Skip to content

chore: v1.0 beta->stable promotion pass (toolbox tools + orchestration/patterns/policy)#130

Merged
johnnichev merged 1 commit into
mainfrom
chore/v1-beta-promotions
Jun 15, 2026
Merged

chore: v1.0 beta->stable promotion pass (toolbox tools + orchestration/patterns/policy)#130
johnnichev merged 1 commit into
mainfrom
chore/v1-beta-promotions

Conversation

@johnnichev

Copy link
Copy Markdown
Owner

First v1.0 API-freeze promotion pass. Markers only — additive, nothing removed or renamed. Driven by the review note (cross-referencing each symbol's marker against its module's ROADMAP promise).

Promoted to @stable

  • 24 mature toolbox tools: calculator, code (incl. execute_shell), db, email, github, linear, notion, pdf, search, slack. 4–15 releases old, frozen signatures, in already-stable category modules.
  • SessionSearchResult, agent.PlanningConfig.

Module promises upgraded to stable

orchestration (+ checkpoint/graph/node/state/supervisor), patterns (+ debate/plan_and_execute/reflective/team_lead), policy. Their public surfaces were already 100% @stable (the v0.25 promotion sweep); this aligns the module-level promise. EXPECTED_MODULE_STABILITY updated so the architecture test stays the gate.

Held @beta deliberately

  • The v0.27.0 additions (make_recall_tool, PromptInjectionGuardrail, calculate_cost_with_cached_input, MongoSessionStore, DynamoDBSessionStore) — one more cycle.
  • evals stays a beta module — a data-driven deviation from the original "upgrade all four" plan: its surface is 71-beta / 7-stable (the 50 evaluator classes are still growing), so promising the module stable would be a false promise. Its 7 core types (EvalSuite/TestCase/EvalReport/…) remain @stable.

Tooling

scripts/stability_audit.py — reproducible report of symbols whose marker disagrees with their module promise. After this pass: promotion candidates 42→17, lagging modules 94→13 (the remainders are the intentional holds).

Verification

  • Architecture stability test (the gate): 1192 passed.
  • mypy clean (180 files); ruff + bandit clean.
  • Full suite: 7638 passed, 165 skipped, 0 failed.

Lands in [Unreleased]; ships with the next release (1.0 freeze or 0.29.0).

…/patterns/policy)

First freeze-prep promotion pass (markers only — additive, nothing removed/renamed).

Promoted to @stable:
- 24 mature toolbox tools (calculator/code/db/email/github/linear/notion/pdf/search/slack), 4-15 releases old with frozen signatures
- SessionSearchResult, agent.PlanningConfig

Module promises upgraded to stable (surfaces already fully @stable since v0.25):
- orchestration (+checkpoint/graph/node/state/supervisor), patterns (+debate/plan_and_execute/reflective/team_lead), policy
- EXPECTED_MODULE_STABILITY updated to match (architecture test is the gate)

Held @beta deliberately:
- v0.27.0 additions (recall tool, injection guardrail, cache-cost, Mongo/Dynamo) — one more cycle
- evals stays a beta MODULE: surface is 71-beta/7-stable (50 evaluator classes still growing); the 7 core types stay @stable

Added scripts/stability_audit.py (reproducible mismatch report). Architecture gate 1192 passed; mypy clean; full suite 7638 passed / 0 failed.
@johnnichev johnnichev merged commit 7850add into main Jun 15, 2026
9 checks passed
@johnnichev johnnichev deleted the chore/v1-beta-promotions branch June 15, 2026 18:09
johnnichev added a commit that referenced this pull request Jun 15, 2026
Ships the beta->stable promotion pass (24 toolbox tools, orchestration/patterns/policy module promises, SessionSearchResult, PlanningConfig) merged in #130. Version 0.28.0->0.29.0, CHANGELOG [0.29.0] synced, README What's New v0.29, version stamps, llms-full regenerated via scripts/build_llms_full.py.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant