Two bugs surfaced wiring dev auto-deploy:
1. Sourcing .image-tags.env clobbered the caller's incoming *_IMAGE_TAG. The
first backend-only deploy persists FRONTEND_IMAGE_TAG=local (sibling unknown);
the next frontend deploy then sourced that and pulled :local -> 'manifest
unknown'. Now the incoming env is captured before sourcing and re-applied.
2. Health-checked services via a host-port curl, but the dev overlay leaves the
frontend port unpublished (reached via the backend proxy) -> false failure.
Now polls each container's own HEALTHCHECK status via docker inspect.
The CI/SSH deploy paths (deploy.yml/deploy-prod.yml over SSH, deploy-dev on the
LAN runner) don't pass REGISTRY_TOKEN and rely on the host already being
docker-logged-in; an unconditional login with an empty token aborts deploy.sh
under set -e. Skip the login when no token is set and trust the host cred;
pull still fails loudly on a real auth problem.
The split backend's Python package data/ lives at /app/data; mounting the
appdata runtime volume there erased data/*.py and broke import at boot. Point
THERMOGRAPH_DATA_DIR + the appdata mount + the singleton lock at /state,
outside the code tree. Pairs with thermograph-backend paths.py making
DATA_DIR/LOGS_DIR env-overridable.
The split branch carried the #34-era render-secrets.sh, missing two fixes
required for prod/beta deploy: (1) sudo-read the root-owned 0400 age key into
SOPS_AGE_KEY when the deploy user can't read it directly; (2) in-place-write
/etc/thermograph.env when it's group-writable (beta's non-root 'deploy' user)
instead of only install(1). Without these, deploy.sh's render fails on both
hosts. Flagged by the terraform-layer workstream.
backend and frontend now reference their own images (BACKEND_IMAGE_PATH/TAG,
FRONTEND_IMAGE_PATH/TAG) instead of the shared emi/thermograph/app. deploy.sh
takes SERVICE=backend|frontend|all, rolls just that service with --no-deps,
persists each service's live tag in deploy/.image-tags.env so a single-service
deploy never disturbs the sibling, and fixes the old IMAGE_TAG:? guard that
made every workflow deploy hard-fail. Health-checks per service (8137/8080).
Claude-Session: https://claude.ai/code/session_01RdARHDJaYC1wSQRTYM6t3Z
Content-change pass following the extraction from the app monorepo (this repo
now stands alone, sourced via git filter-repo to preserve history):
- terraform/variables.tf, secrets.tf, modules/thermograph-host: remove every
app-secret Terraform variable (postgres_password, auth_secret, VAPID keys,
registry_token, Discord/SMTP creds, ...) and the random_password/random_id
generators. The SOPS+age vault (deploy/secrets/*.yaml) is now the sole
source of app secrets, rendered at deploy time by deploy/render-secrets.sh;
Terraform renders only a non-secret /etc/thermograph-topology.env (sizing,
routing) via the renamed thermograph-topology.env.tftpl template.
- hosts gains a required app_image_tag field: the host's own checkout is now
this infra repo, not the app repo, so there is no "current commit" to
derive an image tag from — every host pins one explicitly. repo_url now
points at this repo (private; typically needs an embedded read token).
- deploy.sh: IMAGE_TAG is now required from the environment instead of
derived via `git rev-parse HEAD` of the (now infra-repo) checkout, which
would have silently resolved to the wrong or a nonexistent tag.
- New terraform/modules/gcp-host: creates a GCE VM + minimal VPC/firewall
only, then feeds its IP into the same thermograph-host module every
SSH-managed host already uses — one provisioning path regardless of how a
host came to exist. var.gcp_hosts defaults to {}, so no google_* resource
is planned and the provider is never invoked without it (verified: plan
and validate succeed with no GCP credentials configured).
- terraform/README.md, ACCESS.md (renamed from INFRA.md), README.md: updated
for the new secrets model, the GCP scaffold, and this repo's own identity.
Verified: terraform fmt/validate/init clean; plan succeeds against realistic
dummy hosts (prod+beta shape) and against a populated gcp_hosts entry (plans
6 resources with no live credentials, confirming the composition wires
correctly end to end).
beta (Forgejo's pinned node) is also today's live thermograph.org host,
with Caddy already bound to ports 80/443. A second reverse proxy
binding those same ports would collide with it. Drop Traefik entirely:
forgejo's web port now publishes to 127.0.0.1:3080 only, and a new
Caddy site block (deploy/forgejo/caddy-git.conf) reverse-proxies
git.thermograph.org to it, reusing Caddy's existing automatic-HTTPS
instead of a second ACME flow.
