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
{{ message }}
Repository navigation
[Context] Maintain the music business reference model and adoption map #2147
The team compares the proposed model with current Recoup and authorized Seeker schemas as implementation evolves.
Implementation ownership
API owns shared operations and Vercel Workflows; database changes belong in the database repository. Inspector work stays in the inspector; reference documentation stays in private Business.
Acceptance criteria
Keep the living reference in private Business with explicit proposed versus deployed status.
Maintain identity, recording/composition, release, roster/catalog and scoped evidence relationships; map fields and constraints to adoption tickets.
Validate scenarios and migration implications; never treat the full reference schema as an approved production migration.
Link implementation PRs, meaningful tests and inspector traces. Report implemented, connected, fixture-tested and live-verified states separately. Preserve explicit gaps and paid-call authorization.
Follow #2116's company baseline → linked private evidence → representative source batches → saved assessment → one authorized action. The updated private architecture/onboarding plan is in existing Business226.
Before broad import, map existing artist-account, organization-professional and unresolved-name records to consistent business-subject links, preserving account/access identities. Define the minimum person/company, recording/composition/release, identifiers, source-subject, claim and relationship contracts; adopt only those demanded by actual batches. Never require fake ISRCs, companies-as-artists, or agreement rows for unresolved evidence. Inspect research-to-roster insertion behavior and require explicit roster intent. Treat Spotify listings as observations, not all-source business releases. Private intake must not reuse Arweave knowledge-base publication.
Produce a field-by-field mapping of existing/proposed/missing storage and compatibility/access tests before migrations. Full conceptual schema adoption, universal DDEX coverage and a new graph database are deferred. Source links and proposed claims must not silently become accepted rights, mandates or software permissions. Model corrections require paired reference-schema/relationship/scenario changes; the October 9 sequence update itself changes no SQL.
The first bounded reusable product slice is implemented and pushed, not merged/deployed/live-verified:
Database PR89, head ee58fcaab6816c6f5b79598ede7dcdb400242d10: scoped read_context_company_baseline, combining existing artist links, organization professionals and nonwithdrawn Context source metadata without new business tables or row mutations. Independent 50-row pages; existing IDs remain separate. No raw source text/URLs returned. Explicit gaps prevent treating this register as verified legal companies, source attribution, complete catalog or rights.
API PR981, head 5aeb58e1ed3b8f4bddbbb2b8bafa0977b460a0fe: read_company_baseline shared by HTTP and standard Context MCP, actor from authentication, required organization, validated output and owner. No dispatch, paid provider, billing or model dependency.
63 focused API tests across ten files pass, including synthetic HTTP/MCP parity through the actual storage adapter and the real MCP SDK catalog. Full lint/scoped formatting/diff pass. Full TypeScript still has diagnostics outside changed files. Seven real database tests pass on disposable PostgreSQL 15/17 locally; CI 15/16 and Supabase Preview pass. API latest-head CI/preview and both automated reviews are pending at this checkpoint. Preview success does not prove production rollout.
API#980 landed concurrently and is incorporated. Its professional-data exclusion from delegated OAuth is preserved: this combined read contains those same records and remains excluded from that catalog pending the existing organization-grant audit. The exclusion is tested. Do not call standard MCP fixture parity live delegated connector parity.
Business outcome remains pending: no customer intake or company assessment was created in this slice. This is a reusable inventory read, not completion of the company-baseline milestone. Existing organization names and roster links are shown as registered, not newly confirmed identities or representation. Source version counts are not parsing/review status. No production reads or mutations, collection, prices or flags changed.
Next: finish exact-head reviews; after separate explicit release approval, database first, verify migration/permissions, API second, then authenticated Records readback and denial checks. Continue #2147/#2140 with typed existing-identity/source links and supported private evidence intake, followed by representative imports and saved current-state assessment. Keep the epic's business outcome plus reusable capability acceptance; do not return to unrelated billing. The loop remains paused. No merge approval is inferred from this implementation request.
Sidney explicitly restarted this existing ten-minute heartbeat. Its prompt now follows company onboarding and productization, without the superseded default billing task. No merge or spend approval was inferred.
