The six deploy workflows were one file written six times, differing only in a
branch name, a paths filter, a concurrency group, the service name, its
*_IMAGE_TAG variable and a secret prefix. The two build-push workflows were the
same file twice, differing only in the domain string. The contract into
infra/deploy/deploy.sh (SERVICE + BACKEND_IMAGE_TAG/FRONTEND_IMAGE_TAG) was
already fully parameterised, so the duplication bought nothing and cost eight
files to keep in step.
deploy.yml: branch selects the environment (main -> beta, release -> prod), a
matrix covers backend and frontend, and each leg decides whether this push
actually touched its domain before rolling anything. The workflow-level paths
filter only says backend OR frontend moved; without the per-leg refinement a
backend-only push would also roll the frontend and lose the independent-deploy
property the FE/BE split exists for.
build-push.yml: the same shape for images. Images stay separate and
independently deployable; nothing about the published artefacts changes.
The two *-deploy-dev.yml workflows are deleted rather than ported. They were
already documented as inert -- they call a monorepo path on the LAN box whose
~/thermograph-dev is still a split-era thermograph-infra checkout. LAN dev is a
local `make dev-up` concern, not a CI environment.
Deliberately boring expressions throughout. No dynamic matrix (fromJSON), and
the *_IMAGE_TAG selection is done in shell rather than with a
`matrix.service == 'x' && a || b` ternary. Those are GitHub idioms a Forgejo/act
runner may evaluate differently, and the failure mode is silent: an empty
*_IMAGE_TAG makes deploy.sh fall back to the tag already running, so the job
goes green having deployed nothing. Beta and prod get two explicit, mutually
exclusive steps rather than a ternary over secrets, where an empty host would be
worse still.
Everything load-bearing is preserved: fetch-depth 0 and the domain-keyed 12-hex
tag (the branch tip is often another domain's commit), per-service per-ref
concurrency with cancel-in-progress false, separate PROD_SSH_* credentials, the
v*.*.* both-images exception, and appleboy/ssh-action by full URL.
pr-build.yml is untouched. Its workflow name and `gate` job are the required
status check branch protection is configured against, and renaming that context
makes every PR unmergeable with fully green CI. secrets-guard.yml and
shell-lint.yml are likewise left alone: they are the only other consolidation
candidates that run on pull_request, and the branch-protection API needs
credentials this could not read, so merging them was not worth the risk for two
files.
Logic verified by replaying the plan step against real history: an infra-only
push rolls nothing, a backend-only push rolls backend and leaves frontend
alone, and a run with no usable before-sha defaults to deploying rather than
silently skipping. The computed frontend tag for a backend-only push is
sha-ca84e0ce95f0 -- exactly the tag live on beta -- so the derivation matches
what the old workflows produced.
15 workflows -> 9, 1365 lines -> 1019. Docs naming the deleted files are
updated in the same commit, per the rule the root CLAUDE.md now carries.
main took two direct PRs (#19 shell-lint, #21 the thermograph-daemon Go
service) while dev accumulated the ERA5 lake stack; both sides added compose/
stack services and touched the same seams. Resolutions keep both worlds:
- compose + Swarm stack: lake AND daemon are sibling services; backend env
carries THERMOGRAPH_LAKE_URL and THERMOGRAPH_INTERNAL_TOKEN.
- deploy.sh: backend deploys roll (backend lake daemon); all adds frontend.
- requirements.txt: websockets/apscheduler stay removed (moved to the Go
daemon), duckdb stays (the lake role's engine).
Merged tree: backend suite green, both compose configs validate.
The gateway bot and APScheduler were long-lived stateful I/O loops running
inside the async web app under a leader election. They move into a single Go
binary that owns ONLY that I/O -- websocket, RESUME, heartbeat, backoff, timers.
It owns no grading logic. Anything needing data calls back over a new
internal-only surface (/internal/discord/grade, /internal/jobs/*). Grading
depends on polars and the parquet cache; reimplementing it in Go would let the
bot's grades drift from the API's. The grade route returns gateway-ready JSON
and Go relays the bytes verbatim.
The binary ships in the backend image and runs as a second compose service off
the same tag, so the two ends of the /internal/* contract can never skew.
deploy.sh rolls daemon alongside backend -- without that the service would never
be created, since a single-service deploy uses --no-deps. It also probes the
image first and skips the daemon when rolling a tag that predates the binary:
infra tracks main while image tags are env-staged, so a host can legitimately be
asked to roll an older backend image, and creating the service anyway would
leave a container crash-looping on a missing binary.
replicas: 1 with order: stop-first replaces the leader election -- Discord
permits one gateway connection per bot token.
THERMOGRAPH_INTERNAL_TOKEN is optional: both ends derive it from
THERMOGRAPH_AUTH_SECRET via HMAC under a domain-separation label, so this needs
no new vault entry. The derivation is pinned to a shared cross-language test
vector asserted on both sides, so drift fails CI instead of 401ing every call.
Fail closed when neither secret is set.
Improvements over the Python: a close intended for RESUME uses 4000 rather than
1000 (Discord invalidates a session closed 1000, so the old default defeated its
own resume); MESSAGE_CREATE runs on a bounded worker pool; and a malformed HELLO
returns an error rather than a clean reconnect, which would otherwise reset
backoff and hot-loop against the gateway.
365 Python tests pass; Go build/vet/test -race clean; shellcheck 0 findings.
/opt/thermograph becomes a monorepo checkout: deploy.sh gains INFRA_DIR (git
ops at the root, compose work cd'd into infra/), lock/tags/render paths move
under infra/deploy/, image-path defaults become emi/thermograph/backend|
frontend across compose, stack, and the tag-prune. docker-compose.yml pins
name: thermograph -- without it compose run from .../infra derives project
"infra" and recreates the whole stack beside the running one;
deploy-dev.sh pins COMPOSE_PROJECT_NAME=thermograph-dev (env wins over the
file key) to keep LAN dev's separate project.
