You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
3.1 GA gating — port the v1↔v2 projection layer from the TypeScript reference (adcp-client #1815) into the Python SDK.
Why
Schemas landed in AdCP #3307 / #4770: 12 canonical formats, ProductFormatDeclaration discriminator, 14 reference fixtures, pixel_tracker asset with bidirectional v1↔v2 contract, canonical-projection-ref.json (object-form canonical: on v1 catalog entries), v1-canonical-mapping.json registry.
Without Python SDK support, Python adopters can't consume canonical-formats cleanly — they'd hand-roll v1↔v2 translation per integration. The TS prototype (adcontextprotocol/adcp-client#1815) is the reference; the Python work mirrors it.
What's needed
Type generation — generate Pydantic models from canonical-formats schemas. 13 format_kind values as a tagged union; params typed per canonical; slots[] typed against asset-group vocabulary; pixel_tracker discriminated by event × method.
v2 → v1 projection — mirror the resolution order in v1-canonical-mapping.json:
format_kind: \"custom\" + canonical_formats_only: true → no v1 emit
v1_format_ref[] set → use directly (multi-size fan-out per ref)
Else: registry lookup → structural match → FORMAT_PROJECTION_FAILED advisory
v1 → v2 projection — invert via the v1 catalog's canonical: object-form annotation ({ kind, asset_source?, slots_override? }). Family-level structural match emits FORMAT_DECLARATION_V1_AMBIGUOUS advisory.
Bidirectional pixel_tracker contract — implement the cross-version downgrade/upgrade table in pixel-tracker-asset.json description. Emit PIXEL_TRACKER_LOSSY_DOWNGRADE / PIXEL_TRACKER_UPGRADE_INFERRED advisories on errors[].
Closed-set semantics on format_options[] — sellers MUST reject create_media_buy targeting any format_kind outside the list. SDK helper to validate manifests against the closed set before the seller call.
Test against the 14 reference fixtures at static/examples/products/canonical/ in the main repo. Each fixture should round-trip cleanly: v1 read → v2 projection → v1 reprojection should produce the same format_ids[].
Classification: Feature request Bucket(s): client, validation, cross-repo (no matching labels exist in this repo — flagged in run summary) Status: deferred
What the experts said:
ad-tech-protocol-expert: Blocked on three hard prerequisites — canonical-formats schemas absent from local cache (v3.0.7 has no canonical/ tree), v1-canonical-mapping.json and the four new error codes don't exist locally, and implementing against the draft TS reference risks silent projection-logic drift if the resolution order changes before chore(main): release 9.5.0 #1815 merges.
code-reviewer: Not ready to execute; same three blockers — missing schema bundle, draft TS reference, and 14 round-trip fixtures that live in adcontextprotocol/adcp and must be vendored into tests/fixtures/canonical/ before item 7 is testable in CI.
My take: All three prerequisites must land in sequence before Python implementation begins: (1) a 3.1+ bundle must be published at adcontextprotocol.org and pulled via scripts/sync_schemas.py, (2) adcp-client #1815 must exit draft so the projection resolution order and advisory contract are stable, and (3) the 14 canonical-format fixtures must be vendored locally. Once those are in place, items 1–5+7 are a cohesive single PR; item 6 (closed-set format_options[] validator) can follow.
Status refresh: the original blocker adcontextprotocol/adcp-client#1815 was closed as superseded, and the TypeScript reference work landed through adcontextprotocol/adcp-client#1882 and #1890. This is no longer blocked on a draft TS prototype. Remaining blocker is local Python 3.1/canonical-format schema/codegen availability plus vendoring the reference fixtures; keep open, but classify as deferred to the 3.1/8.0 enablement batch rather than blocked on #1815.
What changed:adcontextprotocol/adcp-client#1815 was closed as superseded; the TypeScript reference implementation landed through adcp-client#1882 and #1890. The prior "Blocked-on: #1815 — resurfaces on merge" trigger is now void.
Remaining blockers (unchanged from original expert assessment):
Python 3.1 canonical-format schema bundle not yet published/available for scripts/sync_schemas.py — the generated Pydantic models (item 1) can't be produced without it.
14 reference fixtures not yet vendored into tests/fixtures/canonical/ — item 7 (round-trip test) requires them.
