lunary-ai / lunary

The production toolkit for LLMs. Observability, prompt management and evaluations.

📈 lunary

Open-source observability, prompt management & evaluations for LLMs

website - docs - self host

Features

Lunary helps LLM developers take their apps to the next level.

• 💵 Analytics (cost, token, latency, ..)
• 🔍 Monitoring (logs, traces, user tracking, ..)
• ⛩️ Prompt Templates (versioning, team collaboration, ..)
• 🏷️ Create fine-tuning datasets
• 🖲️ Chat & feedback tracking
• 🧪 Evaluations

It also designed to be:

• 🤖 Usable with any model, not just OpenAI
• 📦 Easy to integrate (2 minutes)
• 🧑‍💻 Self-hostable

1-min Demo

demo720.mp4

⚙️ Integration

Modules available for:

• JavaScript
• Python

Lunary natively supports:

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

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 free plan of up to 10k requests / month.

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/backend/.env
5. Run npm install
6. Run npm run migrate:db
7. Run npm 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.