That same Caddy block resolves the registry-exposure question (hazard
#15 in the hop1 runbook) as its first listed option: /v2/* (the
registry API) is blocked to everything outside the WireGuard mesh
CIDR, while the git/web UI stays public. README documents the
/etc/hosts override mesh clients need so registry traffic actually
routes over the tunnel rather than the public IP.
Bring the Swarm+Forgejo layer in line with the canonical topology in
docs/runbooks/implementation-handoff.md: three nodes (prod, beta, and
the desktop LAN dev machine) instead of two, with the Forgejo Actions
runner registered on the desktop as a plain systemd service rather than
a Swarm-scheduled Docker-in-Docker sidecar.
- setup-wireguard.sh: full N-peer mesh instead of point-to-point
- docker-stack.yml: drop the runner/runner-dind services, volumes, and
runner-token secret; only forgejo_db_password remains
- register-lan-runner.sh: register under both docker and
thermograph-lan labels, since one runner now covers both job types
- init-swarm.sh, join-swarm.sh, swarm/README.md, forgejo/README.md,
INFRA.md: updated node lists, order of operations, and access-state
table for three nodes
- deploy.yml: renamed to "Deploy to beta VPS" and reconciled the
main-vs-release branch question against terraform.tfvars.example
(this workflow already targets beta; a release-triggered prod
deploy doesn't exist yet and isn't invented here)
warm_cities.py and indexnow.py are one-shot scripts run manually at deploy
today - nothing keeps the curated city set topped up, or pings IndexNow,
between deploys.
Add notifications/scheduler.py: an in-process APScheduler instance driving
two recurring jobs (city warming, daily; IndexNow --if-changed, every 6h by
default), gated by the same leader election as the notifier - only the
process that already won leadership starts it, so the jobs run in exactly
one place under multi-host Swarm rather than once per replica. Neither job
runs immediately on start: warm_cities already runs at deploy time, so an
immediate duplicate on every worker boot/restart would be wasted work.
Extract indexnow.submit_if_changed() from the CLI's --if-changed branch so
the scheduler and the command line share the exact same skip-when-unchanged
logic instead of two copies drifting apart. The CLI's plain (non-flag) path
is unchanged.
APScheduler over a Postgres-native queue: exactly one process ever runs
this, no fan-out/backpressure/dead-letter need, no new infrastructure.
Terraform generates the secrets that have no external meaning
(POSTGRES_PASSWORD, AUTH_SECRET, METRICS_TOKEN, INDEXNOW_KEY) via the random
provider instead of requiring the operator to hand-generate and paste each
into terraform.tfvars. Each is pinned with a static keepers value (secrets.tf)
so apply never regenerates a value already in use - the exact incident class
this guards against: every session invalidated, the app<->DB password
mismatched. Rotation is now a deliberate keepers edit, never a side effect.
postgres_password/auth_secret move from required inputs to optional (default
"") - explicit var wins when supplied (seeding an EXISTING live secret during
a migration onto Terraform, hop-1 cutover runbook Stage 0), else Terraform
generates and owns it. metrics_token/indexnow_key are new: neither existed in
Terraform before, both previously left for the app's own fallback generation.
VAPID deliberately stays a required, non-generated input - an EC keypair
where regeneration breaks every existing push subscription outright, unlike
an opaque token.
Sizing tiers: a locals.sizes t-shirt map (nano/small/medium/large ->
{workers, app_cpus, db_cpus, db_memory}), toward the target Proxmox
sizing-tier model (architecture doc SS6) ahead of actually provisioning VMs -
Proxmox itself stays deferred; today a tier just sizes container caps on the
existing SSH-managed hosts. A host can reference one by name (hosts.<name>.
size) or keep hand-picking the four fields, so existing tfvars are
unaffected; prod's example now uses size = "large" (identical numbers),
beta keeps explicit numbers, and a commented uat example demonstrates the
shortcut for a future ephemeral host.
Strengthened terraform/README.md's local-state caveat: more Terraform-
generated secrets landing in tfstate raises the stakes of the existing
never-commit-cleartext-state guidance, not just the sizing.
