acc[model.id] = {
+        displayName: model.id,
+        model: new ChatOpenAI({
+          openAIApiKey: 'lm-studio',
+          configuration: {
+            baseURL: ensureV1Endpoint(lmStudioEndpoint),
+          },
+          modelName: model.id,
+          temperature: 0.7,
+        }),
+      };
+      return acc;
+    }, {});
+
+    return chatModels;
+  } catch (err) {
+    logger.error(`Error loading LM Studio models: ${err}`);
+    return {};
+  }
+};
+
+export const loadLMStudioEmbeddingsModels = async () => {
+  const lmStudioEndpoint = getLMStudioApiEndpoint();
+
+  if (!lmStudioEndpoint) return {};
+
+  // Check if server is available before attempting to load models
+  const isAvailable = await checkLMStudioAvailability(lmStudioEndpoint);
+  if (!isAvailable) {
+    return {};
+  }
+
+  try {
+    const v1Endpoint = ensureV1Endpoint(lmStudioEndpoint);
+    const response = await axios.get(`${v1Endpoint}/models`, {
+      timeout: 5000, // 5 second timeout for model loading
+      headers: {
+        'Content-Type': 'application/json',
+      },
+    });
+
+    const lmStudioModels = response.data.data;
+
+    const embeddingsModels = lmStudioModels.reduce((acc, model) => {
+      acc[model.id] = {
+        displayName: model.id,
+        model: new OpenAIEmbeddings({
+          openAIApiKey: 'lm-studio', // Dummy key required by LangChain
+          configuration: {
+            baseURL: ensureV1Endpoint(lmStudioEndpoint),
+          },
+          modelName: model.id,
+        }),
+      };
+      return acc;
+    }, {});
+
+    return embeddingsModels;
+  } catch (err) {
+    logger.error(`Error loading LM Studio embeddings model: ${err}`);
+    return {};
+  }
+};

src/routes/config.ts

@@ -6,6 +6,7 @@ import {
 import {
   getGroqApiKey,
   getOllamaApiEndpoint,
+  getLMStudioApiEndpoint,
   getAnthropicApiKey,
   getGeminiApiKey,
   getOpenaiApiKey,
@@ -54,12 +55,13 @@ router.get('/', async (_, res) => {
 
     config['openaiApiKey'] = getOpenaiApiKey();
     config['ollamaApiUrl'] = getOllamaApiEndpoint();
+    config['lmStudioApiUrl'] = getLMStudioApiEndpoint();
     config['anthropicApiKey'] = getAnthropicApiKey();
     config['groqApiKey'] = getGroqApiKey();
     config['geminiApiKey'] = getGeminiApiKey();
     config['customOpenaiApiUrl'] = getCustomOpenaiApiUrl();
     config['customOpenaiApiKey'] = getCustomOpenaiApiKey();
-    config['customOpenaiModelName'] = getCustomOpenaiModelName();
+    config['customOpenaiModelName'] = getCustomOpenaiModelName()
 
     res.status(200).json(config);
   } catch (err: any) {
@@ -88,6 +90,9 @@ router.post('/', async (req, res) => {
       OLLAMA: {
         API_URL: config.ollamaApiUrl,
       },
+      LMSTUDIO: {
+        API_URL: config.lmStudioApiUrl,
+      },
       CUSTOM_OPENAI: {
         API_URL: config.customOpenaiApiUrl,
         API_KEY: config.customOpenaiApiKey,

src/routes/search.ts

@@ -85,10 +85,12 @@ router.post('/', async (req, res) => {
     if (body.chatModel?.provider === 'custom_openai') {
       llm = new ChatOpenAI({
         modelName: body.chatModel?.model || getCustomOpenaiModelName(),
-        openAIApiKey: body.chatModel?.customOpenAIKey || getCustomOpenaiApiKey(),
+        openAIApiKey:
+          body.chatModel?.customOpenAIKey || getCustomOpenaiApiKey(),
         temperature: 0.7,
         configuration: {
-          baseURL: body.chatModel?.customOpenAIBaseURL || getCustomOpenaiApiUrl(),
+          baseURL:
+            body.chatModel?.customOpenAIBaseURL || getCustomOpenaiApiUrl(),
         },
       }) as unknown as BaseChatModel;
     } else if (

ui/app/settings/page.tsx

@@ -223,11 +223,11 @@ const Page = () => {
         setChatModels(data.chatModelProviders || {});
         setEmbeddingModels(data.embeddingModelProviders || {});
 