Pushed Database89 at 7fc831b807def727e3f0951c38d02837d5908112 and API981 at 6ab69fe9423972d296c579533d8057c5c3c1d7c0. Both task worktrees are clean. New regressions reproduced invalid cursor acceptance, null marker rejection and duplicate-role acceptance before correction. All 65 focused API tests across ten files, full lint, formatting and diff checks pass; all eight database tests pass on disposable PostgreSQL 15/17. Populated API fixtures now verify artist, professional and source rows through HTTP and standard MCP. Database tests verify exact source isolation and no source-page overlap.
Foreign, missing, wrong-section and withdrawn page markers now reject generically after access checks. Omitted/null markers consistently restart their section. Professional roles must be distinct. Tool instructions explain these rules. All eleven prior review threads are resolved: seven behavioral/test/instruction findings addressed, and four suggestions for alternate GET transport or broad pre-existing file-size refactors deferred with explanations. This bounded operation retains the established shared Context action endpoint and delegated professional-access restriction.
Latest-head database transaction CI and Supabase preview pass; API format/lint/MCP compatibility and Vercel preview pass. API unit CI and both automated reviews remain pending at this checkpoint; old-head green checks are not used for release. No merge/deployment/production read or mutation occurred.
Next batch: check these exact heads and any new review findings; prepare explicit approval when ready. Meanwhile investigate the minimal typed links between existing roster/business subjects and scoped private source versions for #2147/#2140. The complete company baseline, customer imports and saved assessment remain unfinished; do not report this register as verified company/rights data. No unrelated billing refactor, paid calls or duplicate Records intake.
Baseline ready; next attachment boundary mapped — October 9, 17:07 UTC batch
Database89 head 7fc831b807def727e3f0951c38d02837d5908112 and API981 head 6ab69fe9423972d296c579533d8057c5c3c1d7c0 now pass all nine/seven checks respectively, with no unresolved review threads. Explicit approval for these two new PRs has been requested. No merge/deployment/live read has occurred. The active loop can continue independent intake engineering while the approval is pending.
The field-by-field adoption mapping is now in the existing private Business226 at 674ea4bc821ca6aee5fa0372ef2b5b859da292a5, alongside paired relationship/scenario/gap/decision notes. The conceptual SQL is unchanged: this is a physical-adoption boundary over existing IDs, not a new music-business model or production migration. Source inspection pins Database main f446984 and API main 8c2e1e05; it is not a fresh production schema audit.
Next bounded capability: associate an already retained source version with multiple typed existing targets and read those associations through the shared authenticated human/agent surface. Reuse artist-account IDs, scoped professional IDs and owned Context request/subject pairs; keep company/writer inputs unresolved. Reuse Context source/version hashes and composite owner keys. Do not create accepted results merely to simulate attachments, merge people by name, create company workspaces, enroll roster members or infer rights. An association expresses relevance, not parsing/review or legal authority.
Concrete prerequisite fixtures: exact replay/changed payload; namesakes; foreign owner/professional/candidate/music targets; raw global UUID without an authorized target path; version withdrawal; required target-access loss; changed versions preserving old lineage; instructions treated as data; retrieval by a new authorized session. Preserve historical business evidence where an independently authorized history path exists; removing current roster membership is not permission to erase it. Begin fixtures before an additive owning-database migration, and expose the same shared operation to HTTP/MCP before real Records intake.
Current source audit confirms the latest inspected commit_spotify_context replacement still inserts personal and conditional organization roster rows inside its credited-artist loop. Correct that coupling with explicit intent and tests before broad provider import. The selected private attachment path must not call the collector to establish access.
Full file registration/retained-original handling, parser and pointer validation, import/parse/review projections, identifier-less music/composition adoption, private company/creator identity bridges and saved assessment remain separate unfinished work. No customer imports, financial rows, prices, paid flags, provider calls, rights or production business data were changed in this batch.
Linked-evidence slice prepared in Database #90, exact head 19423815547b8d2f21f7da0386d98899d20637e2, branch codex/context-evidence-attachments.
Retained source versions can be associated with multiple existing artist accounts, scoped organization professionals and owned Context request/subject pairs. Concrete foreign keys and owner constraints retain identity boundaries. This is relevance only, not identity/rights/mandate confirmation.
Writes/read/list/replay recheck membership, source retention and independent target access. Canonical target order reuses the receipt; changed input conflicts. Source/target withdrawal withholds protected receipts without deleting history. Owned requests with retained accepted lineage offer a separate historical path after roster termination; roster links themselves cannot bootstrap access.
