Local work has been stranded on one machine: `make seed` replays
seed/fixtures/seed.json, which is generated from krow-demo and has never
reflected what is in a database. Agents published through importagents,
knowledge ingested, applications screened — none of it had a path to
another deployment.
`make export-data` reads the database itself and writes replayable SQL.
Three things it does deliberately:
Tables are ordered topologically from pg_constraint rather than left in
pg_dump's own order, which sorts by name and so fails on foreign keys in a
way that depends on what the tables are called. Parents always precede
children, and Kahn's algorithm breaks ties alphabetically so two runs
against one schema produce a byte-identical file.
Every statement is INSERT ... ON CONFLICT DO NOTHING. The export can
create rows on a target and cannot modify or delete one. That is a
property of the generated file, not a rule someone has to remember when
they apply it.
The file refuses to apply to a schema older than the one it came from. A
restore into a half-migrated database half-succeeds, and a partial import
is harder to unpick than a failed one.
pg_dump 18 wraps its output in the psql meta-commands \restrict and
\unrestrict. Dumping per table left the closing one without its opener,
which fails with "not currently in restricted mode" and, under
--single-transaction, rolls back having inserted nothing — silently, if
the caller reads psql's output through a pipe instead of its exit code.
Each table's slice is therefore cut at the last line ending in a
semicolon, which no line of pg_dump's epilogue does and every generated
statement does.
sessions and schema_migrations are excluded: sessions are bound to cookies
one deployment issued, and a stale schema_migrations row would make the
target lie about its own version.
Verified against a scratch database migrated to 15: all 25 tables match
the source row for row, a second run is a no-op, and the guard refuses a
version-10 target.
The output is real tenant data — password hashes, personal details — so
seed/exports/ is gitignored. It moves over scp, not through git.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
CORS
cors.go never set Access-Control-Allow-Credentials, so the
cookie-authenticated API was unreadable from any cross-origin frontend:
the server answered correctly and the browser blocked the page from
reading it. Set for allowlisted origins on both the preflight and the
actual response. Three tests added.
HTTP_COOKIE_SAMESITE (lax|none|strict, default lax) is new. CORS is only
half of what a cross-origin browser call needs; SameSite is judged on
registrable domain, so a frontend on an unrelated domain gets perfect CORS
headers and still no cookie. "none" is the only value that survives that,
and validate() refuses it without the Secure flag.
The "*" rejection now explains itself: browsers refuse Allow-Origin "*"
together with credentials, so it would break every authenticated call
rather than loosen anything.
Transactional endpoints (api-contract.md 12.1)
POST /api/v1/job-applications/{id}/hire
POST /api/v1/job-postings/{id}/assignments
Replaces two client-side loops that wrote several records with no
transaction and no rollback. Each is now one endpoint and one transaction,
built over repo.Repo so org scoping, derived columns, type casts and error
translation are not re-derived. Authorization reuses the existing policy
table rather than adding a parallel one: a workflow is exactly as
privileged as the writes it performs. 13 tests, including both rollback
paths.
Bug fix in the repository layer
repo.bindValue handled int64/int/float64/string but not int32, which is
what pgx returns for a PostgreSQL `int` column. Nothing previously read a
record and wrote one of its fields elsewhere, so it never surfaced; the
hire flow does exactly that and failed with "ai_score must be a number".
Both KindInt and KindFloat now accept the widths pgx actually produces.
Deployment
infrastructure/Dockerfile.api multi-stage, cross-compiling (BUILDPLATFORM
+ GOARCH) so linux/amd64 builds from arm64 are compiled rather than
emulated. Alpine runtime, non-root uid 10001, 22.1 MB. Ships api, seed,
setpassword and migrate, plus the migrations, so a Kubernetes
initContainer can apply the schema from the same image and tag as the
API. HEALTHCHECK keys on status code, not body, so a "degraded" instance
is not pulled from rotation during a migration window.
infrastructure/docker-compose.yml migrations run to completion before the
API starts. Assumes a managed PostgreSQL; the local-db overlay adds one
with TLS enabled so APP_ENV=production is met rather than dodged.
scripts/drop_public_tables.go the one-off used to clear an unrelated
schema from krowdb on 2026-08-24, kept for the record. Build-tagged
ignore and gated on CONFIRM_DROP=yes.
Verified against PostgreSQL: 16/16 new tests pass, and the image was built,
run and exercised end to end (login, CORS preflight, authenticated reads,
transaction rollback).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CmQiGq73Uyfq7J4yR8Vxxw
