Deploying the three-tier series onto OHM crash-looped on migration 029:
`NOT NULL constraint failed: cached_branches__new.collection_id`, then a UNIQUE
collision. Root cause: the v0.39.0 default→ohm re-stamp updated
cached_rfcs.project_id but NOT the entry-satellite tables, so ~1.3k
cached_branches rows were stranded at project_id='default' — which 029's
per-project collection backfill can't map (NULL), some of which also duplicate
freshly-re-mirrored 'ohm' rows (UNIQUE), and some of which reference entries
that no longer exist.
Fix — a repair prologue at the top of 029 (no schema change):
- drop stale rows that duplicate an already-correctly-stamped row (keep the
fresh copy) for the branch-keyed tables;
- re-derive each satellite's project_id from its entry (cached_rfcs, by slug);
- drop rows whose entry no longer exists (stale cache, rebuildable from gitea).
A no-op on clean/fresh deployments (empty or consistent satellites).
Validated against a snapshot of the live OHM DB: 029→032 apply cleanly, zero
NULL collection_id, FK check clean, watches/RFCs/branch_visibility preserved,
cached_branches 1397 → 1291 (−67 no-RFC, −39 dups). Fresh-install path
unchanged: existing 029 suite green + a new regression test for the
stale/dup/orphan shape. Full backend suite 547 passed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ships the request side of joining a gated scope, completing the §22.8 pair
(S4 shipped the invite half). A user who knows a project/collection exists
asks to join it naming a desired role; the request fans out to that scope's
Owners across the subtree (the cross-collection inbox, §22.11), who accept
(writing the memberships row via memberships.grant) or decline. Built by
analogy to §28 contribution_requests + the S4 memberships surface.
Backend
- migration 032: join_requests (scope_type ∈ {project,collection}, scope_id,
requester, requested_role, message, status, granted_role); one-open-per
(scope, requester) partial unique index. Additive — no rebuild.
- api_join_requests.py: GET join-target / POST join-requests / POST
{id}/accept / {id}/decline under /api/scopes/{scope_type}/{scope_id}/.
Accept grants via memberships.grant; the request POST does not require the
scope be readable (that is how one joins a gated scope).
- notify: fan_out_join_request (subtree-Owner enumeration via
_scope_owner_user_ids), notify_join_decided, 3 render_summary cases.
- auth.effective_role_at_scope — scope-grain twin of effective_scope_role,
folding global → project for a project target.
- api_collections: viewer.can_request_join on the project + collection blocks.
Frontend
- api.js join verbs; JoinRequestModal; "Request to join" affordance in the
collection directory + catalog footer; JoinRequestRow in the inbox.
Tests: backend test_join_requests_vertical (11) + test_migration_032 (5);
frontend api.joinrequests + CollectionDirectory cases. 546 backend / 36
frontend green.
Per docs/design/2026-06-05-three-tier-projects-collections.md Part E (S6) and
SPEC.md §22.8 / §22.11. Closes the request-to-join item flagged open at
0.45.0; per-type surfaces (§22.4a items 1 & 3) remain the last S6 item.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The release wrap for the S6 scope shipped this slice (SPEC merge, per-collection
enabled_models, type-driven entry noun).
- test_s6_two_project_multicollection.py: an integrative pass proving the new
per-collection knobs (§22.12 enabled_models, §22.4a noun) resolve independently
per collection across two projects with no cross-project/-collection bleed.
- CHANGELOG.md 0.45.0: the §20.4 entry + upgrade-steps (migration 031 is
additive + automatic; per-collection enabled_models and typed collections are
optional). Notes the two OPEN S6 items carried to a follow-up slice (per-type
surfaces; request-to-join + cross-collection inbox).
- VERSION + frontend/package.json → 0.45.0.
Gate: backend 530 passed, frontend 30 passed, build green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The displayed noun for an entry is a framework concept keyed on the collection's
immutable type — document→"RFC", specification→"Spec", bdd→"Feature" — not
deployment content. The chrome reads it from the API instead of hardcoding "RFC".
- collections.ENTRY_NOUN + entry_noun(type) (unknown type → generic "RFC").
- Surfaced as entry_noun on get_collection, list_collections items, the
/api/deployment directory items, and GET /api/projects/:id.
- Frontend: Catalog reads entry_noun for the "+ Propose New <noun>" control;
ProposeModal fetches the active collection's noun for its title + field copy.
- test_s6_entry_noun_vertical.py: map + API surfacing (collection + directory).
Scope note: this is §22.4a item (2) — terminology. The deeper type-module work
(item 1 per-type frontmatter schemas; item 3 specification release-planning +
bdd scenario/coverage surfaces) is flagged OPEN in the design doc and is handed
off to a follow-up spec+slice.
