chore: merge main into next through 9.6.1

.release-please-manifest.json

@@ -1,4 +1,4 @@
 {
-  ".": "9.3.0",
+  ".": "9.6.1",
   "packages/adcp-reporting-storage": "0.1.0-beta.4"
 }

.traycer/environment.json

@@ -0,0 +1,15 @@
+{
+  "setup": {
+    "default": "ROOT=\"$(cd \"$(git rev-parse --git-common-dir)/..\" && pwd)\" || exit 1\nfor f in \"$ROOT\"/.env*; do n=\"${f##*/}\"; [ -f \"$f\" ] && [ ! -e \"$n\" ] && git check-ignore -q \"$n\" && cp \"$f\" \"$n\"; done; true || exit 1\nuv sync --all-extras || exit 1\nmake bootstrap || exit 1",
+    "macos": null,
+    "linux": null,
+    "windows": null
+  },
+  "teardown": {
+    "default": "",
+    "macos": null,
+    "linux": null,
+    "windows": null
+  },
+  "updatedAt": 1793934266991
+}

AGENTS.md

@@ -272,6 +272,19 @@ from adcp.testing import build_asgi_app, build_test_client, make_request_context
 
 ## Property Lists and Identifier Resolution
 
+For publisher manifest sync, resolve the agent's selectors with
+`get_properties_by_agent` or `resolve_properties_for_agent`, then call
+`properties_for_publisher_domain(properties, publisher_domain)` to attribute
+inventory. It keeps publisher subdomains and wildcard domain inventory, and
+excludes domainless inventory (bundle-only apps, venues) unless
+`include_domainless=True`. Permissive resolution requires `publisher_domain=`
+from the trusted fetch origin (`allow_any_domain=True` skips the gate for
+diagnostics). Use `await verify_cross_domain_properties(properties,
+source_domain=..., agent_url=...)` before trusting claims about another
+publisher's domain; its `CrossDomainVerification` result separates
+`verified`, `unverified` and `domainless` properties. All three helpers are
+exported from `adcp`. See `docs/adagents-discovery.md`.
+
 Import `MCPPropertyListFetcher`, `ProductPropertyPolicy`, and
 `check_property_list_reference` from `adcp`. The default seller fetcher resolves
 lists with MCP `get_property_list`, follows pagination, and caches complete

CHANGELOG.md

@@ -236,6 +236,75 @@
   change public consumer defaults. See
   [format-kind migration](docs/canonical-format-kinds-migration.md).
 
