feat: Add LM Studio provider support for local model hosting

README-LMSTUDIO.md

@@ -0,0 +1,219 @@
+# LM Studio Integration with Claude Code Router
+
+This guide explains how to integrate Claude Code Router (CCR) with LM Studio to use local models with Claude Code.
+
+## Prerequisites
+
+1. [LM Studio](https://lmstudio.ai/) installed and running
+2. [Claude Code](https://docs.anthropic.com/en/docs/claude-code/quickstart) installed
+3. Claude Code Router installed: `npm install -g @musistudio/claude-code-router`
+
+## Quick Setup
+
+1. **Start LM Studio**:
+   - Open LM Studio
+   - Load a model (e.g., `google/gemma-3n-e4b`)
+   - Start the local server (typically on `http://localhost:1234`)
+
+2. **Configure CCR for LM Studio**:
+   ```bash
+   ./setup-lmstudio.sh
+   ```
+
+3. **Start CCR**:
+   ```bash
+   ccr start
+   ```
+
+4. **Use Claude Code with LM Studio**:
+   ```bash
+   ccr code "What is the capital of France?"
+   ```
+
+## Manual Configuration
+
+If you prefer to configure manually, create `~/.claude-code-router/config.json`:
+
+```json
+{
+  "Providers": [
+    {
+      "name": "lmstudio",
+      "api_base_url": "http://localhost:1234/v1/chat/completions",
+      "api_key": "lm-studio",
+      "models": ["your-model-name"]
+    }
+  ],
+  "Router": {
+    "default": "lmstudio,your-model-name",
+    "background": "lmstudio,your-model-name",
+    "think": "lmstudio,your-model-name",
+    "longContext": "lmstudio,your-model-name",
+    "longContextThreshold": 60000,
+    "webSearch": "lmstudio,your-model-name"
+  },
+  "API_TIMEOUT_MS": 600000,
+  "LOG": true
+}
+```
+
+## Configuration Options
+
+### Model Configuration
+Update the `models` array with your loaded LM Studio model:
+```json
+"models": ["microsoft/DialoGPT-medium", "meta-llama/Llama-2-7b-chat-hf"]
+```
+
+### Custom API Endpoint
+If LM Studio runs on a different port:
+```json
+"api_base_url": "http://localhost:8080/v1/chat/completions"
+```
+
+### Multiple Models Setup
+Configure different models for different tasks:
+```json
+{
+  "Providers": [
+    {
+      "name": "lmstudio",
+      "api_base_url": "http://localhost:1234/v1/chat/completions",
+      "api_key": "lm-studio",
+      "models": ["google/gemma-3n-e4b", "microsoft/DialoGPT-medium"]
+    }
+  ],
+  "Router": {
+    "default": "lmstudio,google/gemma-3n-e4b",
+    "background": "lmstudio,microsoft/DialoGPT-medium",
+    "think": "lmstudio,google/gemma-3n-e4b"
+  }
+}
+```
+
+## Advanced Configuration
+
+### Adding Transformers
+While LM Studio doesn't typically need transformers (it's OpenAI-compatible), you can add them for custom behavior:
+
+```json
+{
+  "name": "lmstudio",
+  "api_base_url": "http://localhost:1234/v1/chat/completions",
+  "api_key": "lm-studio",
+  "models": ["google/gemma-3n-e4b"],
+  "transformer": {
+    "use": [
+      [
+        "maxtoken",
+        {
+          "max_tokens": 4096
+        }
+      ]
+    ]
+  }
+}
+```
+
+### Custom Router
+Create a custom router for LM Studio specific logic:
+
+```javascript
+// ~/.claude-code-router/custom-lmstudio-router.js
+module.exports = async function router(req, config) {
+  const userMessage = req.body.messages.find((m) => m.role === "user")?.content;
+
+  // Use a specific model for code-related queries
+  if (userMessage && userMessage.toLowerCase().includes("code")) {
+    return "lmstudio,google/gemma-3n-e4b";
+  }
+
+  // Default to configured router
+  return null;
+};
+```
+
+Then in your config:
+```json
+{
+  "CUSTOM_ROUTER_PATH": "$HOME/.claude-code-router/custom-lmstudio-router.js"
+}
+```
+
+## Testing Your Setup
+
+1. **Test basic functionality**:
+   ```bash
+   ccr code "Hello, how are you?"
+   ```
+
+2. **Test with longer context**:
+   ```bash
+   ccr code "Explain the concept of machine learning in detail"
+   ```
+
+3. **Check routing**:
+   ```bash
+   ccr status
+   ```
+
+## Troubleshooting
+
+### Common Issues
+
+1. **Connection refused**: Make sure LM Studio server is running
+2. **Model not found**: Ensure the model name in config matches the loaded model in LM Studio
+3. **Timeout errors**: Increase `API_TIMEOUT_MS` in config for slower local models
+
+### Debug Mode
+Enable logging for troubleshooting:
+```json
+{
+  "LOG": true
+}
+```
+
+View logs:
+```bash
+tail -f ~/.claude-code-router.log
+```
+
+### Testing LM Studio API Directly
+Test the LM Studio API directly with curl (same as your original command):
+
+```bash
+curl http://localhost:1234/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "google/gemma-3n-e4b",
+    "messages": [
+      { "role": "system", "content": "Always answer in rhymes. Today is Thursday" },
+      { "role": "user", "content": "What day is it today?" }
+    ],
+    "temperature": 0.7,
+    "max_tokens": -1,
+    "stream": false
+}'
+```
+
+## Benefits of Using CCR with LM Studio
+
+1. **Unified Interface**: Use Claude Code with local models
+2. **Routing Logic**: Automatically route different types of requests to appropriate models
+3. **Easy Model Switching**: Switch between local and cloud models seamlessly
+4. **Cost Effective**: Use local models for development and testing
+5. **Privacy**: Keep sensitive code on your local machine
+
+## Example Workflows
+
+### Development Workflow
+```bash
+# Use local model for quick iterations
+ccr code "Review this Python function for bugs"
+
+# Switch to cloud model for complex tasks
+/model openrouter,anthropic/claude-3.5-sonnet
+```
+
+### Mixed Model Setup
+Configure different models for different scenarios in your config.json and let CCR automatically route requests based on context, token count, and task type.

