* feat(langfuse): add setting-up-evals reference for eval setup guidance
Adds references/setting-up-evals.md guiding agents through eval setup: teacher-stance behavior, finding the measurement gap before choosing metrics, defining the metric set with the user, and building the right evaluator. Wires it into SKILL.md routing and tightens error-analysis.md scope and human-labeling responsibility.
* fix(langfuse): cross-link error-analysis and setting-up-evals references
Point the error-analysis step in setting-up-evals.md to references/error-analysis.md for the annotation-queue mechanics instead of inlining them, and hand off evaluator setup in error-analysis.md to references/setting-up-evals.md. Keeps a single source of truth for each and aligns the in-reference paths with SKILL.md routing.
* Add issue-detection-triage skill reference
Adds references/issue-detection-triage.md: a workflow for sweeping recent
production traffic across every dimension where issues hide, ranking
findings P0-P3 with a reproducible decision rule, and reporting them with
evidence and ready-to-paste fix prompts.
Also fixes the pagination tip in cli.md (verified against the CLI:
traces/scores are page-based, only observations is cursor-based, metrics
has no pagination flags) and bumps both plugin manifests to 1.3.0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* sharpen descriptions
* Add dataset construction skill reference
* Refine Langfuse skill references
* Clean up issue triage gotchas heading
* Remove improvement loop from dataset skill PR
* Scope PR to dataset creation
* Expand dataset skill discovery
* Clarify dataset skill routing
* Broaden Langfuse skill routing
* Refine dataset skill routing
* Keep dataset PR focused
* Improve Langfuse skill invocation
* Restore numbered invocation description
* Restore CLI and documentation coverage note
* Update best practices guideline for clarity
Clarified the guideline on best practices by specifying 'use-case-specific guidelines' for better understanding.
* Refine dataset creation workflow
* fix(langfuse): fix dataset skill workflow ordering and drop authoring notes
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Lotte Verheyden <48100308+Lotte-Verheyden@users.noreply.github.com>
* feat: route skill feedback through Langfuse intake
* fix: address review — pin write allowlist, discussion fallback, minor bump
- drop the two 'api feedback *' allowed-tools entries so the feedback
submit call goes through the normal permission prompt instead of being
auto-approved (allowlist stays read-only verbs)
- replace the GitHub issue fallback with a prefilled discussion link in
Ideas & Improvements; offered alongside intake in the consent step for
users who want a public, trackable thread
- restore proactive trigger guidance in the references list entry and
frontmatter description (offer once on wrong/outdated guidance)
- drop GITHUB from required_access (no gh CLI usage remains) and align
wording with the shipped API (MCP server tool, correlation ID)
- bump manifests to 1.5.0 per semver policy (new capability)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* docs(feedback): mention reply email for replies
* docs(feedback): simplify submission options
* minor improvements
* nit
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* docs(langfuse): trim prompt-migration reference to migration-specific workflow
Cut ~150 lines of Langfuse API mechanics (create_prompt/get_prompt/
compile, prompt-trace linking, templating tables) that the docs now
cover, replacing them with fetch-at-execution links. Keeps the
migration-specific judgment: prompt inventory, {{var}} conversion,
structure decisions, approval gate, and verification.
Also fixes stale /docs/prompts/* links to /docs/prompt-management/*
and bumps plugin manifests 1.4.1 -> 1.4.2.
* docs(langfuse): apply reviewed prompt-migration content
Adds env-var (LANGFUSE_BASE_URL/HOST) handling, chat-vs-text and
composability fetch guidance, and non-blocking authorization framing
for the migration flow.
* chore(langfuse): refine evaluator upgrade documentation
* docs: add back fallback mechanism to retrieve skills in single GET
* docs: add back notes on dataset eval migration
langfuse-docs#3360 restructures the PostHog and Mixpanel integration pages
v4-first: the old 'Export source (Fast Preview, Langfuse v4)' section became
'Legacy export sources', so the #export-source-fast-preview-langfuse-v4
anchor no longer exists. Point the migration links at the explicit
#migrate-export-source anchor and add links to the new #legacy-vs-enriched
schema-change sections (the analytics counterpart of the blob-storage
field-differences link).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Enabling the dual source is additive — legacy output continues unchanged —
so it needs no downstream sign-off; only the switch that stops legacy
output requires explicit confirmation. Dual-mode duplicates inflate
downstream metrics only where a consumer ingests both shapes, so say
that instead of claiming unconditional inflation.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The docs on main define broader legacy-source cutoffs than the single
2026-05-20 project date (integration-level cutoff on Cloud, self-hosted
events_only write mode), so state eligibility generically and defer to
the docs. Also warn that legacy directories stop receiving files after
the switch, so pipelines watching them go silent without an error.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The "Migrate exports" section restated export mechanics inline and
carried factual drift against the docs. Rework it to reference the
documented upgrade paths, keep only decision-relevant takeaways, and
focus warnings on effects outside Langfuse (customer data pipelines).
Corrections against the current docs:
- Blob Storage integration API uses project-scoped keys, not
organization-scoped credentials
- add the 2026-05-20 Cloud cutoff gate (post-cutoff projects cannot
select legacy sources; nothing to migrate)
- add that a source change applies to future exports only — history
keeps the legacy shape unless re-exported
- note scores exports are unaffected by the source change
Patch version bump per AGENTS.md (content correction).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Remove the standalone sdk-upgrade.md reference file and repoint its
callers at the docs. The v4-project-migration reference now links to the
website's SDK upgrade paths hub, the SKILL.md use-case list no longer
routes to the deleted file, and the codex plugin's v4 migration
defaultPrompt is simplified to a short starter.
dependabot[bot]