Backend 528 passed; frontend 30 passed + build green. Part of v0.45.0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A collection's .collection.yaml may carry an enabled_models list that NARROWS
its project's universe, which narrows the deployment ENABLED_MODELS. Resolution
(extending §6.6/§6.7): funder ∩ per-entry models ∩ collection ∩ project, with
the operator providers as the ceiling.
- migration 031: additive collections.config_json (parallels projects.config_json).
- registry.parse_collection_manifest: read enabled_models into CollectionEntry.config
(absent = inherit, present incl. [] = narrow; [] opts the collection out of AI);
reject a non-list. _upsert_named_collection persists config_json. The default
collection (from projects.yaml) leaves it NULL — it inherits the project.
- models_resolver: _scope_narrowed_universe narrows the operator universe by the
entry's project then collection enabled_models; the funder universe is bounded
by it too. A collection cannot widen its project (narrowing from the ceiling).
- collections.get_collection surfaces enabled_models; GET
/api/projects/:id/collections/:cid returns it.
- test_s6_collection_models_vertical.py: 9 cases (parse, the narrowing chain
incl. opt-out + cannot-widen, API surfacing).
Backend: 525 passed. Per SPEC §22.12. Releases as part of v0.45.0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the global-Owner create-project action and the role-aware deployment
directory, completing slice S5 of the three-tier refactor (release v0.44.0,
minor/non-breaking — no migration). Per
docs/design/2026-06-05-three-tier-projects-collections.md Part E (S5),
Part C.3 (C3.1–C3.2).
Backend:
- auth.can_create_project — global-Owner gate (deployment owner/admin or an
explicit scope_type='global' Owner grant).
- bot.create_project — provision the Gitea content repo (seed README so main
exists), read+append+commit projects.yaml in the registry repo, audit-log
(create_project). The bot stays the only git writer (§1).
- POST /api/projects (api_deployment) — global-Owner gated; validates the id
(slug, not 'default'), name, type, visibility, content_repo; commits via the
bot, then re-mirrors the registry so projects + default collection rows flow
from git (§22.2). GET /api/deployment gains viewer.can_create_project and
default_project_readable. make_router now takes gitea + bot.
Frontend:
- api.createProject; DeploymentProvider surfaces viewer + defaultProjectReadable
+ refresh; DeploymentLanding redirects into the default only when readable
(gated/absent default falls through to the directory, no 404 bounce).
- Directory.jsx role-aware empty states (C3.1 "Create your first project" CTA;
C3.2 "Nothing has been shared with you yet") + Owner-only "New project"
control + CreateProjectModal.
Tests: backend test_create_project_vertical.py (vertical + gates + the
deployment empty-state signals); frontend Directory.test.jsx empty-state cases.
Also: ignore the session-local .superpowers/ tooling dir.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Owner-only scope-role grant surface (Part E S4 / Part C.2). An Owner
grants {owner, contributor} at a scope their reach covers — the project or
one collection within it — to an existing account looked up by email; the
grant writes a `memberships` row immediately and §15-notifies the grantee
(direct grant, no accept round-trip — the C.2 scenarios name existing
accounts and write the row directly).
- auth.can_invite_at_project / can_invite_at_collection — the Owner-reach
invite gates (is_project_superuser / is_collection_superuser).
- memberships.py — grant (with the C.2.6 broader-supersedes-narrower prune,
preserving a stronger child grant — no negative override), revoke, list,
user_by_email.
- api_memberships.py — GET/POST/DELETE /api/projects/:id/members, the single
POST keying on optional collection_id so the invite UI's one control maps
to one endpoint; reach bounded by the inviter's Owner reach (C.2.3);
contributors refused (C.2.4); a pending grantee's row is recorded but
confers no write (C.2.7, the §6 floor).
- notify.notify_scope_role_granted + render_summary — the §15 personal-direct
notification naming the project and role.
- api_collections — surface viewer capabilities (can_create_collection,
can_invite, can_contribute, role) on the project/collection GETs to drive
the C.3 role-aware empty states.
Acceptance: test_s4_invitations_vertical.py covers C.2.1–C.2.7 over the HTTP
surface + resolver, plus the C.3 (@S4) capability flags. Full suite green
(504 passed).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implement slice S3 of the §22 three-tier refactor: the four-layer
most-permissive scope-role resolver (§B.2) over {owner, contributor}
grants at {global, project, collection}, with the §22.5 visibility gate
enforced at the collection grain.
- migration 030: memberships.scope_type += 'global' (the global RFC
Contributor tier; sentinel scope_id '*').