Eight new transaction tests failed before implementation because the operation was absent. All 13 tests now pass locally on disposable PostgreSQL 15 and 17, including concurrent exact-key retry, atomic rollback, foreign existing records, version history, withdrawal, fresh/revoked actors, paging and role/immutability checks. CI15/16, hosted preview and review are pending. No merge, production migration, customer association, paid call or business mutation occurred.
This is a database prerequisite, not completed onboarding. Next: exact-head review, then shared authenticated HTTP/Context MCP adapters with typed input/output and current access checks. Keep professional-containing operations outside delegated OAuth pending the organization-grant audit. Prepared API worktree has no implementation yet. File/original registration, validated locators, parsing and saved assessment remain separate slices; fix research-to-roster side effects before broad import.
Canonical private plan checkpoint is pushed in existing Business #226 at 9bcb9a73; no customer evidence was published here. Baseline Database #89/API #981 remain open at their previously checked heads and await the already-requested explicit approval; no new approval request or merge was attempted.
Evidence association checkpoint (October 9): Database90 at74f7e3eb59c01ba4a48978e149faf22a2d1be88c and API983 at1c8cfd85794afd414fa394067ba154b0cbb1f24b are open, unapproved and unmerged. Shared authenticated HTTP/standard MCP now expose attach/read/list for exact retained source versions and existing typed authorized targets, with persisted relevance-only receipts. Three individually discoverable tools use concrete schemas; all evidence actions remain excluded from delegated OAuth pending the organization-grant audit.
Database corrections prevent target appends/incomplete receipts, validate fingerprints and add the minimal relationship lock grant. The initial preview-applied migration is preserved byte-for-byte; four forward corrections avoid rewriting migration history. Thirty-one fresh/upgrade transaction test executions pass on disposable PG15/17; CI15/16 and the current Supabase Preview pass. API has64 focused tests, full lint/format/typecheck and188 compatibility tests passing (one skipped). Final API policy wording separates write retries, single reads and paging; fresh-head CI and final reviews remain pending. Validated error/discovery/policy findings are fixed; broad pre-existing dispatcher/delegated-registry and new GET surface refactors are explicitly deferred.
No production migration, deployment or customer import. Linking expresses relevance, not identity confirmation, rights or authority. Next bounded run: inspect exact heads/new findings and prepare separate explicit approval when ready. Source-version discovery, private original registration, deterministic parsing and saved assessment remain required; correct provider research-to-roster side effects before broad imports. Existing baseline89/981 approval remains pending without a repeated request. Canonical private plan updated in existing Business226 atcb59f0111a9899cf071cc025ab06fba698e5ee92.
Final wording review: paging now explicitly maps returned next_id to input after_id;12 SDK/catalog tests pass after this correction. Migration-history warning was checked against identical original/current blob hashes and resolved as inapplicable. All observed review threads are resolved with fixes or documented bounded deferrals; latest-head checks/new reviews still require inspection.
Bounded follow-up: Database90 exact head74f7e3eb59c01ba4a48978e149faf22a2d1be88c now passes all11 checks with no unresolved threads. API983 exact head1c8cfd85794afd414fa394067ba154b0cbb1f24b has format/lint/compatibility/CodeRabbit/Vercel passing; unit tests and Cubic are still running, no unresolved threads at this audit. No approval request or merge/deployment/customer operation.
Read-only code audit narrows the next import guard: latest commit_spotify_context inserts every credited artist into personal and selected organization rosters. The authenticated runner and both guest-completion/adoption paths call it. A transport-only fix would miss guest adoption. Prepare a separate forward-only correction with real transaction regressions: retain credits and owned-request lineage without enrolling artists during research/adoption; preserve existing roster links and use the existing explicit roster operation for intentional enrollment. Include both guest timing orders, retry and revoked scope. No provider dispatch is required. This is branch-source inspection, not fresh production function audit.
Canonical private plan checkpoint: Business226 at67a97736921fa648dc9308999df752205e0d77fe. Source/version discovery and private original registration remain later intake slices.
October9 continuation: Database90 exact74f7e3e and API983 exact1c8cfd85 pass all11/7 checks with no unresolved reviews. Explicit approval has now been requested for these two PRs only. Nothing merged, deployed or customer-ingested; DB-before-API release/verification remains required.
