Changelog

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.

Easily use the Langfuse documentation in Cursor and other LLM editors via the new llms.txt file.

Langfuse Prompt Management now integrates natively with the Vercel AI SDK. Version and release prompts in Langfuse, use them via Vercel AI SDK, monitor metrics in Langfuse.

We've updated the sidebar in the Langfuse UI collapsible. This change provides more screen space for viewing traces and other complex pages.

Configure SDK-side masking to redact sensitive information from inputs and outputs sent to the Langfuse server.

Langfuse now supports Amazon Bedrock for LLM Playground and Evaluations.

Tool calling makes Langfuse Evals reliable. Previously, only OpenAI models were supported. With this update, you can use any tool-calling LLM when setting up an LLM-as-a-judge evaluator.

Manage your annotation tasks with ease using our new workflow tooling. Create queues, add traces to them, and get a simple UI to review and label LLM application traces in Langfuse.

Large traces can be hard to read. We've added aggregated latency and cost information to the every span level to make it easier to spot outliers and debug the LLM application.

See previews of all commonly asked questions and feature requests for each page across the Langfuse documentation.

Langfuse is now available on AWS Marketplace, making it the premier choice for AWS customers looking for a powerful LLM observability, analytics and evaluation platform.

Langfuse now provides an integration example for DSPy, offering observability for this powerful framework that optimizes language model prompts and weights.

Prompt management just got more powerful for Langchain users by linking Langfuse prompts to Langchain executions.

Deploy Langfuse behind a custom base path for more flexible self-hosting configurations.

Day 1 support for OpenAI's o1 models including tracking token counts and USD spend.

Automatically initialize Langfuse resources on startup using environment variables, streamlining self-hosted deployments and infrastructure-as-code integrations.

Langfuse prompt management now guarantees instant access to prompts after first use while refreshing the cached version in the background.

Add multiple modalities to a single trace, including text and images.

Add comments to traces, observations, sessions and prompts via the Langfuse UI.

Explore customizable charts for aggregate and time series analytics, grouped by score type, source and name.

You can now replace the documentation, support and feedback links in the UI with references to your internal resources. This helps roll out Langfuse in your organization and ensures that users can find the information they need.

Porter now allows you to deploy a production-grade Langfuse cluster via their Add-On marketplace. If you use Porter, this is the easiest way to get started with self-hosted Langfuse.

Introducing organizations to manage projects and permissions across larger teams working in Langfuse.

Langfuse now supports tracing OpenAI Structured Outputs.

The AI SDK now supports OpenTelemetry-based instrumentation. With the new LangfuseExporter, you can easily collect these traces in Langfuse.

Tables are at the core of Langfuse and we finally made them resizable.

Day 1 support for OpenAI GPT-4o mini including tracking token counts and USD spend.

Display trace or observation inputs and outputs in Markdown format. Enhance the readability of Markdown elements such as headers, lists, blocks of code, tables, and links.

The Langfuse API, Python and JS SDKs now support creating categorical, boolean and numeric scores.

The Langfuse Python and JS SDKs now support fetching traces and observations from the Langfuse API. Use them for custom analysis, few-shotting or fine-tuning workflows.

The Langfuse API now supports direct access to model definitions used for tokenization and model price calculation across the Langfuse platform. You can use the API to synchronize model prices from your model registry or add custom models programmatically.

Introduction of the new Datasets API v2 with improved performance for large datasets. Also, now all datasets-related objects are directly accessible via the API.

Build complex LLM applications with Dify's workflow builder and natively trace and monitor them in Langfuse using the new integration.

Visualize agent flow using our trace timeline chart. This new view helps to identify latency bottlenecks, visualize parallelism, and understand multi-step reasoning in deeply nested chains.

For self-hosted instances, we have added a little update indicator that helps to keep track of the latest changes to Langfuse. On Langfuse Cloud, you always run on the latest version.

LLM security requires effective run-time checks and ex-post monitoring and evaluation. We have added new documentation and a cookbook on how to effectively leverage popular LLM security tools while evaluating them in Langfuse.

Manually annotate traces and LLM responses via the new annotation drawer. You can defined a standardized set of scores and criteria for your project.

The OpenAI Assistants API is a powerful tool for building conversational AI applications while relying on OpenAI to manage state, RAG, code execution and more. The new cookbook provides a step-by-step guide to trace applications built with the API.

Introducing dark mode for the Langfuse UI.

Easily monitor and trace your Haystack pipelines with this new Langfuse integration.

Mirascope is a Python toolkit for building LLM applications and it natively integrates with Langfuse.

Compare metrics across different prompt versions in Langfuse Prompt Management.

Sticky headings to better keep track of your data.

Day 1 support for OpenAI GPT-4o including tracking token counts and USD spend.

Experiment with OpenAI GPT-4o in the playground and use it to power evaluations

Manage prompts via the REST API and add labels to deploy prompts to different environments.

Langfuse's Information Security program has been certified to the SOC 2 Type II standard.

Run model-based evaluations on traces in Langfuse to scale your evaluation workflows. Start with one of our battletested templates or use your own custom templates.