+## [9.6.1](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.6.0...v9.6.1) (2026-11-03)
+
+
+### Bug Fixes
+
+* **reporting:** isolate handler registrations for each installation ([#1847](https://github.com/adcontextprotocol/adcp-client-python/issues/1847)) ([86302d0](https://github.com/adcontextprotocol/adcp-client-python/commit/86302d09404067410a60d3f9d0f850790038a36f))
+
+## [9.6.0](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.5.1...v9.6.0) (2026-11-03)
+
+
+### Features
+
+* **reporting-storage:** accept 255 character row binding IDs and credential refs ([#1828](https://github.com/adcontextprotocol/adcp-client-python/issues/1828)) ([6bb34df](https://github.com/adcontextprotocol/adcp-client-python/commit/6bb34df00821e11f03aed2a018a0c5a6ccab3cdf))
+* **reporting-storage:** collect orphaned row objects by listing with a clone fence ([#1835](https://github.com/adcontextprotocol/adcp-client-python/issues/1835)) ([e004f1b](https://github.com/adcontextprotocol/adcp-client-python/commit/e004f1bc55eff3bc32065a9c785f451af3813f79))
+* **reporting-storage:** order the account change feed by xid and seq ([#1827](https://github.com/adcontextprotocol/adcp-client-python/issues/1827)) ([4449bc9](https://github.com/adcontextprotocol/adcp-client-python/commit/4449bc94cdd47900f39e15f5e99a35a9699d0c30))
+
+
+### Bug Fixes
+
+* **server:** preserve valid creatives in lenient legacy batches ([#1841](https://github.com/adcontextprotocol/adcp-client-python/issues/1841)) ([2bd7751](https://github.com/adcontextprotocol/adcp-client-python/commit/2bd7751cb1312efda5aefc4a11fecdfab0933c9e))
+
+
+### Performance Improvements
+
+* **reporting-storage:** drop the per-object advisory lock where nothing pairs on it ([#1836](https://github.com/adcontextprotocol/adcp-client-python/issues/1836)) ([536b8f8](https://github.com/adcontextprotocol/adcp-client-python/commit/536b8f870f518e7ceab3c7a1d64aa0b229b75e8e))
+
+## [9.5.1](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.5.0...v9.5.1) (2026-11-03)
+
+
+### Bug Fixes
+
+* **ci:** check PR titles without stripping a trailing issue reference ([#1819](https://github.com/adcontextprotocol/adcp-client-python/issues/1819)) ([495593e](https://github.com/adcontextprotocol/adcp-client-python/commit/495593ec319bca75b22174c582fa5417f9730860))
+
+## [9.5.0](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.4.0...v9.5.0) (2026-11-02)
+
+
+### Features
+
+* **decisioning:** let a raised AdcpError reject a request with every error ([#1780](https://github.com/adcontextprotocol/adcp-client-python/issues/1780)) ([#1810](https://github.com/adcontextprotocol/adcp-client-python/issues/1810)) ([272fced](https://github.com/adcontextprotocol/adcp-client-python/commit/272fced52d57632dedf32f2c175cf294100bff63))
+
+
+### Bug Fixes
+
+* **canonical-formats:** check narrowing on seller-asserted references ([#1758](https://github.com/adcontextprotocol/adcp-client-python/issues/1758)) ([617762e](https://github.com/adcontextprotocol/adcp-client-python/commit/617762e94371b5817f400481bbbc866c760a6a1a))
+* **canonical-formats:** read versions from typed legacy formats ([#1757](https://github.com/adcontextprotocol/adcp-client-python/issues/1757)) ([6714203](https://github.com/adcontextprotocol/adcp-client-python/commit/67142036f40e6a68fd9ca3d5c51acaa7bfe402fb))
+* **server:** return task and workflow handoffs uncached from IdempotencyStore.wrap ([#1795](https://github.com/adcontextprotocol/adcp-client-python/issues/1795)) ([126f905](https://github.com/adcontextprotocol/adcp-client-python/commit/126f9052ab79399f1e7bab8a53f34e24999ef95e)), closes [#1781](https://github.com/adcontextprotocol/adcp-client-python/issues/1781)
+* **types:** make the domain package roots lazy so a types package imports first in any order ([#1798](https://github.com/adcontextprotocol/adcp-client-python/issues/1798)) ([7447d00](https://github.com/adcontextprotocol/adcp-client-python/commit/7447d00ceba97490109c46f0cf2f0ca9d6f8e5f3))
+
+## [9.4.0](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.3.1...v9.4.0) (2026-10-31)
+
+
+### Features
+
+* **adagents:** attribute and verify publisher property claims ([#1719](https://github.com/adcontextprotocol/adcp-client-python/issues/1719)) ([0c0ef92](https://github.com/adcontextprotocol/adcp-client-python/commit/0c0ef92515ec5dfe9d1f045b976194d3d245d4e5))
+* **server:** seller helper for wholesale feed versions ([#1718](https://github.com/adcontextprotocol/adcp-client-python/issues/1718)) ([7088ebf](https://github.com/adcontextprotocol/adcp-client-python/commit/7088ebf909852134521ba0ed4796b6f2896a4ad7)), closes [#1676](https://github.com/adcontextprotocol/adcp-client-python/issues/1676)
+* **signing:** accept any well-formed adcp_use purpose in keygen ([#1710](https://github.com/adcontextprotocol/adcp-client-python/issues/1710)) ([787cfb2](https://github.com/adcontextprotocol/adcp-client-python/commit/787cfb215cf9207bfef8c28cc34ff0b737cea39b))
+
+
+### Bug Fixes
+
+* **types:** defer the import-time model rebuilds that cost adcp.client 560 MiB RSS ([#1704](https://github.com/adcontextprotocol/adcp-client-python/issues/1704)) ([aac52ca](https://github.com/adcontextprotocol/adcp-client-python/commit/aac52cac66ecae2d1c2c4e11be23294335a3edd8)), closes [#1694](https://github.com/adcontextprotocol/adcp-client-python/issues/1694)
+
+## [9.3.1](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.3.0...v9.3.1) (2026-10-29)
+
+
+### Bug Fixes
+
+* **server:** compact a pinned seller's tools/list so an SDK 10 buyer can call it ([#1702](https://github.com/adcontextprotocol/adcp-client-python/issues/1702)) ([cc11643](https://github.com/adcontextprotocol/adcp-client-python/commit/cc11643352a078b977449f44dcd2d5687ad29bf6)), closes [#1692](https://github.com/adcontextprotocol/adcp-client-python/issues/1692)
+
 ## [9.3.0](https://github.com/adcontextprotocol/adcp-client-python/compare/v9.2.0...v9.3.0) (2026-10-27)
 