Database91 recoupable/database#91 is open atf0c6a170ccf7c5044cc8fe64f94131d0d294f0b1 on codex/research-roster-isolation. Its forward-only commit_spotify_context replacement removes automatic personal/org roster enrollment while preserving canonical identities, ordered credits, scoped source/version lineage and request output. Both guest adoption orders share the fix. Existing roster links are preserved; later explicit onboarding reuses the researched identity. No new flag or parallel enrollment implementation.
Seven regressions reproduced the bug before the fix;27 transaction tests pass on disposable PG15/17 (nine research/guest/persistence/role plus18 explicit-onboarding regressions). Python compilation/diff checks pass. CI/preview/review are pending; no merge approval for91. No provider calls, hosted business mutation, roster cleanup, financial changes or migration/deployment. Fresh SQL connection readback is persistence evidence, not production HTTP/MCP verification. Canonical private plan updated in Business226 at3c87c6a7ab4a35ac919e42c17d15b3780daac8c8.
Next: inspect91 exact-head reviews/checks, address valid findings, preserve separate pending approvals, and continue source/version discovery/original registration analysis. Before any91 rollout audit expected production predecessor body and then supported authorized behavior.
Partly done. The music-business reference model is kept in the private business workspace with proposed-versus-deployed status, and a field-by-field adoption mapping was drafted there (not yet merged).
Still open: the public adoption-map check against main migrations (recoupable/database#113, unmerged) and keeping the map current as the company baseline and evidence PRs land.
Activity
sidneyswift commentedon Oct 9, 2026
Company onboarding checkpoint — October 9, 2026
The first bounded reusable product slice is implemented and pushed, not merged/deployed/live-verified:
ee58fcaab6816c6f5b79598ede7dcdb400242d10: scopedread_context_company_baseline, combining existing artist links, organization professionals and nonwithdrawn Context source metadata without new business tables or row mutations. Independent 50-row pages; existing IDs remain separate. No raw source text/URLs returned. Explicit gaps prevent treating this register as verified legal companies, source attribution, complete catalog or rights.5aeb58e1ed3b8f4bddbbb2b8bafa0977b460a0fe:read_company_baselineshared by HTTP and standard Context MCP, actor from authentication, required organization, validated output and owner. No dispatch, paid provider, billing or model dependency.API#980 landed concurrently and is incorporated. Its professional-data exclusion from delegated OAuth is preserved: this combined read contains those same records and remains excluded from that catalog pending the existing organization-grant audit. The exclusion is tested. Do not call standard MCP fixture parity live delegated connector parity.
Business outcome remains pending: no customer intake or company assessment was created in this slice. This is a reusable inventory read, not completion of the company-baseline milestone. Existing organization names and roster links are shown as registered, not newly confirmed identities or representation. Source version counts are not parsing/review status. No production reads or mutations, collection, prices or flags changed.
Next: finish exact-head reviews; after separate explicit release approval, database first, verify migration/permissions, API second, then authenticated Records readback and denial checks. Continue #2147/#2140 with typed existing-identity/source links and supported private evidence intake, followed by representative imports and saved current-state assessment. Keep the epic's business outcome plus reusable capability acceptance; do not return to unrelated billing. The loop remains paused. No merge approval is inferred from this implementation request.
sidneyswift commentedon Oct 9, 2026
Resumed loop and review corrections — October 9
Sidney explicitly restarted this existing ten-minute heartbeat. Its prompt now follows company onboarding and productization, without the superseded default billing task. No merge or spend approval was inferred.
Pushed Database89 at
7fc831b807def727e3f0951c38d02837d5908112and API981 at6ab69fe9423972d296c579533d8057c5c3c1d7c0. Both task worktrees are clean. New regressions reproduced invalid cursor acceptance, null marker rejection and duplicate-role acceptance before correction. All 65 focused API tests across ten files, full lint, formatting and diff checks pass; all eight database tests pass on disposable PostgreSQL 15/17. Populated API fixtures now verify artist, professional and source rows through HTTP and standard MCP. Database tests verify exact source isolation and no source-page overlap.Foreign, missing, wrong-section and withdrawn page markers now reject generically after access checks. Omitted/null markers consistently restart their section. Professional roles must be distinct. Tool instructions explain these rules. All eleven prior review threads are resolved: seven behavioral/test/instruction findings addressed, and four suggestions for alternate GET transport or broad pre-existing file-size refactors deferred with explanations. This bounded operation retains the established shared Context action endpoint and delegated professional-access restriction.
