Product Thumbnail

Laminar

Open-source all-in-one platform for engineering AI products

Productivity
Developer Tools
Artificial Intelligence
GitHub

Laminar is an open-source platform where you can trace your LLM app, run evaluations, label production data and use it to improve your prompts. Laminar is fast, reliable and offers best-in-class DX. It’s written in Rust and built on top of a modern tech stack.

Top comment

Hey there 👋, I’m Robert, co-founder and CEO of Laminar (YC S24) -> https://www.ycombinator.com/comp... Good AI engineering is all about data and it boils down to three things – collect data, analyze data, use data. Laminar enables teams to do all of that in a single platform. > Unbelievable alpha in looking at data ©️ Hamel Husain (https://x.com/HamelHusain/status...) Collect data – Tracing - Trace every execution step of your LLM application by adding just 2 lines of code. - Laminar automatically traces common LLM SDKs (OpenAI, Anthropic …), and popular LLM frameworks (Langchain …) - We support images, and soon audio! Analyze data – Labeling - Laminar provides tools for your team to efficiently label the data and analyze human labels distribution. - Use this labeled data to better align LLM-as-a-judge with your app requirements. - Use labels as filters to separate “good” and “bad” data. Use data – Evaluations and Datasets - Run custom evaluations locally, from terminal, code, or as part of CI/CD. Visualize and compare results in Laminar UI. - Run hosted online evaluators (LLM-as-a-judge or Python scripts) as production traces come in. - Store golden datasets in Laminar and run evaluations on them. - Collect “good” examples in datasets and then retrieve most relevant ones via API to use as few-shot examples in your prompts. Laminar is fully open-source, built for scale, and production-ready. You can easily self-host it or use our managed version for production. Laminar is extremely fast, reliable, and built on top of a modern stack. Our frontend is in Next.js, backend is written in Rust🦀, trace ingestion is done over gRPC, messages are sent across services through RabbitMQ and analytics is managed by ClickHouse. Join the ranks of our contributors and star ⭐ our repo here → https://github.com/lmnr-ai/lmnr (we have lots of exciting things to do!) Get started with a managed version today → https://www.lmnr.ai (use our ProductHunt discount to get 3 months free of pro version).

Comment highlights

Although I am not building on AI as of now, Laminar is the only viable way that I envision for myself when I will finally do! Amazing product and even more amazing team, congrats on the launch! Alga!!

I recommend Laminar for developers working with AI products and large language models. This open platform offers all the tools you need to track your LLM applications, run estimates, label production data, and optimize queries.

Congrats lmnr team! I can attest to it being the only reliable and performant LLM monitoring platform I've tried. Founding team is great to talk to and super responsive.

Congrats on your launch! Looks like a great tool and I am sure that it can help many engineering teams

Congrats! I've been looking for tools to manage data labeling and evaluations. Cool product!

Huge congrats to the Laminar team on today's launch! I love how you're streamlining AI product engineering into one open-source platform. Quick question: How do you envision developers leveraging Laminar to improve their LLM app prompts - are there any specific use cases or success metrics you're excited to see emerge from the community?

The labeling feature is something I’ve been looking for in my own workflow. For larger datasets having tools to label efficiently and categorize data can save a lot of hassle.

Congrats on the launch @skull8888888 We're currently evaluating a few players in this space and laminar looks perfect! Awesome work!

Congrats on your launch, team Laminar! I'm fond of the name and logo, is there a story behind them?

Congrats on the launch guys! I love to see the progress - it looks so easy to setup now 👀