 

docs/reporting-ledger-migration.md

@@ -368,3 +368,35 @@ seeds ledgers with the actual published wheel pairs (`adcp==9.3.1` with
 `adcp-reporting-storage==0.1.0b4`, `adcp==10.0.0b2` with `0.2.0b0`) and proves
 the retained evidence reads back unchanged, a restart rebinds the stored
 generations, and the upgraded catalog is identical to a fresh migration.
+
+## Upgrading a ledger that adcp-reporting-storage 0.1.0b5 wrote
+
+0.1.0b5 is the SDK 9 storage release that orders the account change feed by
+`(xid, seq)`, accepts 255-character row binding IDs and credential refs, and
+adds the opt-in orphaned-object collector. Its ledger upgrades to this release
+through the same two entry points and the same numbered migrations as a
+0.1.0b4 ledger (above): this release adds no migration for them. Their DDL is
+in the files `create_schema()` re-applies on every start when the option is
+on (`reporting_changes.sql` with `change_feed=True`,
+`reporting_row_storage.sql` with `row_storage`), as on the SDK 9 line, and
+`python -m adcp migrate` never creates those tables. A ledger created before
+0.1.0b5 gains the account horizon's xid columns, the named feed-consumer and
+row-binding constraints, the `(account_id, xid, seq)` index and the orphan
+collector fence table the first time this release starts with those options;
+see [host change feed and consumers](reporting-row-storage.md#host-change-feed-and-consumers)
+for creating that index concurrently on a large changes table.
+
+adcp-reporting-storage 0.2.0b0 to 0.2.0b2 predate the `(xid, seq)` account
+feed. On a 0.1.0b5 ledger they refuse an account cursor that carries an
+`xid`, refuse a row binding ID over 128 characters, and prune the change feed
+without advancing the account horizon's xid columns. A cursor 0.1.0b5 handed
+out could then pass the account horizon check and skip what 0.2.x pruned. So
+the first time this release starts with `change_feed=True` on a 0.1.0b5 ledger
+that a 0.2 release has already upgraded, it raises each account horizon to the
+deployment horizon, or to just below the account's oldest retained change when
+that is lower. Those cursors then fail closed with `CURSOR_EXPIRED`; a
+consumer starting over still reads every retained change. It runs once (a
+comment on `adcp_reporting_change_account_horizon` records it) and changes
+nothing on a ledger upgraded straight from 0.1.0b5. As for the `(xid, seq)`
+upgrade on SDK 9, stop every older host's change maintenance before this
+release starts.

docs/reporting-release-notes.md

@@ -76,6 +76,23 @@ restart across the upgrade does not turn a replay into a conflict. This is part
 the shared layout in adcontextprotocol/adcp#7996 (Stage 2, tracked in
 [#1457](https://github.com/adcontextprotocol/adcp-client-python/issues/1457)).
 
+## Account change feed orders by `(xid, seq)`
+
+The account change feed (`changes_after(account_id=...)`) now orders by `(xid, seq)` behind
+`xid < pg_snapshot_xmin(pg_current_snapshot())`, as the deployment-wide feed already did, matching the JavaScript SDK
+([adcp-client#3289](https://github.com/adcontextprotocol/adcp-client/pull/3289)). It replaces the page truncation at
+the first in-flight change from #1483, and no longer depends on the account lock for its order: a lower `seq` that
+commits late is returned on a later page, never skipped. Account cursors now carry `xid`. A seq-only account cursor
+saved by an earlier release is still accepted and mapped to a position that skips nothing, at the cost of redelivering
+a few changes, and it still expires exactly (`pruned_through_seq` is kept). `reporting_changes.sql` is re-applied by
+`create_schema` and is idempotent: it adds the account horizon's xid columns, replaces the consumer check constraint,
+and creates `adcp_reporting_changes_account_xid`. Create that index with `CREATE INDEX CONCURRENTLY` first on a large
+changes table, stop older hosts' change maintenance before the new reader runs, and do not roll back after a consumer
+saved a new-format account cursor. Not affected: deployment-wide cursors, the buyer-facing `changes_after` projection and
+anyone who never enabled `change_feed`. See
+[host change feed and consumers](reporting-row-storage.md#host-change-feed-and-consumers)
+([#1825](https://github.com/adcontextprotocol/adcp-client-python/issues/1825)).
+
 ## Feed consumer checkpoints are forward-only
 
 `save_feed_consumer_cursor` used to replace the saved position unconditionally. It now ignores a cursor behind the
@@ -94,6 +111,56 @@ Reads handle both forms. `migrate_inline_rows` moves legacy rows with verificati
 rows do not reproduce its binding, and `retire_legacy_inline_rows` then renames the legacy table so older binaries fail
 closed. See [row storage](reporting-row-storage.md).
 