Verified: terraform validate + fmt clean. A real `terraform plan` against
fake hosts (prod/beta/uat, mixing size="large"/explicit-numbers/size="nano")
resolved every sizing correctly (prod 8/8/4/16g, beta 4/4/2/8g, uat
1/1/1/1g) and planned exactly one instance of each random_password/random_id
resource. Applied just those four resources (real generation, -target to
avoid touching the fake SSH-only host resources) and re-planned: "No
changes" - confirming the keepers pinning holds. Adding an explicit
postgres_password override afterward left the random_password resource
itself completely untouched (0 replace/destroy), confirming the override
path never disturbs the generated resource.
The Swarm/Forgejo standup (a parallel infra track, see INFRA.md) landed real
.forgejo/workflows/{build,pr-build,deploy,deploy-dev}.yml before this PR
merged, making ci.yml a redundant duplicate of their build.yml (same build
gate, same job). Drop it.
Fix the remaining two files to match the real, already-deployed conventions
those files established rather than the guesses this PR shipped with:
runs-on: [self-hosted, thermograph] -> docker (the actual Docker-in-Docker
Swarm-hosted runner label), and appleboy/ssh-action referenced by full GitHub
URL (confirmed not mirrored on this Forgejo instance's default action
registry, per deploy.yml's own header comment) rather than the short form.
actions/checkout needed no change - already confirmed to resolve unchanged
from Forgejo's default mirror.
build-push.yml (image registry push) and ops-cron.yml (backup + IndexNow)
stay: neither duplicates anything in the real mirror, which faithfully
replicates the OLD git-checkout-and-build-in-place deploy model rather than
the registry-based one, and has no scheduled ops jobs at all.
Records the concrete, live state of Track 1 rather than just the setup
procedure: both boxes' roles/IPs, the ssh invocation, where the private key
lives (~/.ssh/thermograph_agent_ed25519, same convention DEPLOY.md already
uses for the CI deploy key), and how to rotate it. Also notes that neither
box has Docker yet and Terraform hasn't been applied to either, which is why
Track 2 is currently paused.
Checks off the two verification-checklist items that are actually confirmed
now (SSH access, password auth dead) rather than leaving them unchecked.
Three workflows for the self-hosted Forgejo instance (Track A chunk 7 of the
infra-design implementation handoff), mirroring/extending the existing
.github/workflows without touching them - GitHub stays the primary repo and
its own build.yml/ci-cd.yml keep gating auto-merge.
ci.yml mirrors build.yml's build gate (deps, backend tests, frontend JS
syntax check, boot + page/API health check) on a Forgejo runner.
build-push.yml builds the app image and pushes it to Forgejo's built-in
registry, tagged by git SHA (every push) and semver (version tags) - the
registry half of the hop-1 cutover's build-once/deploy-everywhere model.
Runs alongside the existing deploy.yml/deploy.sh (git-checkout-and-build-in-
place) without touching it, per the cutover runbook's Stage C.
ops-cron.yml runs a daily pg_dump backup and an IndexNow --if-changed ping,
both via SSH into the prod host (the same SSH secrets deploy.yml already
uses) using the app's already-running compose stack - no new network
exposure, no separate dependency install. These are ops/infra concerns
(scheduled via Forgejo Actions cron), distinct from the app-domain jobs the
worker's own APScheduler runs.
All three need a runner registered with the `thermograph` label and the
registry's public/mesh exposure resolved (Track B).
Verified: raw YAML syntax valid; actionlint clean except the pre-existing,
already-present-on-.github/workflows "unknown custom runner label" warning
(not something these files introduce - the existing thermograph-lan label
triggers the identical warning). One real finding fixed pre-commit: an
unused shellcheck-flagged loop variable in the boot-check retry loop.
* Split web/worker duties with THERMOGRAPH_ROLE
Background work (the subscription notifier) is welded to the same process
that serves requests, so scaling the web tier to N replicas would also scale
notifier instances unless something restricts it further than leader
election alone.
Add THERMOGRAPH_ROLE (web|worker|all, default all - unchanged single-process
behavior). Every replica runs the same image; ROLE only gates whether a
process is allowed to own the notifier at all, layered on top of the
existing leader election: web replicas never start it even if they'd win
leader election, worker replicas start it if they win. The decision is
pulled into _should_run_notifier() so it's unit-testable without booting the
full app (DB init, places index, neighbor warmer).
Add a minimal /healthz liveness route (no DB/upstream I/O, not under BASE)
so a worker replica - which serves no real traffic - still has something
Swarm can health-check.