-        const currentProvider = selectedChatModelProvider;
+        const currentChatProvider = selectedChatModelProvider;
-        const newProviders = Object.keys(data.chatModelProviders || {});
+        const newChatProviders = Object.keys(data.chatModelProviders || {});
 
-        if (!currentProvider && newProviders.length > 0) {
+        if (!currentChatProvider && newChatProviders.length > 0) {
-          const firstProvider = newProviders[0];
+          const firstProvider = newChatProviders[0];
           const firstModel = data.chatModelProviders[firstProvider]?.[0]?.name;
 
           if (firstModel) {
@@ -237,11 +237,11 @@ const Page = () => {
             localStorage.setItem('chatModel', firstModel);
           }
         } else if (
-          currentProvider &&
+          currentChatProvider &&
           (!data.chatModelProviders ||
-            !data.chatModelProviders[currentProvider] ||
+            !data.chatModelProviders[currentChatProvider] ||
-            !Array.isArray(data.chatModelProviders[currentProvider]) ||
+            !Array.isArray(data.chatModelProviders[currentChatProvider]) ||
-            data.chatModelProviders[currentProvider].length === 0)
+            data.chatModelProviders[currentChatProvider].length === 0)
         ) {
           const firstValidProvider = Object.entries(
             data.chatModelProviders || {},
@@ -267,6 +267,55 @@ const Page = () => {
           }
         }
 
+        const currentEmbeddingProvider = selectedEmbeddingModelProvider;
+        const newEmbeddingProviders = Object.keys(
+          data.embeddingModelProviders || {},
+        );
+
+        if (!currentEmbeddingProvider && newEmbeddingProviders.length > 0) {
+          const firstProvider = newEmbeddingProviders[0];
+          const firstModel =
+            data.embeddingModelProviders[firstProvider]?.[0]?.name;
+
+          if (firstModel) {
+            setSelectedEmbeddingModelProvider(firstProvider);
+            setSelectedEmbeddingModel(firstModel);
+            localStorage.setItem('embeddingModelProvider', firstProvider);
+            localStorage.setItem('embeddingModel', firstModel);
+          }
+        } else if (
+          currentEmbeddingProvider &&
+          (!data.embeddingModelProviders ||
+            !data.embeddingModelProviders[currentEmbeddingProvider] ||
+            !Array.isArray(
+              data.embeddingModelProviders[currentEmbeddingProvider],
+            ) ||
+            data.embeddingModelProviders[currentEmbeddingProvider].length === 0)
+        ) {
+          const firstValidProvider = Object.entries(
+            data.embeddingModelProviders || {},
+          ).find(
+            ([_, models]) => Array.isArray(models) && models.length > 0,
+          )?.[0];
+
+          if (firstValidProvider) {
+            setSelectedEmbeddingModelProvider(firstValidProvider);
+            setSelectedEmbeddingModel(
+              data.embeddingModelProviders[firstValidProvider][0].name,
+            );
+            localStorage.setItem('embeddingModelProvider', firstValidProvider);
+            localStorage.setItem(
+              'embeddingModel',
+              data.embeddingModelProviders[firstValidProvider][0].name,
+            );
+          } else {
+            setSelectedEmbeddingModelProvider(null);
+            setSelectedEmbeddingModel(null);
+            localStorage.removeItem('embeddingModelProvider');
+            localStorage.removeItem('embeddingModel');
+          }
+        }
+
         setConfig(data);
       }
 