My take: Disposition is still deferred, but the linkage shifts from "blocked on draft TS ref" to "blocked on 3.1/8.0 schema-enablement batch." Once the 3.1 schema bundle is published and fixtures are vendored, items 1–5+7 form a cohesive single PR.
Resurfaces when: the 3.1 canonical-format schema bundle is available via scripts/sync_schemas.py.
Activity
bokelley commentedon May 19, 2026
Triage
Classification: Feature request
Bucket(s): client, validation, cross-repo (no matching labels exist in this repo — flagged in run summary)
Status: deferred
What the experts said:
canonical/tree),v1-canonical-mapping.jsonand the four new error codes don't exist locally, and implementing against the draft TS reference risks silent projection-logic drift if the resolution order changes before chore(main): release 9.5.0 #1815 merges.adcontextprotocol/adcpand must be vendored intotests/fixtures/canonical/before item 7 is testable in CI.My take: All three prerequisites must land in sequence before Python implementation begins: (1) a 3.1+ bundle must be published at
adcontextprotocol.organd pulled viascripts/sync_schemas.py, (2) adcp-client #1815 must exit draft so the projection resolution order and advisory contract are stable, and (3) the 14 canonical-format fixtures must be vendored locally. Once those are in place, items 1–5+7 are a cohesive single PR; item 6 (closed-setformat_options[]validator) can follow.Blocked-on: adcontextprotocol/adcp-client#1815 — resurfaces on merge
Triaged by Claude Code. Session: https://claude.ai/code/session_01FdKXuxfrUoexLTJLPdahNH
Generated by Claude Code
revert(preview): remove opt-in v3.1 preview surface (#747)
revert(preview): remove opt-in v3.1 preview surface (#747) (#755)
chore: configure release-please for 6.0.0-beta line (closes #802) (#804)
docs: add MIGRATION_v5_to_v6 stub for the 6.0-beta line (#823)
docs(types): refresh import-architecture references (closes #807) (#828)
bokelley commentedon May 23, 2026
Status refresh: the original blocker adcontextprotocol/adcp-client#1815 was closed as superseded, and the TypeScript reference work landed through adcontextprotocol/adcp-client#1882 and #1890. This is no longer blocked on a draft TS prototype. Remaining blocker is local Python 3.1/canonical-format schema/codegen availability plus vendoring the reference fixtures; keep open, but classify as deferred to the 3.1/8.0 enablement batch rather than blocked on #1815.
bokelley commentedon May 23, 2026
Triage (status update — 2026-05-23)
Classification: Feature request
Bucket(s): client, validation, cross-repo
Status: deferred
What changed: adcontextprotocol/adcp-client#1815 was closed as superseded; the TypeScript reference implementation landed through adcp-client#1882 and #1890. The prior "Blocked-on: #1815 — resurfaces on merge" trigger is now void.
Remaining blockers (unchanged from original expert assessment):
scripts/sync_schemas.py— the generated Pydantic models (item 1) can't be produced without it.tests/fixtures/canonical/— item 7 (round-trip test) requires them.My take: Disposition is still deferred, but the linkage shifts from "blocked on draft TS ref" to "blocked on 3.1/8.0 schema-enablement batch." Once the 3.1 schema bundle is published and fixtures are vendored, items 1–5+7 form a cohesive single PR.
Resurfaces when: the 3.1 canonical-format schema bundle is available via
scripts/sync_schemas.py.Triaged by Claude Code. Session: https://claude.ai/code/cse_01THrmbSwj436SbKV9o2jc6F
Generated by Claude Code
fix(canonical-formats): address expert-review findings on #841
ci: re-trigger workflows on PR #841
fix(canonical-formats): address Argus review follow-ups + nits on #841
fix(canonical-formats): address github-code-quality bot comments on #841
feat(canonical-formats): public API + v2→v1 projection (#741, half 1 …
fix(canonical-formats): address expert review on #845
feat(examples): dual-emit v1 format_ids + v2 format_options on refere…
feat(examples): dual-emit v1 format_ids + v2 format_options on refere…
chore(canonical-formats): polish bundle — rename, typed Divergence, f…
chore(canonical-formats): polish bundle — rename, typed Divergence, f…
feat(canonical-formats): v1↔v2 reverse + pixel_tracker + narrowing (a…