Latest-head database transaction CI and Supabase preview pass; API format/lint/MCP compatibility and Vercel preview pass. API unit CI and both automated reviews remain pending at this checkpoint; old-head green checks are not used for release. No merge/deployment/production read or mutation occurred.
Next batch: check these exact heads and any new review findings; prepare explicit approval when ready. Meanwhile investigate the minimal typed links between existing roster/business subjects and scoped private source versions for #2147/#2140. The complete company baseline, customer imports and saved assessment remain unfinished; do not report this register as verified company/rights data. No unrelated billing refactor, paid calls or duplicate Records intake.
sidneyswift commentedon Oct 9, 2026
Baseline ready; next attachment boundary mapped — October 9, 17:07 UTC batch
Database89 head
7fc831b807def727e3f0951c38d02837d5908112and API981 head6ab69fe9423972d296c579533d8057c5c3c1d7c0now pass all nine/seven checks respectively, with no unresolved review threads. Explicit approval for these two new PRs has been requested. No merge/deployment/live read has occurred. The active loop can continue independent intake engineering while the approval is pending.The field-by-field adoption mapping is now in the existing private Business226 at
674ea4bc821ca6aee5fa0372ef2b5b859da292a5, alongside paired relationship/scenario/gap/decision notes. The conceptual SQL is unchanged: this is a physical-adoption boundary over existing IDs, not a new music-business model or production migration. Source inspection pins Database mainf446984and API main8c2e1e05; it is not a fresh production schema audit.Next bounded capability: associate an already retained source version with multiple typed existing targets and read those associations through the shared authenticated human/agent surface. Reuse artist-account IDs, scoped professional IDs and owned Context request/subject pairs; keep company/writer inputs unresolved. Reuse Context source/version hashes and composite owner keys. Do not create accepted results merely to simulate attachments, merge people by name, create company workspaces, enroll roster members or infer rights. An association expresses relevance, not parsing/review or legal authority.
Concrete prerequisite fixtures: exact replay/changed payload; namesakes; foreign owner/professional/candidate/music targets; raw global UUID without an authorized target path; version withdrawal; required target-access loss; changed versions preserving old lineage; instructions treated as data; retrieval by a new authorized session. Preserve historical business evidence where an independently authorized history path exists; removing current roster membership is not permission to erase it. Begin fixtures before an additive owning-database migration, and expose the same shared operation to HTTP/MCP before real Records intake.
Current source audit confirms the latest inspected
commit_spotify_contextreplacement still inserts personal and conditional organization roster rows inside its credited-artist loop. Correct that coupling with explicit intent and tests before broad provider import. The selected private attachment path must not call the collector to establish access.Full file registration/retained-original handling, parser and pointer validation, import/parse/review projections, identifier-less music/composition adoption, private company/creator identity bridges and saved assessment remain separate unfinished work. No customer imports, financial rows, prices, paid flags, provider calls, rights or production business data were changed in this batch.
sidneyswift commentedon Oct 9, 2026
Linked-evidence slice prepared in Database #90, exact head
19423815547b8d2f21f7da0386d98899d20637e2, branchcodex/context-evidence-attachments.Canonical private plan checkpoint is pushed in existing Business #226 at
9bcb9a73; no customer evidence was published here. Baseline Database #89/API #981 remain open at their previously checked heads and await the already-requested explicit approval; no new approval request or merge was attempted.sidneyswift commentedon Oct 9, 2026
Evidence association checkpoint (October 9): Database90 at74f7e3eb59c01ba4a48978e149faf22a2d1be88c and API983 at1c8cfd85794afd414fa394067ba154b0cbb1f24b are open, unapproved and unmerged. Shared authenticated HTTP/standard MCP now expose attach/read/list for exact retained source versions and existing typed authorized targets, with persisted relevance-only receipts. Three individually discoverable tools use concrete schemas; all evidence actions remain excluded from delegated OAuth pending the organization-grant audit.