@@ -278,6 +327,10 @@ const Page = () => {
         localStorage.setItem('chatModelProvider', value);
       } else if (key === 'chatModel') {
         localStorage.setItem('chatModel', value);
+      } else if (key === 'embeddingModelProvider') {
+        localStorage.setItem('embeddingModelProvider', value);
+      } else if (key === 'embeddingModel') {
+        localStorage.setItem('embeddingModel', value);
       }
     } catch (err) {
       console.error('Failed to save:', err);
@@ -436,7 +489,6 @@ const Page = () => {
                         const value = e.target.value;
                         setSelectedChatModelProvider(value);
                         saveConfig('chatModelProvider', value);
-                        // Auto-select first model of new provider
                         const firstModel =
                           config.chatModelProviders[value]?.[0]?.name;
                         if (firstModel) {
@@ -554,6 +606,81 @@ const Page = () => {
                     </div>
                   </div>
                 )}
+
+              {config.embeddingModelProviders && (
+                <div className="flex flex-col space-y-4 mt-4 pt-4 border-t border-light-200 dark:border-dark-200">
+                  <div className="flex flex-col space-y-1">
+                    <p className="text-black/70 dark:text-white/70 text-sm">
+                      Embedding Model Provider
+                    </p>
+                    <Select
+                      value={selectedEmbeddingModelProvider ?? undefined}
+                      onChange={(e) => {
+                        const value = e.target.value;
+                        setSelectedEmbeddingModelProvider(value);
+                        saveConfig('embeddingModelProvider', value);
+                        const firstModel =
+                          config.embeddingModelProviders[value]?.[0]?.name;
+                        if (firstModel) {
+                          setSelectedEmbeddingModel(firstModel);
+                          saveConfig('embeddingModel', firstModel);
+                        }
+                      }}
+                      options={Object.keys(config.embeddingModelProviders).map(
+                        (provider) => ({
+                          value: provider,
+                          label:
+                            provider.charAt(0).toUpperCase() +
+                            provider.slice(1),
+                        }),
+                      )}
+                    />
+                  </div>
+
+                  {selectedEmbeddingModelProvider && (
+                    <div className="flex flex-col space-y-1">
+                      <p className="text-black/70 dark:text-white/70 text-sm">
+                        Embedding Model
+                      </p>
+                      <Select
+                        value={selectedEmbeddingModel ?? undefined}
+                        onChange={(e) => {
+                          const value = e.target.value;
+                          setSelectedEmbeddingModel(value);
+                          saveConfig('embeddingModel', value);
+                        }}
+                        options={(() => {
+                          const embeddingModelProvider =
+                            config.embeddingModelProviders[
+                              selectedEmbeddingModelProvider
+                            ];
+                          return embeddingModelProvider
+                            ? embeddingModelProvider.length > 0
+                              ? embeddingModelProvider.map((model) => ({
+                                  value: model.name,
+                                  label: model.displayName,
+                                }))
+                              : [
+                                  {
+                                    value: '',
+                                    label: 'No models available',
+                                    disabled: true,
+                                  },
+                                ]
+                            : [
+                                {
+                                  value: '',
+                                  label:
+                                    'Invalid provider, please check backend logs',
+                                  disabled: true,
+                                },
+                              ];
+                        })()}
+                      />
+                    </div>
+                  )}
+                </div>
+              )}
             </SettingsSection>
 
             <SettingsSection title="API Keys">

ui/components/MessageBox.tsx

@@ -11,6 +11,8 @@ import {
   StopCircle,
   Layers3,
   Plus,
+  Brain,
+  ChevronDown,
 } from 'lucide-react';
 import Markdown from 'markdown-to-jsx';
 import Copy from './MessageActions/Copy';
@@ -41,26 +43,58 @@ const MessageBox = ({
 }) => {
   const [parsedMessage, setParsedMessage] = useState(message.content);
   const [speechMessage, setSpeechMessage] = useState(message.content);
+  const [thinking, setThinking] = useState<string>('');
+  const [answer, setAnswer] = useState<string>('');
+  const [isThinkingExpanded, setIsThinkingExpanded] = useState(false);
 
   useEffect(() => {
     const regex = /\[(\d+)\]/g;
+    const thinkRegex = /<think>(.*?)(?:<\/think>|$)(.*)/s;
 