- auth.effective_scope_role folds global → project → collection,
most-permissive, no negative override; can_read_collection /
can_contribute_in_collection / is_collection_superuser /
can_create_collection gate reads, writes, admin, and create.
- collection-grain visibility: a gated collection is hidden from the
public (404, omitted from the directory) yet visible+listed for a
scope-role holder; a collection may be set only as strict or stricter
than its project (public < unlisted < gated), validated at create and
clamped at the mirror.
- entry-scoped authority (mark-reviewed, graduate, branch read/contribute,
PR/discussion/contribution moderation) re-pointed from the project grain
to the entry's collection.
- create-collection authority widened to a project/global-scope grant
holder (§B.1), not only a deployment owner/admin.
Keystone reconciliation (session 0076): a plain granted account is a
granted *account*, not a write-everywhere global role; the implicit-public
write baseline is grandfathered onto the migration-seeded `default`
collection only, so the N=1 deployment loses no capability. Reinterprets
§B.1/§B.3 literally — flagged for the SPEC merge (S6).
Completes @S3 (C1.1–C1.8). Tests: test_s3_scope_roles_vertical.py (8 C.1
scenarios + visibility/strictness), test_migration_030_global_scope.py.
Full backend suite 493 passed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Anonymous reader of an empty public collection gets a 200 empty catalog and no
propose action (the propose route rejects anonymous); the Catalog footer's
'Sign in to propose' prompt is the existing anonymous affordance.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
POST /api/projects/<id>/collections, owner/admin-gated, commits a
.collection.yaml to the content repo main via bot.create_collection, then
re-mirrors the registry so the collections row appears (§22.2). Adds GET
list/one collection routes. Extends FakeGitea to model directory listings so
the mirror's content-repo walk is exercised end to end.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Refactor the project-scoped serve/propose internals into collection-grained
helpers (_list_rfcs_for_collection, _get_rfc_for_collection,
_propose_into_collection); add routes under
/api/projects/<id>/collections/<cid>/rfcs[/<slug>|/propose]. Propose writes
the entry under the target collection's <subfolder>/rfcs via a new rfcs_dir
param on bot.open_idea_pr. Default-collection routes preserved as wrappers.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
refresh_meta_repo now iterates a project's collections and keys cached_rfcs by
collection_id; the default collection (subfolder '') keeps the shipped rfcs/
root path. N=1 default path unchanged (469 green).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Discover named collections by walking each project's content-repo root for
<subdir>/.collection.yaml; parse + upsert with immutable-type enforcement
(§22.4a) and project-visibility inheritance. The default collection still
flows from projects.yaml.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
projects.restamp_default_project(config): at startup after the registry mirror,
renames project_id from the M1 bootstrap 'default' to the configured
DEFAULT_PROJECT_ID across every project-scoped table (discovered by column) and
drops the stale 'default' projects row, so a deployment's original corpus lands
at a meaningful /p/<id>/ and 'default' is never a public URL. FK off for the
rename (parent+children move together) + foreign_key_check backstop. Idempotent;
no-op unless DEFAULT_PROJECT_ID is set to a non-'default' value.
test_restamp_default_project.py (3 tests). 450 backend green.
This is the last framework piece for OHM's clean /p/ohm/ cutover.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A new entry can be proposed into a specific project; it lands in that project's
content repo and shows under that project's proposals. A non-default project is
no longer read-only.
- api.py: POST /api/projects/{pid}/rfcs/propose (propose body extracted into a
project-parameterized helper; unscoped /api/rfcs/propose kept as default
compat). Slug uniqueness, idea-PR reservation, landing state, and the
proposed_use_cases row scoped to the target project. GET
/api/projects/{pid}/proposals.
- cache.py: refresh_meta_pulls loops every project's content_repo, stamping
cached_prs.project_id; projects.content_repo(pid) helper.
- frontend: proposeRFC(projectId,…)/listProposals(projectId); ProposeModal
takes projectId; App resolves current project from the /p/<id>/ URL; Catalog
lists that project's proposals.
- tests: test_project_scoped_propose.py (lands scoped + gated 404). 447 backend
+ 11 Vitest green; clean build.
Known limitation: branch/PR/graduation edit flows + default-id re-stamp not yet
scoped (next slice). Per docs/superpowers/specs/2026-06-04-m3-backend-planb-design.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A second project's corpus now renders under /p/<id>/, isolated by its own
slug namespace. Completes step (1) "RFC app supports multiple projects" for
the read path.
- api.py: GET /api/projects/{pid}/rfcs (scoped catalog) + /{slug} (entry by
(project_id,slug)), behind the §22.5 read gate. Unscoped /api/rfcs stay as
default-project compat.
