lunary

Toolkit for LLM chatbots

website - docs - llm sdk

Features

Lunary helps developers of LLM Chatbots develop and improve them.

• ??? Conversation & feedback tracking
• ?? Analytics (costs, token, latency, ..)
• ?? Debugging (logs, traces, user tracking, ..)
• ?? Prompt Directory (versioning, team collaboration, ..)
• ??? Datasets (for evaluation, fine-tuning, ..)
• ?? Topic & sentiment analysis

It also designed to be:

• ?? Easy to use
• ?? Fast to integrate (2 minutes)
• ????? Self-hostable

1-min Demo

demo-annotated.mp4

?? Integration

Modules available for:

• JavaScript
• Python

Lunary natively supports:

• Abso
• LangChain (JS & Python)
• OpenAI module
• LiteLLM
• Flowise

Additionally you can use it with any other LLM by manually sending events.

?? Documentation

Full documentation is available on the website.

?? Hosted version

We offer a hosted version with a generous free tier.

With the hosted version:

• ?? don't worry about devops or managing updates
• ?? get priority 1:1 support with our team
• ???? your data is stored safely in Europe

Running locally

1. Clone the repository
2. Setup a PostgreSQL instance (version 15 minimum)
3. Copy the content of packages/backend/.env.example to packages/backend/.env and fill the missing values
4. Copy the content of packages/frontend/.env.example to packages/frontend/.env and fill the missing values
5. Run bun install
6. Run bun run migrate:db
7. Run bun run dev

You can now open the dashboard at http://localhost:8080. When using our JS or Python SDK, you need to set the environment variable LUNARY_API_URL to http://localhost:3333. You can use LUNARY_VERBOSE=True to see all the event sent by the SDK

?? Support

Need help or have questions? Chat with us on the website or email us: hello [at] lunary.ai. We're here to help every step of the way.

License

This project is licensed under the Apache 2.0 License.