-    if (
+    // Check for thinking content, including partial tags
-      message.role === 'assistant' &&
+    const match = message.content.match(thinkRegex);
-      message?.sources &&
+    if (match) {
-      message.sources.length > 0
+      const [_, thinkingContent, answerContent] = match;
-    ) {
+      
-      return setParsedMessage(
+      // Set thinking content even if </think> hasn't appeared yet
+      if (thinkingContent) {
+        setThinking(thinkingContent.trim());
+        setIsThinkingExpanded(true); // Auto-expand when thinking starts
+      }
+
+      // Only set answer content if we have it (after </think>)
+      if (answerContent) {
+        setAnswer(answerContent.trim());
+        
+        // Process the answer part for sources if needed
+        if (message.role === 'assistant' && message?.sources && message.sources.length > 0) {
+          setParsedMessage(
+            answerContent.trim().replace(
+              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>`,
+            ),
+          );
+        } else {
+          setParsedMessage(answerContent.trim());
+        }
+        setSpeechMessage(answerContent.trim().replace(regex, ''));
+      }
+    } else {
+      // No thinking content - process as before
+      if (message.role === 'assistant' && message?.sources && message.sources.length > 0) {
+        setParsedMessage(
           message.content.replace(
             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>`,
           ),
         );
-    }
+      } else {
-
-    setSpeechMessage(message.content.replace(regex, ''));
         setParsedMessage(message.content);
+      }
+      setSpeechMessage(message.content.replace(regex, ''));
+    }
   }, [message.content, message.sources, message.role]);
 
   const { speechStatus, start, stop } = useSpeech({ text: speechMessage });
@@ -92,6 +126,49 @@ const MessageBox = ({
                 <MessageSources sources={message.sources} />
               </div>
             )}
+            <div className="flex flex-col space-y-4">
+              {thinking && (
+                <div className="flex flex-col space-y-2 mb-4">
+                  <button 
+                    onClick={() => setIsThinkingExpanded(!isThinkingExpanded)}
+                    className="flex flex-row items-center space-x-2 group text-black/70 dark:text-white/70 hover:text-black dark:hover:text-white transition duration-200"
+                  >
+                    <Brain size={20} />
+                    <h3 className="font-medium text-xl">Reasoning</h3>
+                    <ChevronDown 
+                      size={16}
+                      className={cn(
+                        "transition-transform duration-200",
+                        isThinkingExpanded ? "rotate-180" : ""
+                      )} 
+                    />
+                  </button>
+                  
+                  {isThinkingExpanded && (
+                    <div className="rounded-lg bg-light-secondary/50 dark:bg-dark-secondary/50 p-4">
+                      {thinking.split('\n\n').map((paragraph, index) => {
+                        if (!paragraph.trim()) return null;
+                        
+                        const content = paragraph.replace(/^[•\-\d.]\s*/, '');
+                        
+                        return (
+                          <div key={index} className="mb-2 last:mb-0">
+                            <details className="group [&_summary::-webkit-details-marker]:hidden">
+                              <summary className="flex items-center cursor-pointer list-none text-sm text-black/70 dark:text-white/70 hover:text-black dark:hover:text-white">
+                                <span className="arrow mr-2 inline-block transition-transform duration-200 group-open:rotate-90 group-open:self-start group-open:mt-1">▸</span>
+                                <p className="relative whitespace-normal line-clamp-1 group-open:line-clamp-none after:content-['...'] after:inline group-open:after:hidden transition-all duration-200 text-ellipsis overflow-hidden group-open:overflow-visible">
+                                  {content}
+                                </p>
+                              </summary>
+                            </details>
+                          </div>
+                        );
+                      })}
+                    </div>
+                  )}
+                </div>
+              )}
+
               <div className="flex flex-col space-y-2">
                 <div className="flex flex-row items-center space-x-2">
                   <Disc3
@@ -113,6 +190,7 @@ const MessageBox = ({
                 >
                   {parsedMessage}
                 </Markdown>
+              </div>
               {loading && isLast ? null : (
                 <div className="flex flex-row items-center justify-between w-full text-black dark:text-white py-4 -mx-2">
                   <div className="flex flex-row items-center space-x-1">

ui/tsconfig.json

@@ -1,5 +1,6 @@
 {
   "compilerOptions": {
+    "target": "es2018",
     "lib": ["dom", "dom.iterable", "esnext"],
     "allowJs": true,
     "skipLibCheck": true,