- cache.py: refresh_meta_repo iterates every projects row, mirroring each
content_repo into cached_rfcs stamped with its project_id (_upsert_cached_rfc
gains a project_id arg).
- frontend: api.listRFCs(projectId)/getRFC(projectId,slug); Catalog + RFCView
pass useProjectId(); ProjectLayout's NotServedPlaceholder guard removed (every
project serves), 404/not-readable branch kept; NotServedPlaceholder deleted.
- tests: test_project_scoped_serving.py (catalog scoping, entry isolation,
gated 404); ProjectLayout.test updated (non-default renders). 445 backend +
11 Vitest green; clean build.
Known limitation (next slice): write path + default-id re-stamp not yet scoped
(non-default project read-only); no live impact (no 2nd project in prod yet).
Per docs/superpowers/specs/2026-06-04-m3-backend-planb-design.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Lands the table rebuilds migration 026 deferred until a second project exists,
shipped separately from per-project serving for migration hygiene. No behavior
change (deployments stay single-project 'default').
- migration 028_project_scoped_keys.sql: folds project_id into the slug-keyed
PK/UNIQUE of 13 tables (cached_rfcs PK (slug)->(project_id,slug); the
UNIQUE/PK on cached_branches, branch_visibility, branch_contribute_grants,
stars, watches, pr_seen, branch_chat_seen, funder_consents, rfc_collaborators,
contribution_requests, proposed_use_cases gain project_id) and makes the FKs
to cached_rfcs on rfc_invitations/rfc_collaborators/contribution_requests
composite (project_id, rfc_slug) -> cached_rfcs(project_id, slug).
- db.run_migrations: a `-- migrate:no-foreign-keys` migration runs with
PRAGMA foreign_keys toggled OFF around it (required by SQLite's table-rebuild
procedure) + foreign_key_check after, failing loudly on dangling refs.
- ON CONFLICT upsert targets for the rebuilt tables gain project_id so they
match the new composite indexes (value still defaults to 'default').
- test_migration_028_project_scoped_keys.py: proves two-project same-slug
coexistence, within-project uniqueness, composite-FK enforcement. 442 pass.
- design doc §2 marked shipped.
Per docs/superpowers/specs/2026-06-04-m3-backend-planb-design.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Implements the M3-frontend slice of the §22 multi-project track, per
docs/superpowers/specs/2026-06-03-m3-frontend-design.md (design merged in
#10). Completes the runtime-config cut 0.33.0 (M3-backend Plan A) began.
Frontend:
- DeploymentProvider boots GET /api/deployment → {name, tagline,
defaultProjectId, projects}; brandTitle() neutral 'RFC' pre-fetch fallback.
- /p/:projectId/* routing with generic /e/<slug> segment. ProjectLayout
fetches /api/projects/:id, applies per-project theme (reset on switch),
provides ProjectContext, guards the corpus (served only for the default;
others get NotServedPlaceholder — decouples this slice from Plan B).
- Directory at / (2+ projects) with N=1 redirect into the single project;
ProjectSwitcher in deployment chrome; entry-noun by project type.
- VITE_APP_NAME hard cut: removed from vite.config + index.html; the 6 brand
reads now use deployment.name via context; static <title>RFC</title> + JS
document.title. Internal /rfc·/proposals links → /p/<project>/e|proposals
via lib/entryPaths.
Backend:
- GET /api/deployment returns default_project_id (the guard contract).
- Server-side 308s: /rfc/<slug>, /rfc/<slug>/pr/<n>, /proposals/<n> →
/p/<default>/… . nginx (testing + prod) routes /rfc/ and /proposals/ to
the backend.
Tests: 3 new backend redirect/deployment tests (438 pass); Vitest unit for
DeploymentProvider, ProjectLayout (theme/guard/404), Directory (11 pass);
clean build with no VITE_APP_NAME. Playwright e2e deferred until Tier-1 seeds
a registry (see CHANGELOG 0.35.0 step 5).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds two read endpoints that replace the build-time VITE_APP_NAME with a
runtime-config surface: /api/deployment returns deployment name/tagline plus
the projects visible to the caller (gated filtered by membership, unlisted
omitted from enumeration per §22.5); /api/projects/:id returns one project's
config and optional theme overlay, gated behind the §22.5 read gate.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Integrates the multi-project work so far — the §22 design drafts, M1 (the
schema spine), and M2 (project-scoped authorization + the §22.7 resolver) —
on top of the v0.32.0 retire/§13 changes that landed on main meanwhile.