* Add the Swarm interim stack file, a pinnable TimescaleDB tag, and a Caddy health-gate
Three changes toward the hop-1 interim cutover, all inert until Track B
stands up the platform:
docker-stack.yml: the Swarm stack file for the interim cutover, distinct
from docker-compose.yml (today's plain-compose deploy, unaffected). Pulls a
pre-built image (IMAGE_TAG) instead of building in place; app/worker publish
no host port (127.0.0.1:8137:8137 has no Swarm equivalent - Swarm's routing
mesh publishes on 0.0.0.0, which would expose the plaintext app un-fronted),
reaching Caddy only over an MTU-lowered overlay network (VXLAN-over-WireGuard
needs a smaller MTU or large payloads silently stall); db is placement-
pinned to a labelled node; app/worker skip inline migrations
(RUN_MIGRATIONS=0) so the runbook's one-shot migrate task is the only thing
that ever runs Alembic; secrets are real Swarm secrets mounted at
/run/secrets, read by the entrypoint shim rather than plain env vars.
TIMESCALEDB_TAG: docker-compose.yml's db image now reads this (default
latest-pg18, today's behavior unchanged), wired through Terraform
(timescaledb_tag, default "latest-pg18") so it can actually be pinned to an
exact minor without hand-editing the host - required before any host of the
stack could replicate with another (a floating tag risks mismatched
extension minors, which blocks a physical replica and risks compressed-
chunk corruption on restore).
Caddy active health-gate: both the Terraform-rendered Caddyfile and the live
deploy/Caddyfile now health-check the app on the same cheap /healthz route
its own Docker HEALTHCHECK uses (now /healthz instead of the SSR homepage,
so it's cheap enough for a tight interval and works identically for a
worker replica, which serves no public traffic at all) - Caddy won't
forward into a container that's still booting or unhealthy.
Verified live: built and booted the real image via docker compose - both
containers report healthy via the new /healthz-based HEALTHCHECK, and GET /
still renders the full SSR homepage unchanged. Both Caddyfiles validated
with the real caddy binary. docker-stack.yml validated with docker compose
config (required-var guards fire with clear messages; secrets correctly
mount at /run/secrets/<name>, matching the entrypoint shim's mapping).
docker-compose.yml validated with and without TIMESCALEDB_TAG set, alongside
the existing openmeteo overlay. terraform validate + fmt clean.
Three additive infrastructure layers on top of the two VPS boxes Terraform
already provisions (prod: new 48 GB/12-core box, thermograph.org; beta: old
VPS, 75.119.132.91). None of this touches backend/, Dockerfile,
docker-compose*.yml, terraform/, or deploy/db/ — that stays owned by the
app-containerization work in flight elsewhere; this is strictly the layer on
top. See INFRA.md for the full runbook and order of operations.
- deploy/provision-agent-access.sh: a dedicated, auditable full-sudo login
(not raw root) for agent-driven ops — passwordless sudo under a distinct
username, sshd hardened to key-only auth, auditd logging every
root-effective command. One line to revoke.
- deploy/swarm/: a 2-node Swarm (prod=manager, beta=worker) joined over a
WireGuard tunnel rather than trusting the public internet with the
overlay data plane, which Docker's own guidance says should never face it
directly. Swarm ports locked to the tunnel interface once joined. This
cluster's only workload is Forgejo — it does not orchestrate the
Terraform-managed app deploys, so nothing here can strand the app's
single-writer database.
- deploy/forgejo/ + .forgejo/workflows/: Forgejo + Traefik + a
Docker-in-Docker-sandboxed runner as a Swarm stack pinned to beta, plus
Forgejo Actions workflows mirroring .github/workflows/*.yml. The custom
auto-merge workflow step is dropped — it existed only to work around
GitHub's paywalled branch protection on private free-tier repos, which
Forgejo has no such tier for; native "auto merge when checks succeed"
replaces it, and as a real git push (unlike GitHub's non-triggering
token-merge) it fires the LAN deploy naturally with no double-trigger
logic needed. appleboy/ssh-action is referenced by full URL (not mirrored
on Forgejo's default action registry); actions/checkout and
actions/setup-python resolve unchanged.
Migration is mirror-first: the GitHub repo import and workflow files land
here, but cutting deploy secrets over and retiring GitHub happens only after
verification (INFRA.md 3d) — GitHub stays live as a fallback throughout.
One flagged, unresolved mismatch: deploy.yml still triggers on `main`, but
terraform.tfvars.example names prod's deploy branch `release`. Left as a
faithful mirror rather than guessed at — reconcile with whoever's driving
Terraform/deploy.
