v0.2.0

[jrcks67]

[0.2.0] - 2026-05-01

Fabric earns the "open-source observability + control plane for AI agents" framing by capturing LLM operations natively. Three substantive additions over the v0.1.x line: per-LLM-call child spans with
gen_ai.* semantic conventions, auto-instrument extras for the popular LLM SDKs, and a trace pipeline on the OTel collector's custom guard processor so the chart's privacy promise actually applies to the
SDK's spans (not just future L2 bridge log records).

This release subsumes the never-tagged 0.1.3 audit follow-up work (round-2 audit fixes across SDK, components, charts, docs, and specs). 0.1.3 was prepped on the chore/v0-1-3-audit-followup branch and
merged to main, but the new public API additions in that branch (Decision.llm_call, Decision.tool_call, the [openai]/[anthropic]/... extras, fabricguardprocessor trace processing) are semver-minor
work, not patch — so we skip the 0.1.3 tag and ship everything as 0.2.0.

…followed by the full per-component changelog: Added (SDK), Added (OTel collector), Changed (docs / scope), Fixed (CI), Round-3 audit follow-up (PR #49), Audit follow-up (folded from never-tagged 0.1.3),
Fixed (SDK), Fixed (components), Fixed (charts), Fixed (docs), Operator action required.

---

This discussion was created from the release v0.2.0.