Integration decisions:
- Renumbered the M1 projects migration 025_projects.sql -> 026_projects.sql.
v0.32.0 shipped 025_retired_state.sql, which rebuilds cached_rfcs and
predates the project_id column; running projects *after* the retire rebuild
is required so project_id survives on a fresh database (the runner applies
*.sql in filename order). Comment/doc references bumped to match.
- Resolved the get_rfc conflict in api.py by composing both gates: compute the
viewer once, apply the §22.5 visibility gate, then v0.32.0's §13.7
retired-entry owner-only check.
- api_graduation.py / api_discussion.py auto-merged cleanly (M2's threaded
viewer/visibility gate coexists with the new retire endpoints + retired
state).
Full suite 401 passed on a fresh DB.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Builds the three-tier authorization resolver on M1's schema spine: the
most-permissive union of deployment role (§6.1), project role (§22.6), and
per-RFC authority (§6.3/§12), with the §22.5 visibility gate subtractive on
top (§22.7). Pure app-layer — no migration (M1 shipped the tables), no
behavior change on the public default project. Verifiable on the single
default project by flipping its visibility and granting/revoking roles.
- app/auth.py: the §22.7 resolver primitives — project_visibility,
project_member_role, project_of_rfc, is_project_superuser, can_read_project,
effective write/discuss standing (can_contribute_in_project /
can_discuss_in_project), require_project_readable, visible_project_ids.
The three per-RFC capability helpers (can_discuss_rfc / can_contribute_to_rfc
/ can_invite_to_rfc) now compose all three tiers, so the ~20 call sites
inherit M2 unchanged.
- Read pass: the §22.5 visibility gate (404 to non-members) threaded into
every RFC-resolution helper across api.py / api_branches / api_prs /
api_graduation / api_discussion / api_contributions / api_invitations, plus
the catalog + proposals listings filtered by visible_project_ids.
- Write pass: the deep gates (branch read/contribute/owner, PR merge/withdraw/
edit, graduation, discussion resolve) fold in project_admin via
is_project_superuser; the "any-contributor" branch mode routes through
can_contribute_in_project; propose + claim gate on project contribute
standing. Deployment-level surfaces (admin idea-PR merge/decline, account
notification-mute) stay deployment-scoped.
- Operator decisions: implicit-on-public (a granted deployment contributor
keeps its pre-M2 write baseline on a public project, no membership row, so
the N=1 case stays whole) and preserve-curation (that baseline does not
override per-RFC owner curation — only an explicit project grant or a
deployment owner/admin does), keeping the v0.16.0 per-RFC invite contract
intact on public.
- tests: test_multi_project_authz_vertical.py — 9 vertical assertions on the
resolver tiers, public-unchanged regression, the gated 404 read gate, the
gated contribution gate, union + subtractive visibility, and revocation.
Full suite 390 passed.
- docs: mark Part C M2 "(landed)" with the two operator decisions.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Optional number (§13.2/§13.3): GraduateBody.rfc_id is now optional.
A blank/absent id graduates to `active` with `id: null`, leaving the
slug as the canonical identifier (§2.3). /graduate/check treats a blank
id as valid (ok:true); /graduate only validates the RFC-NNNN regex +
collision check when a number is supplied. The Graduate dialog allows an
empty number and renders number-less entries by slug (no RFC-undefined).
Retire (§3, §3.1, §13.7): new `retired` soft-delete state. An RFC's own
owners (frontmatter) and site `owner`-role holders — not app admins —
retire via POST /api/rfcs/<slug>/retire, an auto-merged meta-repo flip
PR (graduation machinery reused). Retired entries leave every browsing
surface: catalog, get_rfc (404 except site owner), discussion/branch
reads. Un-retire is site-owner-only (POST .../unretire), discoverable
via the owner-gated GET /api/admin/retired-rfcs ("Retired" admin tab).
Migration 025 widens the cached_rfcs.state CHECK to include 'retired'
(table rebuild, all columns preserved).
Tests: graduation no-number cases (active+null id by slug; check accepts
blank) added to test_graduation_vertical.py; new test_retire_vertical.py
covers perms (owner/site-owner allowed, admin/contributor 403),
catalog/read exclusion, and a graduate→retire→unretire round-trip. Full
backend suite 386 passing; frontend builds clean. SPEC §3/§3.1/§13
updated; CHANGELOG + VERSION → 0.33.0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Lays the additive foundation for hosting N projects per deployment, with
today's single corpus as the N=1 case. No behavior change: the app runs
exactly as before, single project, with the spine underneath.
- migration 025: `projects` + `project_members` tables; seed the `default`
project (visibility=public, preserving open-by-default); thread
`project_id NOT NULL DEFAULT 'default'` onto all 19 slug-bearing tables,
backfilling existing rows. Additive — no table rebuilds; the slug-keyed
uniqueness/PK rework is enumerated in the migration header and deferred to
the slice that activates project #2.
