v1.2.1

What's Changed

• fix(hook): narrow the state lock so concurrent sessions keep their turns by @milanagm in #89
• feat: attach project name to trace metadata by @johnbuccola in #60
• Remove install-required flags from API-key userConfig fields by @SalesforceRocks in #40
• docs: verify Claude traces via Observations API v2 by @maxdeichmann in #91
• docs(readme): slim the README and add a contributing guide by @milanagm in #94
• fix(hook): trace subagents launched by subagents by @pushkar in #102

New Contributors

• @johnbuccola made their first contribution in #60
• @SalesforceRocks made their first contribution in #40
• @maxdeichmann made their first contribution in #91
• @pushkar made their first contribution in #102

Full Changelog: v1.2.0...v1.2.1

v1.2.0

What's Changed

• feat(hook): trace Claude Code subagent transcripts by @milanagm in #11
• docs(plugin): clarify Claude Observability setup by @milanagm in #14
• feat(hook): opt-in deterministic trace IDs via CC_LANGFUSE_TRACE_SEED by @jannikmaierhoefer in #23
• fix(hook): tag skills invoked via slash commands by @milanagm in #20
• fix(hook): eliminate silent failure modes and heal desktop GUI tracing by @milanagm in #26
• fix(hook): persist open turns across Stop firings by @milanagm in #19
• feat(hook): attach headless Claude Code runs to an existing parent trace by @milanagm in #32
• fix(hook): record usage.speed in generation metadata by @milanagm in #36
• feat(hook): drop numbering from LLM call observation names by @milanagm in #44
• fix(hook): configurable state directory via CC_LANGFUSE_STATE_DIR by @milanagm in #46
• fix(hook): let per-repo env vars win over machine-wide wizard config by @milanagm in #47
• fix(plugin): align test pin, wizard default and docs with the EU code default by @milanagm in #53
• fix(hook): trace Workflow-spawned subagent transcripts by @milanagm in #45
• feat(hook): configurable trace name style for name-based grouping by @milanagm in #55
• feat(hook): capture image blocks as markers and opt-in Langfuse media by @milanagm in #56
• perf(hook): read workflow run directories only when a turn references them by @milanagm in #64
• chore(hook): align the Langfuse SDK pin with the documented v4 minimum by @milanagm in #65
• fix(hook): attribute named-teammate subagent transcripts by @milanagm in #66
• fix(plugin): bump the version so updates reach installed users by @milanagm in #71
• docs(plugin): restructure the README and document how to update by @milanagm in #72
• fix(hook): end a generation at its assistant row, not its tool results by @milanagm in #79
• feat(hook): capture assistant thinking blocks as ChatML thinking parts by @milanagm in #76
• fix(hook): restore logging when keys are missing instead of silent fail by @milanagm in #78
• fix(hook): price cache writes by the lifetime they were written under by @milanagm in #73
• feat(hook): record the conversation history as each LLM call's input by @milanagm in #80
• fix(hook): emit tool calls in the nested shape Langfuse renders by @milanagm in #77
• fix(hook): skip history a fork copied from its original session by @milanagm in #81
• fix(hook): preserve tool and turn failure levels by @milanagm in #84
• test(hook): update stale tool_calls expectations after #77 by @milanagm in #85
• feat(hook): accept caller trace tags via CC_LANGFUSE_TRACE_TAGS by @milanagm in #87
• ci: add CI, release and workflow-security workflows by @milanagm in #86
• chore(plugin): bump the version to 1.2.0 for release by @milanagm in #88

New Contributors

• @jannikmaierhoefer made their first contribution in #23

Full Changelog: v1.0.0...v1.2.0

v1.0.0 — Marketplace release

Various improvements to the Claude x Langfuse plugin. This release publishes the newest version to the Claude Marketplace.