Sage is an open-source tool designed to help developers understand and integrate with any codebase quickly, functioning as an alternative to GitHub Copilot. It targets developers who need to learn new codebases efficiently, offering a significant time-saving benefit by reducing manual code exploration.

How It Works

Sage employs a retrieval-augmented generation (RAG) approach, allowing users to chat with their codebase. It supports flexible retrieval mechanisms, ranging from lightweight strategies requiring only an LLM API key to more traditional RAG setups that involve indexing the codebase. This modular design allows for pluggable embedding models, LLM providers (local via Ollama or cloud APIs like OpenAI/Anthropic), and vector stores (like Marqo).

Quick Start & Requirements

• Install: Follow the quickstart guide.
• Prerequisites: Python, Ollama (for local LLMs), Marqo (for local vector store), or API keys for third-party LLM providers.
• Resources: Local setup with Ollama and Marqo requires sufficient disk space for indexed codebases and LLM models.
• Links: Hosted app, Documentation

Highlighted Details

• Chat with any codebase in under two minutes.
• Fully local or cloud-based operation for privacy and quality optimization.
• Modular architecture for easy integration of custom LLMs, embedding models, and vector stores.
• Well-documented experiments profiling various retrieval strategies.

Maintenance & Community

The project is actively seeking contributions and feature requests via email or pull requests. They offer to host OSS repositories for free on their platform.

Licensing & Compatibility

The repository's license is not explicitly stated in the provided README.

Limitations & Caveats

The README does not detail specific limitations, known bugs, or deprecation status. The "fully local" setup requires managing local LLM and vector store deployments.