- app/projects.py: §22.13 startup backfill of the default project's
content_repo from META_REPO (idempotent — never clobbers a value the
future registry mirror sets).
- config: REGISTRY_REPO wired (optional through M1; consumed by the M3
mirror), documented in .env.example as META_REPO's successor.
- tests: 6 vertical assertions on the spine (seed, backfill, column shape,
role CHECK, idempotency, config). Full suite 381 passed.
- docs: align Part C M1/M3 boundaries with the landed code (registry mirror
+ redirect move to M3).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The provisioning docstring claimed the invitee row gets
`last_seen_at = NULL` as the "not yet arrived" discriminator. It does
not: the column is NOT NULL and the INSERT omits it, so it defaults to
datetime('now') — the longer note below already explained this, but the
bullet contradicted it. Rewrite the bullet to state the real behavior
(both timestamps default to now; the pending-invite state lives in the
unclaimed user_invite_tokens row) and note that consumers must treat a
pending-invite row as never-seen. Comment-only; no runtime change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Retire the per-RFC-repo model. RFCs now live in their meta-repo entry
(rfcs/<slug>.md) for their whole life; graduation is an in-place
super-draft → active state flip that keeps the body in the entry — no
repo creation, no body-strip, no five-step transaction, no rollback.
SPEC: §1 topology rewritten (one meta/content repository, no per-RFC
repos) with a deployer-facing "single content repository" framing; §2
(repo: always-null), §3 (active is in-place), §4, §9.8 (handoff
frictions dissolve), and §13 fully rewritten (two-field dialog, "The
flip", §13.6 RFC-0001 fold-back record).
Code: graduation collapses to open+merge one frontmatter PR; branch/PR/
chat dispatch re-keyed on meta-residency (repo IS NULL) so active RFCs
edit on the meta repo exactly as super-drafts do; the two "RFC has no
repo" 409 guards removed; promote-to-branch slug-embeds an active RFC's
auto-branch (edit-<slug>-<hex>) for shared-repo cache attribution;
refresh_meta_branches + hygiene branch-resolution include meta-resident
actives; the §9.8 read-only guard + pre_graduation_history scoped to
legacy per-repo only; dead bot primitives + GraduateDialog repo field +
blocking-PR popover removed. The repo: frontmatter field and the
/blocking-prs endpoint are retained (schema stability / informational).
Tests: graduation suite rewritten to the flip model; e2e + hygiene
updated. Full backend suite 375 passed; frontend builds.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
refresh_meta_pulls / refresh_rfc_repo recover a PR's slug from its
Gitea head.ref, which collapses to the refs/pull/<N>/head sentinel
once a merged PR's branch is deleted. The slug then parsed to None,
the row was skipped, and cached_prs.state froze at 'open' — so the
entry showed as both a super-draft and a pending idea. Recover the
real branch name from the stored cached_prs row when Gitea reports an
empty or sentinel ref.
Surfaced via the ROADMAP #35 operator authoring lane (CLI merge with
--delete-branch); the web UX leaves branches in place so it never hit
this. Regression test added; full suite 375 green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the v0.26.0 (#28 Part 1) read-time scanner into three buckets in
one pass — active link (Part 1), pending-RFC contribute offer (Part 3),
create-RFC offer (Part 2) — precedence active > pending > candidate. The
backend still emits only structured segments (never HTML), so the surface
stays XSS-safe by construction.
Part 2 — create-RFC offers: a multi-word tag from the #27 taxonomy with no
defining RFC renders, for a create-rights viewer, as an inline "+ create
RFC" affordance that opens the propose modal pre-filled (?propose=<term>;
ProposeModal gained initialTitle). Conservative multi-word gate; broader
heuristics + the Haiku path are deferred.
Part 3 — contribute-to-pending offers: a term matching a super-draft
renders, for a signed-in non-owner, an "ask to contribute" affordance with
the owner's display name. It opens a 3-field request form (who/why/optional
use-case); submitting lands a contribution_requests row (migration 024) and
one actionable §15 notification per owner (new kind
contribution_request_on_pending_rfc, personal-direct). The owner's inbox
shows who/why/use-case inline with Accept/Decline. Accept fires #12's
owner-invite flow with the requester as invitee and echoes a notification
back; decline notifies the requester. Pre-merge idea PRs are out of scope.
New endpoints: GET /api/rfcs/{slug}/contribution-target,
POST /api/rfcs/{slug}/contribution-requests,
.../{id}/accept, .../{id}/decline. The invite issue path was refactored
into one reusable api_invitations.issue_invitation(...) chokepoint shared
by the manual invite endpoint and Part 3's accept.
