feat: Add LM Studio Support and Thinking Model Panel

LM Studio Integration:
- Added LM Studio provider with OpenAI-compatible API support
- Dynamic model discovery via /v1/models endpoint
- Support for both chat and embeddings models
- Docker-compatible networking configuration

Thinking Model Panel:
- Added collapsible UI panel for model's chain of thought
- Parses responses with <think> tags to separate reasoning
- Maintains backward compatibility with regular responses
- Styled consistently with app theme for light/dark modes
- Preserves all existing message functionality (sources, markdown, etc.)

These improvements enhance the app's compatibility with local LLMs and
provide better visibility into model reasoning processes while maintaining
existing functionality.

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 +1,89 @@
+import { OpenAIEmbeddings } from '@langchain/openai';
+import { ChatOpenAI } from '@langchain/openai';
+import { getKeepAlive, getLMStudioApiEndpoint } from '../../config';
+import logger from '../../utils/logger';
+import axios from 'axios';
+
+interface LMStudioModel {
+  id: string;
+  // add other properties if LM Studio API provides them
+}
+
+interface ChatModelConfig {
+  displayName: string;
+  model: ChatOpenAI;
+}
+
+export const loadLMStudioChatModels = async (): Promise<Record<string, ChatModelConfig>> => {
+  const lmStudioEndpoint = getLMStudioApiEndpoint();
+
+  if (!lmStudioEndpoint) {
+    logger.debug('LM Studio endpoint not configured, skipping');
+    return {};
+  }
+
+  try {
+    const response = await axios.get<{ data: LMStudioModel[] }>(`${lmStudioEndpoint}/models`, {
+      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: lmStudioEndpoint,
+          },
+          modelName: model.id,
+          temperature: 0.7,
+        }),
+      };
+      return acc;
+    }, {});
+
+    return chatModels;
+  } catch (err) {
+    logger.error(`Error loading LM Studio models: ${err}`);
+    return {};
+  }
+};
+
+export const loadLMStudioEmbeddingsModels = async () => {
+  const lmStudioEndpoint = getLMStudioApiEndpoint();
+
+  if (!lmStudioEndpoint) return {};
+
+  try {
+    const response = await axios.get(`${lmStudioEndpoint}/models`, {
+      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: lmStudioEndpoint,
+          },
+          modelName: model.id,
+        }),
+      };
+      return acc;
+    }, {});
+
+    return embeddingsModels;
+  } catch (err) {
+    logger.error(`Error loading LM Studio embeddings model: ${err}`);
+    return {};
+  }
+};