
Langfuse April Update
Langfuse April Update: 10,000 GitHub Stars, Langfuse User Meetup, Q2 Roadmap Highlights & More!
April was a big month at Langfuse—here’s everything we shipped and what’s coming next.
⭐ 10,000 GitHub Stars
On April 3rd, we crossed the 10k-star milestone. Thank you for helping us build the best open-source observability tool for AI! → If you haven’t yet, show us some love.
🕹️ Langfuse User Meetup – San Francisco
Join us for a Langfuse user meetup in San Francisco! Marc will demo the latest features, and several users will showcase how Langfuse powers their AI products. Plenty of time for Q&A and networking.
When: May 12th, 6:30 pm Where: San Francisco (location to be announced) → Sign up here
🗺️ Q2 Roadmap Highlights
You told us what you need—here’s what we’re building this quarter:
- Advanced Evaluation: Session-level scoring; richer evaluation views; LLM-as-judge templates; non-LLM evaluators for CI/CD, rule-based routing
- Tracing: Full-text search across traces & sessions; custom dashboards; query API; OpenTelemetry-native SDKs, webhooks & alerts
- Agent Observability: Generalised agent graphs; filterable tool calls; opinionated evaluations
- Prompt Management: Track prompt variables in prod; LLM-assisted prompt engineering; A/B testing
📈 Configurable Dashboards
Being able to create custom dashboards on top of Langfuse data has been one of the most requested features. We’ve released an initial beta and will add many improvements to this over the next days/weeks.
→ We showed a live demo during the last community town hall → Please share your thoughts/feedback on this here
⚕️ HIPAA Cloud
We’re excited to offer HIPAA-compliant Langfuse Cloud instances! This enables healthcare organizations to safely use Langfuse Cloud while ensuring patient data remains secure and confidential.
→ Reach out to [email protected] for access to the Langfuse HIPAA Cloud.
🛠️ Tool Calling & Structured Outputs
Ship agents faster: Langfuse now supports tool calling and JSON-typed responses end-to-end in the Playground. You can now fetch real-time data, pass it to your LLM, and render structured outputs in your UI—all while tracing every call.
👍 Session-Level Scores
Langfuse now supports session-level scores, enabling comprehensive evaluation of conversational experiences across multiple interactions rather than just individual traces or observations.
➕ More April Releases
- Protected Prompt Labels: Lock critical prompts before they hit production. Admins can now mark prompt labels as protected, preventing accidental edits or deletions.
- Admin API: Automate everything - Organization CRUD, project management, and scoped API keys are now available via Admin API endpoints.
- OpenAI o3, o4-mini & 4.1 support---same day full compatibility with OpenAI’s latest models.
- Copy docs as Markdown---paste straight into ChatGPT/Cursor and speed up your workflow.