Lunary provides a production toolkit for LLM applications, offering observability, prompt management, and evaluation features. It targets developers building LLM chatbots, aiming to simplify development and improvement through features like conversation tracking, cost analytics, debugging tools, and prompt versioning.

How It Works

Lunary operates by integrating with LLM applications via SDKs (JavaScript, Python) or manual event sending. It captures conversation data, LLM interactions, and user feedback, processing this information for analytics, debugging, and prompt management. The system supports various LLM frameworks like LangChain and LiteLLM, with a focus on ease of integration and self-hostability.

Quick Start & Requirements

• Install/Run: Self-hostable via cloning the repository.
• Prerequisites: PostgreSQL (version 15 minimum), Node.js (via bun).
• Setup: Requires setting environment variables (.env files) and running backend/frontend development servers.
• Docs: https://lunary.ai/docs
• Demo: https://github.com/user-attachments/assets/4f95d952-a89b-442b-854a-e71e036073bb

Highlighted Details

• Features conversation tracking, cost/token/latency analytics, debugging logs, and prompt versioning.
• Supports integration with Abso, LangChain (JS & Python), OpenAI module, LiteLLM, and Flowise.
• Offers a self-hostable option and a hosted version with a free tier.
• Includes dataset management for evaluation and fine-tuning.

Maintenance & Community

• Active development indicated by recent commits.
• Support available via website chat or email (hello@lunary.ai).

Licensing & Compatibility

• Licensed under Apache 2.0.
• Compatible with commercial use and closed-source linking.

Limitations & Caveats

Self-hosting requires setting up and managing a PostgreSQL instance and Node.js environment. The project is primarily focused on LLM chatbots, and integration with non-standard LLM setups might require manual event implementation.