Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions PLAN.md
Original file line number Diff line number Diff line change
Expand Up @@ -49,9 +49,10 @@ Deliver a standalone, installable tool that deterministically checks OpenAI Agen
- [x] (2026-08-14) Proposed a minimal community-tool entry for AgentRunProof on the official v0.21
testing-guide PR #4381. The maintainer declined the listing to keep that guide focused on
SDK-maintained APIs, while explicitly welcoming future reproducible findings backed by the tool.
- [ ] (2026-08-15) Publish v0.2.0 with packaged-wheel coverage for SDK 0.20.0 and 0.21.0, an
- [x] (2026-08-15) Published v0.2.0 with packaged-wheel coverage for SDK 0.20.0 and 0.21.0, an
opt-in provider-free generation span for observability integration tests, and a fresh canonical
comparison bundle bound to the v0.2.0 wheel.
comparison bundle bound to the v0.2.0 wheel. The immutable GitHub Release and PyPI artifacts
have matching wheel and sdist digests, and fresh installs pass against both exact SDK baselines.

## Milestones

Expand Down
4 changes: 3 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -89,7 +89,9 @@ JSON restoration. The revised head `1725a898` passes the built-in restored-appro
runner modes and was squash-merged as `50d65f65`; the upstream 24-case regression also covers
approval and rejection before and after restoration across two and three nested edges. The
immutable v0.1.2 comparison bundle pins the released failure, the intermediate merged behavior,
and the final recursive and serialized PASS results by wheel hash and Git provenance.
and the final recursive and serialized PASS results by wheel hash and Git provenance. The v0.2.0
release carries a freshly reproduced bundle for the same causal ladder, bound to the v0.2.0
harness wheel.

For library scenarios, the top-level package exposes `Scenario`/`ScenarioCase` for one run and `ScenarioPlan`/`ScenarioPhase`/`ResumeInput`/`StateProbe` for ordered multi-run contracts, together with `DeterministicModel`, `RecordingSession`, `run_scenario()`, and certificate helpers. The built-in scenario and the two multi-phase historical scenarios are executable examples.

Expand Down
Loading