The subscription notifier elects one leader via a host-local flock
(THERMOGRAPH_SINGLETON_LOCK) so multiple uvicorn workers on one host don't
each run it. Under multi-host Swarm that guard is insufficient: each host
would independently elect its own leader, multiplying Open-Meteo quota use
N-fold again.
Add claim_pg(key) alongside the existing claim(lock_path): a cluster-wide
Postgres advisory lock, visible to every host talking to the same database.
The holding connection is dedicated and kept for the process lifetime
(advisory locks are session-scoped); a dead connection is dropped and
re-election retried on the next call.
claim_leader() dispatches between the two mechanisms from env:
THERMOGRAPH_SINGLETON_PG (+ Postgres) -> claim_pg; else
THERMOGRAPH_SINGLETON_LOCK -> claim; else always leader, unchanged. Wired in
at web/app.py in place of the direct claim() call. Off by default, so
today's single-host behavior is unaffected.
Replace the per-cell parquet cache with TimescaleDB hypertables as the
production backend for the raw daily climate record, and drop pg_duckdb.
Parquet stays the backend whenever THERMOGRAPH_DATABASE_URL is not a
Postgres URL (dev, tests, offline tooling), the same dialect switch the
accounts DB and derived store already use, so CI stays Postgres-free.
- data/climate_store.py: psycopg + polars bridge over climate_history
(a hypertable), climate_recent, and climate_sync (per-cell freshness).
Reads via pl.read_database, writes via COPY + ON CONFLICT upsert;
fail-soft to a cache miss so a DB hiccup degrades to upstream refetch.
- data/climate.py: route every cache/mtime touchpoint through a backend
dispatch. recent_stamp becomes int(recent_synced_at) on Postgres; the
stale-serve path still avoids bumping it, so derived-payload tokens
invalidate on exactly the same events as before.
- alembic 0002: CREATE EXTENSION timescaledb plus the hypertable schema
(compression policy on year-old chunks), guarded to no-op off Postgres.
- migrate_cache_to_pg.py (make migrate-cache): idempotent backfill of the
parquet cache into the hypertables, preserving file mtimes as sync
timestamps so recent_stamp is unchanged across cutover.
- db image -> stock timescale/timescaledb:latest-pg18; drop the custom
pg_duckdb Dockerfile, the read-only /parquet mount, and the duckdb
tuning GUC. Docs updated for the new backend and cutover.
Co-authored-by: Claude <noreply@anthropic.com>
Two prod-readiness hardening changes:
DB tuning scales with the container budget. Replace the fixed 8 GB
20-tuning.sql with 20-tuning.sh, which derives shared_buffers (25%),
effective_cache_size (75%), work_mem, maintenance_work_mem and
duckdb.max_memory (50%) from the DB_MEMORY the compose db service now passes in.
The ratios reproduce the historical 8 GB tuning exactly and scale linearly, so
the 48 GB prod box (db_memory 16g) gets shared_buffers 4 GB / duckdb 8 GB with
no separate edit. Beta/local (8g default) are unchanged. Docs that told
operators to raise the tuning by hand are updated.
Boot ordering for the self-hosted archive. On an openmeteo host, install a
docker.service drop-in (Wants/After rclone-om.service) so Docker starts after
the object-storage mount is ready on every boot — the restart-policy containers
never bind an empty mount point. rclone-om is Type=notify, so After waits for
the mount to actually be ready. Cleaned up when openmeteo is toggled off.
Get the 45-year historical record off the rate-limited public Open-Meteo
archive API by running a private Open-Meteo instance that serves the
era5_seamless blend (0.1° ERA5-Land + 0.25° ERA5 for gusts) from the
compressed .om archive in object storage, mounted on the host with rclone.
- climate.py: make ARCHIVE_URL env-driven (THERMOGRAPH_ARCHIVE_URL) and pin
models=era5_seamless on the archive fetches only, so a self-hosted instance
serves the same 0.1° resolution; forecast path unchanged. The public API's
default is already seamless, so dev/beta (URL unset) behave identically.
- docker-compose.openmeteo.yml: open-meteo-api + two rolling sync workers
(era5_land 0.1°, era5 0.25° for gusts), bind-mounting the object-storage
mount; the overlay points the app at the local instance.
- Makefile: om-up / om-down / om-backfill (one-time full-history backfill).
