Changelog

Launch Week 4 🚀

Mixpanel integration

We teamed up with Mixpanel to integrate LLM-related product metrics into your existing Mixpanel dashboards.

Langchain v1 Support

Langfuse SDKs now support Langchain v1 for both Python and JS/TS. The integration remains stable and backward compatible.

Natural Language Filtering for Traces

Filter your traces using plain English queries. Powered by AWS Bedrock with zero data retention.

Mutable Score Configs

Score configurations can now be updated after creation. Modify existing configs via API/SDK and UI while keeping all your data safe.

TypeScript SDK v4 (GA)

The new OpenTelemetry-based TypeScript SDK v4 is now generally available with improved DX, modular packages, and seamless integrations.

Full-Text Search Across Dataset Items

Find dataset items by searching through their actual content with our new full-text search capability

Docs now available as Markdown (.md) endpoints

Append .md to any docs URL to fetch the page as Markdown. Built at compile time for fast, reliable access.

OpenAI GPT-5 pricing now available in Langfuse

Day 1 support for OpenAI GPT-5 including tracking token counts and USD spend.

Annotation Queue Assignments

Assign users to annotation queues to make it easier for team members to focus on relevant tasks.

LLM Playground with Side-by-Side Comparison

The LLM Playground now supports side-by-side prompt comparison with parallel LLM execution.

Sessions in Annotation Queues

Annotation Queues now support session-level annotation, making it easier to evaluate multi-turn interactions in your LLM applications.

Trigger Remote Custom Evaluations from the Langfuse UI

Enable non-technical users to trigger external experiment pipelines directly from the Langfuse UI on datasets.

Full-Text Search Across Prompt Content

Find prompts by searching through their actual content with our new full-text search capability

AWS SDK Default Credential Provider Chain Support

Self-hosted deployments can now use the AWS SDK default credential provider chain for Amazon Bedrock LLM connections.

Gemini 2.5 Pro, Flash, and Flash Lite support for playground, evaluations, and cost tracking

Langfuse says hello to Gemini 2.5 Pro (GA), Gemini 2.5 Flash (GA), and Gemini 2.5 Flash Lite (Preview). Support added for the LLM playground, evaluations, and cost tracking.

Message Placeholders in Chat Prompts

Inject context or message history into your chat prompts at just the right spot.

Search to Navigate Large Prompt Collections

Now it is easier to find prompts in prompt management via the new search bar

Improved JSON Handling in LangChain Prompts

Simplified JSON handling in prompts designated for LangChain consumption - no more extra escaping required.

Exports of Datasets and Audit Logs

Datasets and Audit Logs can now be exported as CSV and JSON directly from the UI.

CSV Download for Dashboard Charts

Download metrics from any chart in your custom dashboards as CSV files for further analysis and reporting.

OpenAI o3-pro and updated prices for o3 reflected in Langfuse cost tracking

Langfuse adds same day support for OpenAI's o3-pro model and updated o3 pricing

Pipecat Tracing Integration

Trace realtime audio agents built with Pipecat via OpenTelemetry and Langfuse

Claude Sonnet and Opus 4 support

Langfuse says hello to Anthropic Claude Sonnet 4 & Opus 4 with day 1 support for both the LLM playground, LLM-as-a-judge and cost tracking.

Launch Week 3 🚀

Save and share table views

Pin your favorite table settings and share them with a single link.

Launch Week 3 🚀

Full Text Search

Find content in inputs and outputs of your traces and observations with full text search.

Customize Product Module Visibility

Control which product modules are visible in the main menu through environment variables when self-hosting Langfuse

Metrics API for custom analytics

Build custom reports and dashboards with our flexible Metrics API. Query traces, observations, and scores with custom dimensions, metrics, and time granularity.