Database corrections prevent target appends/incomplete receipts, validate fingerprints and add the minimal relationship lock grant. The initial preview-applied migration is preserved byte-for-byte; four forward corrections avoid rewriting migration history. Thirty-one fresh/upgrade transaction test executions pass on disposable PG15/17; CI15/16 and the current Supabase Preview pass. API has64 focused tests, full lint/format/typecheck and188 compatibility tests passing (one skipped). Final API policy wording separates write retries, single reads and paging; fresh-head CI and final reviews remain pending. Validated error/discovery/policy findings are fixed; broad pre-existing dispatcher/delegated-registry and new GET surface refactors are explicitly deferred.
No production migration, deployment or customer import. Linking expresses relevance, not identity confirmation, rights or authority. Next bounded run: inspect exact heads/new findings and prepare separate explicit approval when ready. Source-version discovery, private original registration, deterministic parsing and saved assessment remain required; correct provider research-to-roster side effects before broad imports. Existing baseline89/981 approval remains pending without a repeated request. Canonical private plan updated in existing Business226 atcb59f0111a9899cf071cc025ab06fba698e5ee92.
PRs: recoupable/database#90 and recoupable/api#983.
Final wording review: paging now explicitly maps returned next_id to input after_id;12 SDK/catalog tests pass after this correction. Migration-history warning was checked against identical original/current blob hashes and resolved as inapplicable. All observed review threads are resolved with fixes or documented bounded deferrals; latest-head checks/new reviews still require inspection.
sidneyswift commentedon Oct 9, 2026
Bounded follow-up: Database90 exact head74f7e3eb59c01ba4a48978e149faf22a2d1be88c now passes all11 checks with no unresolved threads. API983 exact head1c8cfd85794afd414fa394067ba154b0cbb1f24b has format/lint/compatibility/CodeRabbit/Vercel passing; unit tests and Cubic are still running, no unresolved threads at this audit. No approval request or merge/deployment/customer operation.
Read-only code audit narrows the next import guard: latest commit_spotify_context inserts every credited artist into personal and selected organization rosters. The authenticated runner and both guest-completion/adoption paths call it. A transport-only fix would miss guest adoption. Prepare a separate forward-only correction with real transaction regressions: retain credits and owned-request lineage without enrolling artists during research/adoption; preserve existing roster links and use the existing explicit roster operation for intentional enrollment. Include both guest timing orders, retry and revoked scope. No provider dispatch is required. This is branch-source inspection, not fresh production function audit.
Canonical private plan checkpoint: Business226 at67a97736921fa648dc9308999df752205e0d77fe. Source/version discovery and private original registration remain later intake slices.
sidneyswift commentedon Oct 9, 2026
October9 continuation: Database90 exact74f7e3e and API983 exact1c8cfd85 pass all11/7 checks with no unresolved reviews. Explicit approval has now been requested for these two PRs only. Nothing merged, deployed or customer-ingested; DB-before-API release/verification remains required.
Database91 recoupable/database#91 is open atf0c6a170ccf7c5044cc8fe64f94131d0d294f0b1 on codex/research-roster-isolation. Its forward-only commit_spotify_context replacement removes automatic personal/org roster enrollment while preserving canonical identities, ordered credits, scoped source/version lineage and request output. Both guest adoption orders share the fix. Existing roster links are preserved; later explicit onboarding reuses the researched identity. No new flag or parallel enrollment implementation.
Seven regressions reproduced the bug before the fix;27 transaction tests pass on disposable PG15/17 (nine research/guest/persistence/role plus18 explicit-onboarding regressions). Python compilation/diff checks pass. CI/preview/review are pending; no merge approval for91. No provider calls, hosted business mutation, roster cleanup, financial changes or migration/deployment. Fresh SQL connection readback is persistence evidence, not production HTTP/MCP verification. Canonical private plan updated in Business226 at3c87c6a7ab4a35ac919e42c17d15b3780daac8c8.
Next: inspect91 exact-head reviews/checks, address valid findings, preserve separate pending approvals, and continue source/version discovery/original registration analysis. Before any91 rollout audit expected production predecessor body and then supported authorized behavior.
sidneyswift commentedon Nov 6, 2026
Partly done. The music-business reference model is kept in the private business workspace with proposed-versus-deployed status, and a field-by-field adoption mapping was drafted there (not yet merged).
Still open: the public adoption-map check against main migrations (recoupable/database#113, unmerged) and keeping the map current as the company baseline and evidence PRs land.