- Terraform: per-host openmeteo flag layers the overlay, renders OM_DATA_DIR,
and provisions the host rclone systemd mount from the bucket credentials.
- deploy/openmeteo: operator runbook + rclone mount unit template.
Terraform config under terraform/ manages the two existing VPS hosts and hands the
app to docker-compose, with local state:
- prod: the new 48GB/12-core VPS (release branch, thermograph.org), sized larger.
- beta: the old VPS 75.119.132.91 (main branch, testing tier), no public domain.
- The LAN dev box stays on deploy/deploy-dev.sh (dev branch) — out of Terraform.
A reusable module (modules/thermograph-host) SSHes each host to install docker/
compose/ufw (+ Caddy when a domain is set), sync the checkout to the host's branch,
render /etc/thermograph.env from Terraform variables (secrets pushed via provisioner
content, never on local disk), `docker compose up -d`, and health-check. Named
volumes are preserved on re-apply, so the Postgres data is never recreated.
Container resources are now env-driven in docker-compose.yml (APP_CPUS/DB_CPUS/
DB_MEMORY/WORKERS) with unchanged defaults, so Terraform can size each host.
deploy/db/init/20-tuning.sql sets the memory budget via ALTER SYSTEM (applied on a
fresh volume, effective after the post-init restart): shared_buffers 2GB,
effective_cache_size 6GB, work_mem 64MB, maintenance_work_mem 512MB, max_wal_size
4GB, plus duckdb.max_memory 4GB for parquet processing.
Compose: the db container gets an 8g memory ceiling + 1g shm_size (parallel-query
shared memory) in prod; the dev overlay resets mem_limit so dev memory is uncapped
too (the ~8 GB budget still comes from the tuning). Verified the settings take
effect and pg_duckdb still loads.
Run Thermograph as a docker-compose stack (app + Postgres 18) and standardize the
data layer on Postgres, while keeping the test suite on SQLite.
- accounts/db.py: DSN-driven engines. On Postgres, a per-worker read-write +
read-only asyncpg pair (the RO engine pins read-only transactions, used by the
pure-GET endpoints) plus a sync psycopg engine for the notifier thread; the
SQLite path is preserved for tests/local (selected when THERMOGRAPH_DATABASE_URL
is unset). models.py: boolean server_default -> sa.false().
- store.py / metrics.py: dialect-flexible — Postgres UNLOGGED tables via psycopg
when configured, else the existing raw-sqlite3 paths byte-for-byte; sync
interfaces and every fail-soft contract preserved.
- Alembic (backend/alembic/) manages the accounts schema; the container entrypoint
runs `alembic upgrade head` before uvicorn (4 workers). migrate_accounts_to_pg.py
copies the accounts data SQLite->PG through the ORM (UUID/bool/JSON coerced),
skips access_token, and resets identity sequences.
- Dockerfile + docker-compose.yml: app image (uvicorn, 4 workers, loopback 8137)
and a Postgres 18 db (2 CPUs) running pg_duckdb (deploy/db/) so the parquet
climate cache is queryable in-DB via read_parquet('/parquet/cache/*.parquet').
- deploy.sh/thermograph.service rewired to manage the compose stack; env example,
Makefile targets (up/down/db-up), and deploy/POSTGRES-MIGRATION.md cutover runbook.
Tests stay on SQLite (dialect fallback) — 323 pass. The full Postgres stack was
verified via docker compose: alembic migrations, register/login, the RO endpoint,
store/metrics round-trips, and the accounts data migration.
The deploy scripts pulled code and restarted but never applied the
hand-written SQL migrations in deploy/migrations/, so column additions to
the long-lived accounts DB had to be run by hand.
Add deploy/migrate-db.py, an idempotent runner that resolves the accounts
DB the same way the app does (THERMOGRAPH_ACCOUNTS_DB or
<repo>/data/accounts.sqlite), applies any files not yet recorded in a
schema_migrations table, and:
- backs the DB up before touching it;
- baselines a fresh DB (no file, or no user table) rather than ALTERing
a table create_all() will build with the current schema on next start;
- tolerates an already-present column/index (hand-applied or model-built)
by recording it instead of failing.
Wire it into deploy.sh (prod) and deploy-dev.sh (dev) just before the
service restarts, so new code never queries a column an older DB lacks.
No env or systemd changes: the migration runs as the deploy user, who owns
the data dir.
Claude-Session: https://claude.ai/code/session_01XXxmNFy9cZ6Gh8Y9thZn62
