langfuse

v3.177.0 Feature

This release adds 5 notable features for engineering teams evaluating rollout.

Published 2d Tracing

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

analytics autogen evaluation langchain large-language-models llama-index

+12 more

llm llm-evaluation llm-observability llmops monitoring observability openai playground prompt-engineering prompt-management self-hosted ycombinator

Summary

AI summary

Adds LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag, AI telemetry toggle, claude‑opus‑4‑8 model, UI notifications for code evals and MCP v2, plus fixes in mcp, search, traces, eval performance, and dependency chores.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Low	Adds LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag to disable FTS on v3 tables Adds LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag to disable FTS on v3 tables Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Low	Adds toggle for AI telemetry Adds toggle for AI telemetry Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Low	Adds claude-opus-4-8 model price entry Adds claude-opus-4-8 model price entry Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Low	Adds notification for code evals launch in UI Adds notification for code evals launch in UI Source: llm_adapter@2026-06-01 Confidence: high	—
Feature	Low	Adds notification for MCP v2 in UI Adds notification for MCP v2 in UI Source: llm_adapter@2026-06-01 Confidence: high	—
Performance	Low	Improves blob-export query performance by replacing FINAL with LIMIT 1 BY Improves blob-export query performance by replacing FINAL with LIMIT 1 BY Source: llm_adapter@2026-06-01 Confidence: high	—
Performance	Low	Improves eval performance by skipping FINAL and unused aggregations in checkTraceExistsAndGetTimestamp Improves eval performance by skipping FINAL and unused aggregations in checkTraceExistsAndGetTimestamp Source: llm_adapter@2026-06-01 Confidence: high	—
Bugfix
Bugfix	Medium	Refines handling of LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag Refines handling of LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag Source: llm_adapter@2026-06-01 Confidence: high	—
Bugfix	Medium	Tightens matches operator semantics Tightens matches operator semantics Source: llm_adapter@2026-06-01 Confidence: high	—
Bugfix	Medium	Fixes infinite recursion in buildStepGroups of traces module Fixes infinite recursion in buildStepGroups of traces module Source: llm_adapter@2026-06-01 Confidence: high	—
Bugfix	Low	Stabilizes score config pagination order in API and MCP Stabilizes score config pagination order in API and MCP Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Stops retrying context overflow errors in evals Stops retrying context overflow errors in evals Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Hides duplicate SDK warning in fast preview of evals Hides duplicate SDK warning in fast preview of evals Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Removes intersection or union JSON schemas usage in MCP Removes intersection or union JSON schemas usage in MCP Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Adds missing audit logs in MCP Adds missing audit logs in MCP Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Renames create tool operation to upsert in MCP Renames create tool operation to upsert in MCP Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Throws error if entity not exists in deleteAnnotationQueueAssignment MCP call Throws error if entity not exists in deleteAnnotationQueueAssignment MCP call Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Uses IDs instead of names for dataset tools in MCP Uses IDs instead of names for dataset tools in MCP Source: granite4.1:30b@2026-06-01-audit Confidence: low	—
Bugfix	Low	Matches escaped Unicode content in full‑text search Matches escaped Unicode content in full‑text search Source: granite4.1:30b@2026-06-01-audit Confidence: low	—

Full changelog

What's Changed

Features

feat: introducing LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH flag to allow disabling FTS on v3 tables by @sumerman in https://github.com/langfuse/langfuse/pull/13912
feat(ai): Add toggle for AI telemetry by @bezbac in https://github.com/langfuse/langfuse/pull/13939
feat(model-prices): add claude-opus-4-8 by @hassiebp in https://github.com/langfuse/langfuse/pull/13919
feat(ui): notification for code evals launch by @jannikmaierhoefer in https://github.com/langfuse/langfuse/pull/13894
feat(ui): notification for MCP v2 by @jannikmaierhoefer in https://github.com/langfuse/langfuse/pull/13895

Fixes / Improvements

fix: refine LANGFUSE_DISABLE_LEGACY_TRACING_IO_SEARCH handling by @sumerman in https://github.com/langfuse/langfuse/pull/13929
fix: tighten matches operator semantics by @sumerman in https://github.com/langfuse/langfuse/pull/13928
fix(agents): require exhaustive datadog alert sweep by @maxdeichmann in https://github.com/langfuse/langfuse/pull/13968
fix(api,mcp): Stabilize score config pagination order by @peter941221 in https://github.com/langfuse/langfuse/pull/13832
fix(eval): stop retrying context overflow errors by @wochinge in https://github.com/langfuse/langfuse/pull/13930
fix(evals): hide duplicate SDK warning in fast preview by @marliessophie in https://github.com/langfuse/langfuse/pull/13947
fix(mcp): Do not use intersection or union JSON schemas by @bezbac in https://github.com/langfuse/langfuse/pull/13927
fix(mcp): Missing audit logs by @bezbac in https://github.com/langfuse/langfuse/pull/13915
fix(mcp): Rename relevant tools from create to upsert by @bezbac in https://github.com/langfuse/langfuse/pull/13913
fix(mcp): Throw if not exists in deleteAnnotationQueueAssignment by @bezbac in https://github.com/langfuse/langfuse/pull/13914
fix(mcp): Use ids instead of names for dataset tools by @bezbac in https://github.com/langfuse/langfuse/pull/13916
fix(search): match escaped unicode content in full-text search - Issue #11538 by @sophyphreak in https://github.com/langfuse/langfuse/pull/13644
fix(traces): Infinite recursion in buildStepGroups by @bezbac in https://github.com/langfuse/langfuse/pull/13900
perf(blob-export): replace observations FINAL with LIMIT 1 BY by @niklassemmler in https://github.com/langfuse/langfuse/pull/13935
perf(eval): skip FINAL and unused aggregations in checkTraceExistsAndGetTimestamp by @Steffen911 in https://github.com/langfuse/langfuse/pull/13934

Chores

chore: adding a few more tests after #13644 by @sumerman in https://github.com/langfuse/langfuse/pull/13967
chore(deps): dedupe and bump by @nimarb in https://github.com/langfuse/langfuse/pull/13911
chore(mcp): Update MCP server version by @bezbac in https://github.com/langfuse/langfuse/pull/13883
ci: notify slack on critical workflow failures by @wochinge in https://github.com/langfuse/langfuse/pull/13910
ci(deps): bump the github-actions group with 3 updates by @dependabot[bot] in https://github.com/langfuse/langfuse/pull/13924

New Contributors

@peter941221 made their first contribution in https://github.com/langfuse/langfuse/pull/13832
@sophyphreak made their first contribution in https://github.com/langfuse/langfuse/pull/13644

Full Changelog: https://github.com/langfuse/langfuse/compare/v3.176.0...v3.177.0

View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track langfuse

Get notified when new releases ship.

About langfuse

🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, LiteLLM, and more. YC W23

All releases →