README.md

@@ -9,7 +9,7 @@
 ## ✨ Features
 
 - **Model Routing**: Route requests to different models based on your needs (e.g., background tasks, thinking, long context).
-- **Multi-Provider Support**: Supports various model providers like OpenRouter, DeepSeek, Ollama, Gemini, Volcengine, and SiliconFlow.
+- **Multi-Provider Support**: Supports various model providers like OpenRouter, DeepSeek, Ollama, Gemini, Volcengine, SiliconFlow, and LM Studio.
 - **Request/Response Transformation**: Customize requests and responses for different providers using transformers.
 - **Dynamic Model Switching**: Switch models on-the-fly within Claude Code using the `/model` command.
 - **GitHub Actions Integration**: Trigger Claude Code tasks in your GitHub workflows.
@@ -151,6 +151,12 @@ Here is a comprehensive example:
         "claude-opus-4-20250514",
         "gemini-2.5-pro"
       ]
+    },
+    {
+      "name": "lmstudio",
+      "api_base_url": "http://localhost:1234/v1/chat/completions",
+      "api_key": "lm-studio",
+      "models": ["google/gemma-3n-e4b"]
     }
   ],
   "Router": {

config.example.json

@@ -102,11 +102,17 @@
           "enhancetool"
         ]
       }
+    },
+    {
+      "name": "lmstudio",
+      "api_base_url": "http://localhost:1234/v1/chat/completions",
+      "api_key": "lm-studio",
+      "models": ["google/gemma-3n-e4b"]
     }
   ],
   "Router": {
-    "default": "deepseek,deepseek-chat",
-    "background": "ollama,qwen2.5-coder:latest",
+    "default": "lmstudio,google/gemma-3n-e4b",
+    "background": "lmstudio,google/gemma-3n-e4b",
     "think": "deepseek,deepseek-reasoner",
     "longContext": "openrouter,google/gemini-2.5-pro-preview",
     "longContextThreshold": 60000,

config.lmstudio.json

@@ -0,0 +1,20 @@
+{
+  "Providers": [
+    {
+      "name": "lmstudio",
+      "api_base_url": "http://localhost:1234/v1/chat/completions",
+      "api_key": "lm-studio",
+      "models": ["google/gemma-3n-e4b"]
+    }
+  ],
+  "Router": {
+    "default": "lmstudio,google/gemma-3n-e4b",
+    "background": "lmstudio,google/gemma-3n-e4b",
+    "think": "lmstudio,google/gemma-3n-e4b",
+    "longContext": "lmstudio,google/gemma-3n-e4b",
+    "longContextThreshold": 60000,
+    "webSearch": "lmstudio,google/gemma-3n-e4b"
+  },
+  "API_TIMEOUT_MS": 600000,
+  "LOG": true
+}

setup-lmstudio.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# Setup script for integrating Claude Code Router with LM Studio
+echo "Setting up Claude Code Router for LM Studio..."
+
+# Create .claude-code-router directory if it doesn't exist
+mkdir -p "$HOME/.claude-code-router"
+
+# Copy the LM Studio configuration
+echo "Copying LM Studio configuration..."
+cp config.lmstudio.json "$HOME/.claude-code-router/config.json"
+
+echo "Configuration copied to $HOME/.claude-code-router/config.json"
+echo ""
+echo "To use CCR with LM Studio:"
+echo "1. Make sure LM Studio is running on http://localhost:1234"
+echo "2. Load the model 'google/gemma-3n-e4b' in LM Studio"
+echo "3. Run: ccr start"
+echo "4. Run: ccr code 'your prompt here'"
+echo ""
+echo "You can also edit the configuration at:"
+echo "$HOME/.claude-code-router/config.json"
+echo ""
+echo "To change the model, update the 'models' array in the lmstudio provider"
+echo "and update the Router section accordingly."