Langfuse Changelog

Projects using app-root-aware Langfuse SDKs now open the observations table with a focused view of outer roots and application entry points by default.

Trace graph view now offers two layouts: an aggregated overview and an expanded as-it-ran DAG, backed by a rebuilt deterministic renderer.

Langfuse tables now support filtering by true or false score values.

Recorded tool calls can now be accessed directly in code and LLM-as-a-Judge evaluators.

Home is now a customizable dashboard where you can choose which dashboard shows on it, reuse Home widgets in your own dashboards, and copy, duplicate, and share widgets as portable JSON.

Scheduled blob storage exports can now be written as Apache Parquet files, with Parquet set as the default for new integrations and configurable via the public API.

Experiments and their items can now be fetched programmatically through new public API endpoints, or explored by AI agents via the Langfuse MCP server.

Inspect Langfuse media references, data URIs, and media URLs as compact preview tags in JSON viewers, table cells, and dataset item editors.

All posts

Process an annotation queue without the mouse: arrow through items, jump between score fields, pick options with number keys, and complete with Cmd+Enter.

Use @langfuse/vercel-ai-sdk 5.9.0 to trace Vercel AI SDK 7 calls in Langfuse.

Describe the filters you want in plain language and the search bar drafts them as editable query pills. Opt-in on Langfuse Cloud, powered by AWS Bedrock with zero data retention.

Create Langfuse dataset items with images, audio, video, documents, and other attachments for SDK-based multi-modal experiments.

Filter traces and observations by typing. A fast query bar with operators, full-text search, wildcards, and autocomplete.

Create monitors that watch cost, quality, and latency metrics and notify your team over Slack, webhooks, or GitHub Actions when they move outside expected ranges.

The Langfuse Assistant is now in public beta for everyone on Langfuse Cloud. Ask questions about your traces, observations, and metrics in plain language.

Use @langfuse/browser to send frontend feedback scores with only a public key.

Use the new @langfuse/vercel-ai-sdk beta package to trace Vercel AI SDK v7 beta calls in Langfuse.

Call any HTTP endpoint from your trace, observation, or session views to trigger external workflows from Langfuse.