agent-coderag 1.2.0

Lightweight semantic code search and distillation utility for AI coding agents. It solves the API knowledge gap via real-time local signature extraction and intent analysis without PyTorch. Optimized for token efficiency, it compresses codebase context into compact semantic summaries stored in a local DuckDB vector similarity index.

Agent-CodeRAG

Fast. Local. Agent-First. Token-Efficient. Bridge the gap between AI coding agents and your local environment.

---

Table of Contents

• Why Agent-CodeRAG?
• Quick Start
• Features
• How It Works
• API Discovery
• For AI Agents
• Development
• Contributing
• License

---

Why Agent-CodeRAG?

AI coding agents often hallucinate when calling library APIs because their training data is static. This leads to a "Fail-Fix-Fail" cycle — broken code, token waste, and frustration.

The Problem: Your agent knows Pydantic v1 (model.dict()), but your environment uses v2 (model.model_dump()). Result: 5000+ wasted tokens trying to "fix" something it doesn't understand.

The Solution: Agent-CodeRAG extracts actual API signatures from your installed libraries and provides the LLM with real-time, environment-specific context — saving up to 80% of context window tokens.

🔝 Back to top

---

Quick Start

# 1. Install
pip install agent-coderag

# 2. Setup (download ONNX model)
agent-coderag setup

# 3. Configure LLM (optional, for AI distillation)
agent-coderag config --url "http://localhost:11434" --provider "ollama" --model "qwen2.5-coder:7b"

# 4. Index your project
agent-coderag sync --all

# 5. Search
agent-coderag search "how to handle errors"

Docker:

docker build -t agent-coderag .
docker run -v ~/.cache/agent-coderag:/root/.cache/agent-coderag agent-coderag setup

🔝 Back to top

---

✨ Features

• ⚡ No PyTorch — Uses onnxruntime and tokenizers (Rust) for instant startup
• 💾 DuckDB VSS — High-performance vector search in a single local file
• 🏗️ Multi-Language — Native indexing for Python (AST) and Java (javalang)
• 🔄 Delta-Sync — SHA-256 hashing re-distills only changed code
• 🔌 Hybrid Intelligence — Works offline; adds AI-distilled reasoning when LLM is connected
• 📡 API Discovery — Extract live API signatures from Python modules or Java JARs

🔝 Back to top

---

🛠 How It Works

graph TD
    A[Local Python Code] --> B[AST Parser]
    B --> C{Delta-Sync}
    C -- Changed/New --> D[LLM Distiller]
    C -- Unchanged --> E[Local Cache]
    D --> F[Semantic Summary]
    E --> F
    F --> G[ONNX Embedder]
    G --> H[(DuckDB VSS)]
    H --> I[Semantic Search / JSON API]

1. AST Parser — Parses your Python code
2. Delta-Sync — Uses SHA-256 to detect changes
3. LLM Distiller — Generates semantic summaries (optional)
4. ONNX Embedder — Creates embeddings locally
5. DuckDB VSS — Stores vectors for fast similarity search

🔝 Back to top

---

📡 API Discovery

agent-coderag api pydantic

Returns the live public API, methods, and signatures for any installed library.

🔝 Back to top

---

🤖 For AI Agents

Agent-CodeRAG is built for programmatic consumption:

1. Search First: agent-coderag --json search "topic" --limit 1
2. Use Intent: The summary field provides technical intent — skip reading unnecessary files

🔝 Back to top

---

🔧 Development

# Run tests
pytest tests/
pytest e2e_tests/

# Setup pre-commit hooks
pip install pre-commit
pre-commit install

🔝 Back to top

---

🤝 Contributing

Contributions are welcome! Here's how to get started:

1. Fork the repository
2. Clone your fork: git clone https://github.com/YOUR_USERNAME/agent-coderag.git
3. Create a branch: git checkout -b feature/your-feature
4. Make changes and commit with Conventional Commits
5. Run tests: pytest tests/
6. Push to your fork and create a Pull Request

See CONTRIBUTING.md for detailed guidelines.

🔝 Back to top

---

📄 License

MIT © 2026 Igor Boloban

🔝 Back to top