Coding Assistant

Coding Assistant

Your Local AI Pair Programmer: Point this AI assistant to your project folder and start asking questions!

It reads your code locally to provide truly context-aware help, explanations, and suggestions directly related to your files. Code smarter, locally.

Prerequisites 📝

• Python: Version 3.11 or higher recommended.
• tree command: The get_tree_folders tool relies on this.
• Linux (Debian/Ubuntu): sudo apt update && sudo apt install tree
• macOS (using Homebrew): brew install tree

Installation ⚙️

1. Clone the repository:

``bash git clone https://github.com/Bessouat40/coding-assistant cd coding-assistant ``

1. Install Python dependencies:

``bash pip install -r requirements.txt ``

Configuration 🔑

1. API Keys & Settings: Sensitive information like LLM API keys should be stored in a .env file in the project's root directory. - Copy the example file or create a new file named .env:

``bash cp .env.example .env ``

• Edit the .env file and add your keys:

``dotenv # .env MISTRAL_API_KEY=your_mistral_api_key_here # GOOGLE_API_KEY=your_google_api_key_here # Add other variables if needed, e.g., MAX_CONTEXT_TOKENS=7000 ``

1. LLM Provider: Change LLM provider if you want :

• Go to api/utils.py and modify loadl_llm function. Uncomment the line with the provider you need.

By default, it's set to Google :

def load_llm(logger):
    try:
        model = ChatOllama(model="llama3.1:8b")
        # model = ChatMistralAI(model="codestral-latest")
        # model = ChatGoogleGenerativeAI(model="gemini-2.0-flash-001")
        logger.info(f"ChatGoogleGenerativeAI model '{model.model}' initialized successfully.")
        return model
    except Exception as e:
        logger.error(f"Failed to initialize the LLM model: {e}")
        raise RuntimeError(f"Could not initialize LLM: {e}") from e

Running the Application ▶️

A convenience script launch_assistant.sh is provided to start all components.

1. Make the script executable:

``bash chmod +x launch_assistant.sh ``

1. Run the script:

``bash ./launch_assistant.sh ``

This script will:

• Load environment variables from .env.
• Start the MCP Tool Server (agent_tools.py) in the background (logs to mcp_server.log).
• Start the FastAPI Backend (api.py) in the background (logs to fastapi_api.log).
• Start the Streamlit UI (streamlit_app.py) in the background (logs to streamlit_ui.log).
• Print the URLs and PIDs for each component.

You can view logs in the specified .log files for debugging. Press Ctrl+C in the terminal where you ran the script to stop all components gracefully.

Usage 🧑‍💻

1. Access the UI: Open your web browser and navigate to the Streamlit URL (usually http://localhost:8501).
2. Set Project Directory: In the sidebar, enter the absolute path to the local project directory you want the assistant to work with. Click "Set Directory".
3. Chat: Use the chat input at the bottom to ask questions about the code in the specified directory. Examples:

• "Explain the purpose of the run_agent function in api.py."
• "Show me the file structure of this project." (Uses tree)
• "What arguments does the generate_response_api function take?"
• "Read the contents of prompt.py." (Uses cat)
• "Can you suggest improvements to the error handling in streamlit_app.py?"

1. Follow-up: The assistant remembers the context of your current chat session. Ask follow-up questions naturally.
2. Clear History: Use the "Clear Chat History" button in the sidebar to start a fresh conversation.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Bessouat40
• Source: Bessouat40/coding-assistant
• License: MIT

Install and usage instructions live in the source repository linked above.