Tests: 9 new (3 scanner-bucket unit + 6 e2e). Full suite 374 passing.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two informational findings from the Session-0026 audit, both
framework-internal defense-in-depth. No operator action: no migration,
no schema/config/overlay change, no deployment-facing surface.
- I3: guard the dead text/html branch in email_envelope.build_envelope.
No send path passes body_html; the unused branch would emit HTML built
from possibly-unescaped user content (C1 stored-XSS class in the mail
channel). Passing body_html now raises NotImplementedError; the arg is
kept for documented future symmetry, enabling HTML mail becomes a
deliberate escape-then-unguard change.
- I4: make turnstile.verify_token async. The sync httpx.post ran inside
the async /auth/otc/request handler, blocking the event loop up to the
10s timeout on a slow CloudFlare call. It now awaits httpx.AsyncClient
via a narrow _siteverify_post seam (tests patch the seam, not the
shared AsyncClient). The sole caller (main.py) now awaits it.
Tests: full backend suite 365 passed. Added a coroutine-contract unit
test for verify_token and flipped the email_envelope HTML test to assert
the guard.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Remediates the rfc-app application + deploy-config findings from the
Session 0026 security audit. Cut as the "v0.25.0-security-hardening"
branch (from v0.24.0); reversioned to 0.27.0 on rebase onto main since
v0.26.0 (#28) shipped while this was in flight.
- C1 (Critical): single sanitizeHtml.js chokepoint (DOMPurify) for every
marked→innerHTML / dangerouslySetInnerHTML sink (MarkdownPreview,
ProposalView x2, Editor); rel=noopener hook on target=_blank links.
- H1: per-account OTC-verify lockout (migration 023, auto-applied) +
per-IP throttle via new ratelimit.py; wired on otc verify/request +
passcode check/verify.
- M1: device_trust.lookup() single indexed-row read — cookie value is now
"<row_id>.<raw_token>"; bcrypt-checks one row, not a global table scan.
(Behavior change: existing device-trust cookies re-prompt once.)
- M2: HTTP security headers (CSP/HSTS/XFO/XCTO/Referrer-Policy) at nginx.
- M4: session cookie Secure-by-default (SESSION_COOKIE_SECURE opt-out).
- M5: bounce webhook fails CLOSED (503) when secret unset, instead of open;
RFC_APP_INSECURE_BOUNCE_WEBHOOK=1 dev opt-in.
- L2/L3: per-IP cooldown + check-endpoint throttle.
- L4: systemd sandbox knobs. L8/I1: nginx server_tokens off + TLS1.0/1.1 out.
VERSION + frontend/package.json → 0.27.0; CHANGELOG documents the upgrade
steps (incl. the out-of-band nginx + systemd apply, which the flotilla
deploy gesture does not perform).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Part 1 of item #28: references to existing accepted RFCs inside PR
descriptions and comment text now render as inline links to the
referenced RFC. Parts 2 (offer-to-create) and 3 (offer-to-contribute-
to-pending) are deliberately deferred — Part 1 ships first as the easy
win, per the roadmap row.
Shipped in parallel with the v0.25.0 security-hardening session; this
took the next free version slot (0.26.0) per the roadmap's
"claims the next available version number" rule. Expect a top-of-file
CHANGELOG/VERSION merge with 0.25.0 — distinct concerns, trivial to
resolve.
Backend:
- rfc_links.py (new): builds a term index from the live accepted
(state='active') RFC corpus and segments plain text into text /
rfc-link segments. Conservative matching — links only rfc_id tokens
(RFC-0001), multi-word titles (Open Human Model), and hyphenated
slugs (open-human-model); a single common-word title/slug is NOT
linked (would turn every prose "human" into a link). Case-insensitive,
word-boundary-anchored, longest-match-wins, self-reference suppressed.
- api_prs.py get_pr(): enriches the PR description (description_segments)
and every PR comment (text_segments).
- api_discussion.py: enriches PR-less discussion comments (text_segments).
Read-time, not submit-time: the roadmap says "at submit time" but the
intent it names is "not as live compose preview", which read-time
honors. Chosen for correctness (links track the live active set —
newly-accepted RFCs start linking, withdrawn ones stop), zero migration,
and cheapness (small cache-resident corpus). Recorded as a §19.3-rule-2
note in the session transcript.
Frontend:
- LinkedText.jsx (new): maps backend segments onto React text nodes +
anchors. No dangerouslySetInnerHTML — XSS-safe by construction,
independent of any HTML-sanitization layer. Falls back to raw text
when segments are absent.
