code-search-chat

Code Search Chat Interface

A web interface for the Probe code search tool, powered by AI (Anthropic Claude or OpenAI GPT).

Features

• Interactive chat interface for code search
• Support for both Anthropic Claude and OpenAI GPT models
• Multiple search tools: search, query, and extract
• Markdown rendering with syntax highlighting
• Mermaid diagram support
• Streaming responses for real-time feedback
• API endpoints for programmatic access
• Optional basic authentication
• Uses the @buger/probe npm package

Setup

Environment Variables

Create a .env file in the web directory with the following variables:

# Required: At least one of these API keys must be provided
ANTHROPIC_API_KEY=your_anthropic_api_key
OPENAI_API_KEY=your_openai_api_key

# Optional: Override the default model
MODEL_NAME=claude-3-7-sonnet-latest  # or gpt-4o, etc.

# Optional: Override the default API URLs
ANTHROPIC_API_URL=https://api.anthropic.com
OPENAI_API_URL=https://api.openai.com/v1

# Optional: Configure the port (default: 3000)
PORT=8080

# Optional: Enable debug mode
DEBUG=true

# Optional: Enable raw request debugging to see the exact prompts sent to LLMs
DEBUG_RAW_REQUEST=true

# You can combine both for maximum debugging information
DEBUG=true DEBUG_RAW_REQUEST=true

# Required: Configure folders to search
ALLOWED_FOLDERS=/path/to/repo1,/path/to/repo2

# Optional: Authentication settings
AUTH_ENABLED=true  # Set to true to enable authentication
AUTH_USERNAME=admin  # Custom username (default: admin)
AUTH_PASSWORD=secure_password  # Custom password (default: password)

Running Locally

• Install dependencies:

  npm install
  

• Start the server:

  npm start
  

• Open your browser and navigate to http://localhost:8080 (or whatever port you configured)

Docker

Building the Docker Image

docker build -t code-search-chat .

Running with Docker

docker run -p 8080:8080 \
  -e ANTHROPIC_API_KEY=your_anthropic_api_key \
  -e ALLOWED_FOLDERS=/app/code1,/app/code2 \
  -v /path/to/local/code1:/app/code1 \
  -v /path/to/local/code2:/app/code2 \
  code-search-chat

Or with OpenAI and authentication:

docker run -p 8080:8080 \
  -e OPENAI_API_KEY=your_openai_api_key \
  -e MODEL_NAME=gpt-4o \
  -e ALLOWED_FOLDERS=/app/code1,/app/code2 \
  -e AUTH_ENABLED=true \
  -e AUTH_USERNAME=admin \
  -e AUTH_PASSWORD=secure_password \
  -v /path/to/local/code1:/app/code1 \
  -v /path/to/local/code2:/app/code2 \
  code-search-chat

Environment Variables in Docker

All the environment variables mentioned in the Setup section can be passed to the Docker container using the -e flag.

API Documentation

The application provides a full OpenAPI specification at /openapi.yaml. You can use this specification with tools like Swagger UI or Postman to explore and test the API.

API Endpoints

The application provides the following API endpoints:

1. Search Code (POST /api/search)

Search code repositories using the Probe tool.

Request:

{
  "keywords": "search pattern",
  "folder": "/path/to/repo",
  "exact": false,
  "allow_tests": false
}

Parameters:

• keywords (required): Search pattern
• folder (optional): Path to search in (must be one of the allowed folders)
• exact (optional): Use exact match (default: false)
• allow_tests (optional): Include test files in results (default: false)

Response:

{
  "results": "search results text",
  "command": "probe command that was executed",
  "timestamp": "2025-08-03T07:10:00.000Z"
}

2. Query Code (POST /api/query)

Search code using ast-grep structural pattern matching.

Request:

{
  "pattern": "function $NAME($$$PARAMS) { $$$BODY }",
  "path": "/path/to/repo",
  "language": "javascript",
  "allow_tests": false
}

Parameters:

• pattern (required): AST pattern to search for
• path (optional): Path to search in (must be one of the allowed folders)
• language (optional): Programming language to use for parsing
• allow_tests (optional): Include test files in results (default: false)

Response:

{
  "results": "query results text",
  "timestamp": "2025-08-03T07:10:00.000Z"
}

3. Extract Code (POST /api/extract)

Extract code blocks from files based on file paths and optional line numbers.

Request:

{
  "file_path": "src/main.js:42",
  "line": 42,
  "end_line": 60,
  "allow_tests": false,
  "context_lines": 10,
  "format": "plain"
}

Parameters:

• file_path (required): Path to the file to extract from
• line (optional): Start line number
• end_line (optional): End line number
• allow_tests (optional): Allow test files (default: false)
• context_lines (optional): Number of context lines (default: 10)
• format (optional): Output format (default: "plain")

Response:

{
  "results": "extracted code text",
  "timestamp": "2025-08-03T07:10:00.000Z"
}

4. Chat with AI (POST /api/chat)

Send a message to the AI and get a response.

Request:

{
  "message": "your question about the code",
  "stream": true
}

Parameters:

• message (required): The message to send to the AI
• stream (optional): Whether to stream the response (default: true)

Response (stream=false):

{
  "response": "AI response text",
  "toolCalls": [
    {
      "name": "searchCode",
      "arguments": {
        "keywords": "search pattern",
        "folder": "/path/to/repo"
      },
      "result": "search results"
    }
  ],
  "timestamp": "2025-08-03T07:10:00.000Z"
}

Response (stream=true): Text stream of the AI response.

Authentication

When authentication is enabled (AUTH_ENABLED=true), all endpoints (both UI and API) require basic authentication. The default username is admin and the default password is password, but these can be customized using the AUTH_USERNAME and AUTH_PASSWORD environment variables.

To authenticate API requests, include the Authorization header with the value Basic <base64-encoded-credentials>, where <base64-encoded-credentials> is the Base64 encoding of username:password.

Example:

Authorization: Basic YWRtaW46cGFzc3dvcmQ=

API Model Support

The application will use the first available API in this order:

• Anthropic Claude (if ANTHROPIC_API_KEY is provided)
• OpenAI GPT (if OPENAI_API_KEY is provided)

You can override the default model by setting the MODEL_NAME environment variable.

Default models:

• Anthropic: claude-3-7-sonnet-latest
• OpenAI: gpt-4o

Custom API URLs

If you're using a proxy or a custom endpoint for the APIs, you can override the default URLs:

• ANTHROPIC_API_URL: Default is https://api.anthropic.com
• OPENAI_API_URL: Default is https://api.openai.com/v1

Dependencies

This application uses the following key dependencies:

• @buger/probe: Node.js wrapper for the probe tool
• Vercel AI SDK: The AI SDK used for model integration
• Marked.js: For Markdown rendering
• Highlight.js: For syntax highlighting
• Mermaid.js: For diagram rendering