Skip to content

v0.2.3 — Result packs, fingerprint reporting, drift, publication primitives

Objective

Ship publishable campaign result packs, fingerprint-informed comparative and longitudinal reporting, readable campaign/semantic outputs, committed-artifact drift guards, pack and directory comparison tooling, and honest browser/MCP capability boundaries—while keeping default CI offline.

Scope (delivered)

  • Verification matrix — canonical just ci, validate-artifacts, and uv run fallbacks documented in docs/workflows/verification.md.
  • Drift checks — tests/test_schema_drift_contracts.py over committed examples/ / fixtures/; docs/audits/schema-drift-contracts-inventory.md.
  • Campaign result packs — alwm benchmark campaign pack / pack-check; campaign_result_pack; examples/campaign_result_packs/.
  • Pack comparison — alwm benchmark campaign compare-packs → campaign_result_pack_comparison; example examples/campaign_result_packs/compare_minimal_vs_multi/.
  • Campaign directory comparison — alwm benchmark campaign compare → campaign_compare (see CHANGELOG and examples/campaign_compares/).
  • Fingerprint interpretation — analysis JSON + Markdown (longitudinal, comparative, attribution).
  • Browser reporting — fixture-backed evidence in reports; partial MCP stdio path documented and tested.

Non-goals (unchanged from roadmap)

  • Dashboards, cloud runners, remote/IDE MCP, or mandatory live stacks in default CI.

Verification

  • uv run just ci
  • just validate-artifacts (or uv run pytest tests/test_schema_drift_contracts.py)
  • Spot checks from docs/roadmap/v0.2.3.md — Verification commands.

Next milestone

v0.2.4 shipped (docs/releases/v0.2.4.md). v0.2.5 shipped (docs/releases/v0.2.5.md).