DATABASE_URL_SYNC is now set as a Railway reference variable pointing to
${{pgvector.DATABASE_URL}}, which resolves to the correct postgres superuser
credentials per environment (production, PR preview, fresh DBs). This handles
the bootstrap case where resolutionflow_admin doesn't exist yet.
Falls back to ADMIN_DATABASE_URL (sync-converted) for local dev only.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Prior approach (ADMIN_DATABASE_URL first) broke PR preview environments: fresh
Railway PostgreSQL instances have no resolutionflow_admin role yet, so the admin
URL fails before the create_db_roles migration can run (bootstrap deadlock).
New priority order in _alembic_sync_url():
1. PGHOST/PGUSER/PGPASSWORD/PGDATABASE — Railway auto-links these from the
PostgreSQL service per-environment, giving correct superuser creds for every
env including fresh PR preview DBs where no custom roles exist yet.
2. ADMIN_DATABASE_URL (resolutionflow_admin, BYPASSRLS, asyncpg→sync) — local
dev and stable envs where the role already exists.
3. DATABASE_URL_SYNC — legacy fallback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
DATABASE_URL_SYNC uses the postgres superuser whose password is unavailable
in Railway after Phase 1 switched runtime to the app role. resolutionflow_admin
(BYPASSRLS) is the correct role for migrations. Derive a psycopg2 sync URL from
ADMIN_DATABASE_URL; fall back to DATABASE_URL_SYNC for local dev environments
where ADMIN_DATABASE_URL is not set separately.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Run alembic upgrade head before tests so DB roles and RLS policies exist
- Set TEST_DB_NAME=resolutionflow_test so test_rls_isolation.py connects to
the correct database (was defaulting to patherly_test which doesn't exist in CI)
- Set DB_APP_ROLE_PASSWORD so create_db_roles migration creates resolutionflow_app
with a known password that the RLS isolation tests can connect with
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Backend: start_session, prepare_session, batch_launch_sessions all missing
account_id=current_user.account_id — Phase 1 NOT NULL constraint made these
500 in test suite (test_ratings.py fixture couldn't create sessions).
Frontend ESLint:
- TaskLane.tsx: suppress react-refresh/only-export-components for clearTaskState
- TeamSummary.tsx: init loading from isAccountOwner to avoid sync setState in effect
- ScriptBodyEditor.tsx: move lastValueRef.current assignment into useEffect
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Migration 019 only backfills trees with team_id IS NOT NULL.
Migration 3a40fe11b427 only covered is_default=TRUE trees.
Trees with team_id=NULL and is_default=FALSE (e.g. inactive test trees,
pre-team-system content) fell through both passes and triggered the NULL
guard.
Add two new UPDATE steps after the is_default pass:
1. Assign remaining trees to their author's account (if author has one)
2. Final fallback to PLATFORM_ACCOUNT_ID for any still-NULL rows
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After migration 174f442795b7 enforces NOT NULL on account_id, all
platform/global content must use the sentinel platform account instead
of NULL. Three categories of fixes:
1. trees.py: is_default trees now get PLATFORM_ACCOUNT_ID (not None)
2. admin_categories.py: global category CRUD now uses PLATFORM_ACCOUNT_ID
3. categories.py, tags.py, step_categories.py: creation endpoints coerce
None → PLATFORM_ACCOUNT_ID; IS NULL filter queries updated to
== PLATFORM_ACCOUNT_ID (IS NULL queries returned empty after migration
backfilled all global rows to the platform account)
Defines PLATFORM_ACCOUNT_ID constant in app/core/service_account.py.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The INSERT into template_trees incorrectly referenced `tags` as a column
on the `trees` table. Tags are a relationship via the `tree_tag_assignments`
join table — there is no direct column. Migration was failing with:
UndefinedColumn: column "tags" does not exist ... FROM trees
Fixed by replacing COALESCE(tags, '[]') with a correlated subquery that
aggregates tag names from tree_tag_assignments → tree_tags.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Migration 057 inserts 6 AD script templates with NULL team_id and NULL
created_by. Neither backfill path (created_by→users, team_id→team admin)
could attribute them to an account, causing the verify check to fail.
Fix: pre-create the platform sentinel account (ON CONFLICT DO NOTHING,
safe since 3a40fe11b427 also creates it idempotently) and add a final
fallback UPDATE assigning any remaining NULL script_templates to it.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
PostgreSQL UPDATE...FROM does not allow the updated table to be
referenced inside the FROM clause's JOIN conditions. Replace the
LEFT JOIN psa_connections with a correlated subquery.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
session_resolution_outputs is created in migration 067 (sequential branch
from 064). On fresh databases, Alembic could run cc214c63aa30 before 067,
causing "table does not exist" errors. depends_on ensures 067 always runs
first regardless of branch traversal order.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Combines the Phase 1 tenant isolation chain (064 → ... → 174f442795b7)
with the main sequential chain (064 → ... → 070) into a single Alembic
head (a9f3b2c1d4e5) so `alembic upgrade head` in the Dockerfile works
without ambiguity.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
b8d2f4a6c091 was NOT the production head. The true head was 064
(064_normalize_script_builder_messages) via the chain:
b8d2f4a6c091 → f0aad74ea51b → 062 → 063 → 064
This caused 'multiple head revisions' on Railway deployment.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The tags column was accidentally omitted from the is_default tree copy.
Now uses COALESCE(tags, '[]'::jsonb) to preserve source tree tags.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
All previously-nullable account_id columns are now NOT NULL.
tree_embeddings and feedback backfilled before constraint applied.
Global content assigned to platform sentinel account (00000000-...-0001)
in preceding migration.
Tables updated: users, trees, tree_categories, tree_tags,
step_categories, step_library, tree_embeddings, feedback
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Creates template_trees and platform_steps (no account_id, no RLS).
Migrates is_default=TRUE trees and public steps into them.
Creates sentinel platform account (00000000-...-0001) for global
tree_categories, tree_tags, step_categories, step_library, and
is_default trees — clearing all NULL account_id rows in those tables
as prerequisite for Group 9 SET NOT NULL.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Zero rows in production — this is a schema-only migration in practice.
team_id kept for app code compatibility. Drop deferred to later cleanup.
Backfill: team_id → team admin user → account_id; fallback: created_by.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
team_id is kept in all three tables — drop deferred until app code
is fully migrated off team_id references.
Tables: script_builder_sessions, script_templates, script_generations
Backfill: user_id/created_by → users.account_id
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
psa_post_log: backfill via psa_connection, fallback to posted_by user
psa_member_mappings: backfill via psa_connection
notification_logs: backfill via notification_config
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Backfill from rater/user's account_id (not the step's account_id).
This is an explicit design decision — step rating data is attributed
to the account that performed the rating.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
