Ask Your Codebase CLI

An agentic CLI tool that lets a developer ask natural-language questions about their codebase and have an AI agent search, explain, edit, and fix code — using RAG for retrieval and LangGraph for reasoning.

Tech Stack

• Language: Python 3.11+
• Agent Framework: LangGraph + LangChain
• Embeddings: sentence-transformers (all-MiniLM-L6-v2 runs locally)
• Vector Database: ChromaDB (local persistence)
• LLM Providers: Groq (default), Google Gemini, and Ollama (local)
• CLI Framework: Typer

---

Installation

Install in editable mode:

pip install -e .

To run the agent, export the environment variable for your LLM provider:

# For Groq
export GROQ_API_KEY="your-groq-api-key"

# For Google Gemini
export GEMINI_API_KEY="your-gemini-api-key"

---

Configuration

You can customize indexing and LLM settings by creating a .askcodebase.yml file in the root of your project:

ignore_patterns:
  - "node_modules/**"
  - ".git/**"
  - "venv/**"
  - ".venv/**"
  - "dist/**"

chunk:
  size: 1000
  overlap: 200

agent:
  max_iterations: 5
  auto_apply: false

llm:
  provider: "groq"
  model: "llama-3.1-8b-instant"
  temperature: 0.0

---

CLI Commands

1. Index Codebase

Recursively scans the directory, respects .gitignore, and generates embeddings:

ask-codebase index

To index a specific folder path:

ask-codebase index /path/to/project

2. Ask a Question

Get answers with precise file and line citations:

ask-codebase ask "How does the authentication middleware work?"

3. Automatically Fix Bugs

Ask the agent to locate a bug, modify code, and verify it with the project's tests:

ask-codebase fix "Fix the division-by-zero issue in calculator.py"

4. Interactive Chat (REPL)

Run a persistent, multi-turn conversation session:

ask-codebase chat

Use the --verbose or -v flag on any command to inspect the agent's full plan, tool calls, and observations.