Langfuse

🚀 Big news: We're open sourcing ALL product features under the MIT license! Today marks a major milestone in our journey to build the leading open source LLM Engineering Platform. We're releasing the following features under the MIT License: ✅ LLM-as-a-Judge Evaluations ✅ Annotation Queues ✅ Prompt Experiments ✅ Playground ✅ And more... 𝗪𝗵𝘆 𝗮𝗿𝗲 𝘄𝗲 𝗱𝗼𝗶𝗻𝗴 𝘁𝗵𝗶𝘀? The LLM landscape changes every week. Developers need a platform they can trust, extend, and iterate on without barriers. By removing commercial gates from core product features, we're fostering deeper community collaboration and accelerate development speed. 𝗧𝗵𝗲 𝗻𝘂𝗺𝗯𝗲𝗿𝘀 𝘀𝗽𝗲𝗮𝗸 𝗳𝗼𝗿 𝘁𝗵𝗲𝗺𝘀𝗲𝗹𝘃𝗲𝘀: >7,000,000 monthly SDK installs, > 5,500,000 Docker pulls, and 8,000+ monthly active self-hosted Langfuse instances that we know of. This is our commitment to the developer community - your data stays yours, your workflows stay flexible, and your platform stays open. 👉 Star us on GitHub 👉 Deploy with our new Terraform modules 👉 Join the conversation on our roadmap We're incredibly grateful to work with such an amazing community. The future of LLM engineering is open. Let's build it together.

✨ New integration: Langfuse × BytePlus BytePlus is a suite of AI-powered APIs and services developed by ByteDance, including speech, video, and recommendation technologies. Langfuse integrates with BytePlus to trace and evaluate LLM workflows, enabling observability across generation and user interaction. Link in comments to learn more.

🦎 Introducing the Reptyl Stack: A production-grade foundation for resilient AI applications After months of building complex AI workflows at Pond Labs, I found that libraries like LangGraph make it easy to prototype AI features, but fall short in production. The Reptyl Stack tackles the problems we encountered head-on and includes: 🚀 Restate - Durable execution that makes AI workflows automatically resilient 📊 Langfuse - AI-specific observability (token tracking, cost monitoring, prompt versioning) Why this matters: - AI agents can execute complex, reversible operations (like booking travel) with automatic rollback - Human-in-the-loop workflows suspend gracefully without consuming resources - Multi-agent systems coordinate reliably with built-in fault tolerance 🔥 Hot reload everywhere for rapid iteration and great developer experience. The stack is intentionally flexible—start with this foundation, then add vector databases, specialised AI tools, or additional infrastructure as needs evolve. Built from real experience shipping MeetQu, our AI-powered meeting prep tool. The template is open source and ready to use. Links to the introductory blog post and the GitHub repo in the comments. #AI #SoftwareEngineering #OpenSource

New Integration: Trace LiveKit voice AI Agents with Langfuse LiveKit is an open-source Python and Node.js framework for building production-grade multimodal and voice AI agents. This integration captures traces for key agent activities in Langfuse, like session management, agent turns, LLM node executions, and more, leveraging OpenTelemetry data. Link in comments to get started.

💬 New Slack Integration: Stay informed about your prompt changes directly in Slack! Langfuse now offers a native integration that sends prompt change notifications right to your Slack channels. Whether you're monitoring production or ensuring team coordination, this feature keeps everyone updated. Check out the details on our changelog (link in comments)

⏯️ New in Langfuse: The LLM Playground now features side-by-side prompt comparison, allowing for parallel or selective execution of prompts. Configure independently, test variations simultaneously, and directly save successful prompts to your project. 🔗 Link in Comments to learn more

📊 More Langfuse Dashboards Improvements: CSV Download for Charts You can now download metrics from any chart in your custom dashboards as CSV files, making it easier for further analysis and reporting. Hover over any chart and click the download button to receive your CSV file. This simplifies creating reports and integrating dashboard data into existing workflows. 🔗 Link in comments for more details.

📊 New in Langfuse : Histogram Charts in Custom Dashboards We added histogram charts as a new visualization option in Langfuse custom dashboards. They help you better understand the distribution of your data, highlighting patterns, outliers, and trends in your LLM metrics. Perfect for latency analysis, cost patterns, score distributions, and more! Create your own histogram in a few easy steps: 1. Dashboards → Widgets → New Widget 2. Select data source 3. Choose Histogram as your chart type 4. Select metric to visualize 5. Apply filters, save, and add to your dashboard 🔗 Explore more in our changelog (link in comments)

🔎 New in Langfuse: Search your Prompt Table. You can now search across prompt names and tags, complementing our newly introduced folder system to improve your prompt management experience, helping you locate and organize prompts. More in our changelog (🔗 link in comments)

🧩 New in Langfuse: Prompt Version Webhooks