- PRView.jsx: description + PR conversation comment bodies render via
LinkedText.
- RFCDiscussionPanel.jsx: discussion comment bodies render via LinkedText.
- App.css: .rfc-autolink (subtle accent + dotted underline, tokenized).
Tests: 12 new (test_rfc_links_vertical.py) — 9 scanner units + 3
end-to-end (PR description / review comment / discussion comment all
surface *_segments; self-reference suppression). Full suite 363 green;
frontend builds clean.
No upgrade steps: additive, no migration, no secret, no config.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The §9.1 Slice-2 AI-suggested tag chips, deferred since the propose
modal landed, now wired up. As the propose-RFC draft fills in, the
backend asks Claude Haiku for tags drawn ONLY from the corpus's
existing tag set (the de-facto taxonomy — v1 tags are free-form chip
input, there is no curated list), surfaced as clickable suggestion
chips. Nothing auto-applies; the user clicks to add.
Backend:
- tag_suggest.py: universe gather (distinct corpus tags, most-common
first), Haiku prompt + tolerant reply parser (drops invented tags,
dedupes, clamps confidence), in-process per-user rate limit.
- providers.construct_haiku(): dedicated Haiku provider from the
operator key, independent of ENABLED_MODELS — tag suggestion always
uses the cheap+fast model. No RFC slug at propose time, so the §6.7
funder path does not apply.
- POST /api/rfcs/suggest-tags: contributor-gated, rate-limited.
Degrades to an empty list (never an error) when no Anthropic key is
bound, the corpus has no tags, or the draft is empty.
Frontend:
- ProposeModal: debounced suggestion fetch (700ms, stale-response
guarded), clickable suggestion chips, and the required inline
disclosure that the draft text is sent to Anthropic.
- api.suggestTags(): forgiving — any non-OK resolves to [].
Tests: 11 new (vertical contributor-gating / filtering / no-key /
empty-corpus / 429, plus units for gather, parser tolerance, max,
short-circuit, provider-failure). Full suite 351 green.
New secret on deploy: ANTHROPIC_API_KEY (see CHANGELOG Upgrade steps).
Disclosure copy wants a counsel pass before deploy, per #22 discipline.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Track each authenticated user's last-viewed route + light view state
server-side, and on next sign-in redirect them to that state, falling
back to the empty-state home only when there's no recorded state.
Backend:
- migration 022_user_session_state.sql: one row per user
(user_id PK/FK, last_route, last_route_state JSON-as-TEXT,
resume_enabled default 1, last_updated_at).
- PUT /api/me/last-state (require_user): upserts route + light state;
no-ops when resume_enabled=0. Localized in the /me region.
- /api/auth/me payload now carries resume_enabled + last_route +
decoded last_route_state (no extra round-trip).
- test_session_resume_vertical.py: auth-required, upsert/read-back,
per-user isolation, resume_enabled=0 disable.
Frontend:
- lib/useLastState.js: debounced (~1s) route-change PUT for
authenticated users; one-time resume redirect on sign-in, gated on
identify having fired (preserves #21 Part C identify-then-track).
- api.js: putLastState() client call.
- App.jsx: import + call the hook; set identifyReady after identify.
Header region untouched.
Per-user (not per-device); profile-settings opt-out toggle UI
deferred (column + default-on behavior ship now). Stored state is
route + light view state ONLY, never draft buffers — documented in
SPEC §6.8.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds an optional "What will you be using this for?" capture as a sibling
to the required justification on both propose surfaces, per roadmap #26.
- propose-RFC modal (ProposeModal): optional textarea below the required
"Why is this RFC needed?" pitch, labeled "What will you be using this
RFC for? (optional)".
- propose-PR modal (PRModal): optional textarea below the required
description, labeled "What will you be using this change for?".
- Backend: ProposeBody / OpenPRBody gain an optional `proposed_use_case`
(NULL/omitted accepted, no min, 8000-char cap matching the existing
free-text bound). Persisted to a new canonical side table
`proposed_use_cases` keyed by PR number, mirrored onto the cache
columns added by migration 021. Returned on the proposal list/detail,
RFC detail (by slug), and PR detail endpoints.
- Display: ProposalView, RFCView (main only), and PRView render the
captured use case with a muted "left blank" treatment when NULL.
- migration 021: nullable `proposed_use_case` on cached_rfcs/cached_prs
plus the reconcile-proof `proposed_use_cases` truth table.
- New vertical test_proposed_use_case_vertical: persists+returns when
supplied, accepted as NULL/omitted, for both surfaces.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ben Stull
Ben Stull
Ben Stull