[bokelley]

Forward the canonical-format repairs from #1757 and #1758 to the SDK 10 line for AdCP 3.3.

• Typed legacy VAST/DAAST formats retain their version constraints during projection.
• Narrowing covers the eleven published conflict/acceptance vectors, exact numeric values, range boundaries, nullable duration endpoints and exact-duration precedence.
• Declaration/product projection accepts an optional provisioned requirements lookup and emits existing divergence advisories for each full seller reference. Missing lookups remain explicitly unchecked, and lookup failures propagate.

Refs #1750 and #1749. The three source files match the reviewed main fixes. Preserve next’s existing serialization tests; no schema pins, generated models, package versions or dependencies change.

Give the context-ID integration test fixture the unique private name _ContextIdEchoHandler. This prevents another test module’s _EchoHandler registration from hiding create_media_buy during full-suite collection. The test assertions and library behavior are unchanged.

Validation

• Full local gates in Docker: make lint, make typecheck-all, and make test passed. 20603 passed, 2810 skipped, 10 deselected, 1 xfailed, 1484 warnings in 4076.94s (1:07:56); coverage 92.19%.
• Docker with the declared SDK 10 development dependencies: 188 focused tests passed across v1/v2 projection, narrowing, registry, round trips and public API.
• Normal commit hooks passed: formatting, Ruff, source/adopter type checks, Bandit and file/commit-message checks.
• Independent code and security reviews passed on this forward port.
• The vendored vector file matches the published AdCP v3.2.7 source byte-for-byte (SHA-256 3e9ecb82e8103c29cc14cf8ca6175ed71bf68d170ff1020a3065bd2f376e6510).

[aao-secretariat]

Ladon verdict: Approve

Approve — clean forward-port of the #1757/#1758 canonical-format repairs to the SDK 10 / next line.

Checked:

• No findings at any severity (Critical/High/Medium/Low all clear).
• high_risk: false; gated_paths: false; no no-auto-approve team match; no prior decision.
• Reviewer traced the new interval/dedup path in check_narrows against all shipped vectors (None-bound handling, exact-vs-range double-count suppression, numeric guards), the v2→v1 operator-lookup threading (exceptions propagate, missing entries leave refs unchecked, refs preserved on advisory), and the typed-model path in _v1_version_constraints — all correct and test-pinned.
• New params are additive/backward-compatible; new test fixtures and test files accompany the change.

No decision-table trigger fires (rows 1–8 all pass), so this falls through to row 9 → approve.

[aao-secretariat]

Ladon verdict: Approve

Approve — test-only rename, no blocking or medium findings.

This PR's delta since the prior approve is a single test-only change: the helper class _EchoHandler was renamed to _ContextIdEchoHandler in tests/integration/test_a2a_context_id.py, across its definition and all four _running_server(...) call sites. All five references in the base were renamed with no dangling identifier left behind. No source, public surface, or wire shape was touched.

Checks:

• No critical/high findings (decision table row 1 not triggered).
• gated_paths: false, and review_decision is APPROVED regardless (row 2 not triggered).
• high_risk: false — no (deleted)/(modified) sensitive-path triggers (rows 3-5 not triggered).
• Prior decision was approve, not escalate (row 6 not triggered).
• No no-auto-approve team match (row 7 not triggered).
• Zero medium findings (row 8 not triggered).

Falls through to row 9 → approve.