Changelog

Define and test tool calls and structured output schemas directly in the Langfuse Playground

Quickly navigate trace details in the table view via keyboard shortcuts or by clicking on the table rows.

Our redesigned prompt view brings enhanced visualization, version comparison, and search capabilities.

Introducing our redesigned trace view with improved hierarchy visualization, timeline/tree toggle, and powerful search capabilities.

Trace your OpenAI Response API calls with our Python and JS SDKs

Evaluation scores can now be exported in bulk directly from the Langfuse UI for easier analysis and reporting. Supports CSV and JSON formats.

Access and manage your annotation queues programmatically with our new Public API endpoints.

Create modular, reusable prompts by referencing existing prompts within other prompts.

Use native environment to separate staging, dev, and production systems within the same Langfuse project.

When self-hosting or on Langfuse Cloud Teams Plan, you can now use WorkOS as an Identity Provider (IdP).

Dataset items and runs can now also be permanently deleted via the API and UI. Until now, these could only be archived.

You can now duplicate dataset items into other datasets and add them to many datasets at once when creating a new dataset item from a trace.

Langfuse offers a new Java API client for easy integration with your Java applications.

Langfuse now offers full support for OpenAI's GPT-4.5, featuring day-one integration with our LLM playground, LLM-as-a-judge evaluation system, and comprehensive cost tracking capabilities.

New onboarding screens that introduce the capabilities of Langfuse features that are not yet used within a project.

Langfuse says hello to Anthropic Claude 3.7 Sonnet with day 1 support for both the LLM playground, LLM-as-a-judge and cost tracking.

Langfuse now supports access to Gemini models via Google AI Studio.

You can now choose to execute LLM-as-a-judge evaluations on existing, new, or all data.

Use Langfuse prompts in your LLM Agent systems, Claude Desktop, and Cursor and other MCP Clients via the new MCP Server.

Push your OpenTelemetry Spans to Langfuse

Together with LiteLLM, we launch the OSS LLMOps Stack which is a popular, well-integrated and battle-tested LLMOps stack that can easily be self-hosted and extended.

Follow your LangGraph agent execution on a graph view

Quickly select all items subject to your filters and perform actions on them.

Focus on the most relevant elements of your trace by using log-level filters

Langfuse says hello to Gemini 2.0 Flash, Gemini 2.0 Flash-Lite, and Gemini 2.0 Pro. Support added for both the LLM playground and cost tracking.

Access all of Langfuse public API resources via a typed client in our SDKs

Two hours ago, OpenAI released the latest version of their o3-mini model. Langfuse now supports this model in both the LLM playground and cost tracking.

Quickly access different Langfuse features and switch between projects via the Cmd+K menu.

Automatically remove event data outside of a specified retention period.

Langfuse now has a new API Reference with interactive examples.

Complex prompts are easier to edit with our new prompt editor, which supports variable highlighting.

Trace sampling is now supported in the JS SDK and integrations.

Document your updates to prompts

Create your Datasets in seconds

Configure additional HTTP headers for LLM API keys to include custom headers in requests to provider endpoints, particularly useful for OpenAI-compliant LLM proxies.

See a detailed comparison of changes between prompt versions. Optionally, you can also review changes before creating a new prompt version.

As part of our ongoing effort to improve Langfuse for larger teams, we are excited to announce that audit logs are now available on Enterprise/Teams plans.

After running for some time in public beta, we are excited to announce that the PostHog integration is now fully available. It runs on a hourly basis and is available to all Langfuse plans (cloud) and Pro/Enterprise (self-hosted).

Langfuse now supports cost tracking all usage types such as cached tokens, audio tokens, reasoning tokens, etc.

Langfuse v3 is now stable and ready for production use when self-hosting Langfuse, including many scalability and architectural improvements.

To make it easier to get started with the Langfuse JS/TS SDK, we've created an extensive end-to-end example notebook. It includes a general introduction and examples for Anthropic, OpenAI SDK and LangChain.

Langfuse now supports GitHub Enterprise and Keycloak as SSO providers on Langfuse Cloud and self-hosted instances

Comprehensive guides for tracing Google Vertex AI and Gemini models with Langfuse

Langfuse now supports Google Vertex AI incl. Gemini models for LLM Playground and Evaluations.

Move fast on prompts without breaking things! Run experiments on Datasets and directly compare evaluation results side-by-side. Experimentation speeds up the feedback loop when working on prompts and prevents regressions when making rapid changes.

We've completely rebuilt the documentation for Datasets and Evals to make it easier to get started with offline evaluation. To celebrate Launch Week, we've also summarized all the documentation improvements we've made over the past year.

Add multi-modal attachments to LLM traces in Langfuse to observe, and view them within the Langfuse UI.

Introducing support for managed LLM-as-a-judge evaluators for dataset experiments.

After running experiments on datasets, you can now compare results side-by-side, view metrics, and peek into details of each dataset item across runs.