+## The per-object lock is kept only where it pairs
+
+Retirement, the intent sweep and a commit that adopts an object its own intent recorded no longer take the per-object
+advisory lock, and no longer hold a transaction or pooled connection across the provider delete. The lock stays in the
+two places that pair with each other: the orphan collector's delete, which locks every binding's key for the object, and
+a commit that adopts an object its intent did not record. The lock text is unchanged, so a Python deleter and a
+JavaScript adopter still agree, and rolling upgrades between a version that locks everywhere and this one are safe (every
+pair that needs exclusion keeps the lock on both sides). A commit now closes its intent with `RETURNING created_objects`
+to know which adopted objects it recorded. No schema or wire change. See
+[the object lock contract](reporting-row-storage.md#the-object-lock-contract)
+([#1492](https://github.com/adcontextprotocol/adcp-client-python/issues/1492), adcp-client#3606).
+
+## Opt-in orphaned-object collector, fenced against clones
+
+`PgReportingLedgerStore(orphan_collection=ReportingOrphanCollectionOptionsV1(...))` and
+`collect_orphaned_row_objects` list each object binding under its owned prefix and delete the objects no ledger record
+or open upload names, which the write-intent sweep cannot see (an upload that succeeded before its intent recorded it,
+intents for a binding no host configures, intents lost to a restore). It mirrors the JavaScript SDK
+([adcp-client#3291](https://github.com/adcontextprotocol/adcp-client/pull/3291) and the clone fence in
+[#3593](https://github.com/adcontextprotocol/adcp-client/pull/3593)). Nothing changes unless you opt in. An object is
+deleted only when it is a well-formed row key under `{prefix}/{namespace_key}/`, older than `min_age_seconds` (default
+7 days, minimum 1 day), named by no record or upload at any version under any binding, and unreferenced under the
+per-object lock of every binding. A real pass first claims the namespace in a create-only fence in the bucket and
+refuses to delete while another database holds a newer claim, so a clone or restore of the ledger database that shares
+its namespace key and its opt-in cannot delete the live database's newer objects; `take_over_orphan_collector_fence`
+is the deliberate step after a restore.
+
+Providers gain an optional `list`. The provider conformance suite checks it when present, so its result now ends with a
+`list` entry for providers that implement it; compare accordingly. A provider without `list` keeps working and is
+reported `unsupported`. `create_schema` adds the shared `adcp_reporting_orphan_collector_fences` table, additively and
+outside the required-schema manifest. The production worker reports `fenced` bindings as stalled work and `unsupported`
+bindings as pending work. **Roll out:** start with `dry_run=True`, claim from the live database first, and never enable
+the collector on a copy of the deployment. See [the listing collector](reporting-row-storage.md#orphaned-objects-the-opt-in-listing-collector)
+([#1829](https://github.com/adcontextprotocol/adcp-client-python/issues/1829)).
+
+## Row binding IDs and credential references accept 255 characters
+
+`row_binding_id` and a binding's `credential_ref` were capped at 128 characters. Every comparable AdCP identifier
+(`reporting_revision_id`, `resource_ref`, `destination_ref`) allows 255, so an ID derived from an AdCP destination
+reference could be refused for no protocol reason. Both now accept 1 to 255 characters of `A-Za-z0-9_.:-`, matching the
+JavaScript SDK ([adcp-client#3595](https://github.com/adcontextprotocol/adcp-client/pull/3595)). The configuration
+check, the `adcp_reporting_row_bindings` `CHECK` and the provider `clients` map share the bound, which is exported as
+`REPORTING_ROW_BINDING_ID_MAX_LENGTH` and `REPORTING_ROW_CREDENTIAL_REF_MAX_LENGTH`. A malformed value that the old
+check let through (a trailing newline) is now refused too. `create_schema` replaces the original unnamed 128-character
+`CHECK` with `adcp_reporting_row_bindings_row_binding_id_bounds` in place, idempotently, with no data migration; stop
+older writers first, because an older binary requires the original constraint and fails closed against an upgraded
+database. Nothing that was valid before is refused. See
+[stored names follow AdCP](reporting-row-storage.md#stored-names-follow-adcp)
+([#1498](https://github.com/adcontextprotocol/adcp-client-python/issues/1498)).
+
 ## Row storage field alignment (breaking for row storage adopters)
 
 Row storage stores every value in the format AdCP defines for it, matching the JavaScript SDK byte for byte: