Merge pull request 'Promote dev to main (reconciled bot + hardening merge, test harness, in-image CI)' (#15) from dev into main

.forgejo/workflows/build.yml

@@ -9,9 +9,9 @@ name: Build check
 # the runner's shared capacity=2 contention) without ever settling into a
 # reliably green check. The image DOES boot correctly standalone --
 # independently verified by hand: `docker run` + real alembic migrations +
-# `/healthz` returning 200 + a real `/api/v2/place` 200. Full pytest-suite
+# `/healthz` returning 200 + a real `/api/v2/place` 200. The full pytest
-# migration is separate, deferred follow-up work too, same category as
+# suite DOES run here now -- inside the built image (see the test step below),
-# thermograph-copy's deferred vendoring.
+# which sidesteps those runner quirks entirely.
 
 on:
   # Reusable (uses: ./.forgejo/workflows/build.yml) by pr-build.yml and
@@ -38,3 +38,21 @@ jobs:
 
       - name: Build
         run: docker build -t thermograph-backend:ci .
+
+      # The suite is hermetic (sqlite/tmpdir; no network), and the image already
+      # contains tests/ + every runtime dep (COPY . /app/), so run pytest INSIDE
+      # the image we just built: the exact interpreter/deps that ship, none of
+      # the runner-environment quirks that sank the old host-side boot check.
+      # requirements-dev.txt only layers pytest on top, so the install is tiny.
+      # -u 0: the image's default user can't write to site-packages.
+      # --entrypoint sh: the image's entrypoint is alembic-migrate + uvicorn;
+      # without the override the test command is swallowed as entrypoint args
+      # and the container just boots the server forever. unset THERMOGRAPH_BASE:
+      # the image bakes the prod root-mount (/), which moves every route off the
+      # /thermograph prefix the tests are written against -- requests would fall
+      # through to the frontend-proxy catch-all and fail with ConnectError.
+      - name: Run tests in the built image
+        run: |
+          docker run --rm -u 0 --entrypoint sh thermograph-backend:ci \
+            -c "unset THERMOGRAPH_BASE; pip install -q --no-cache-dir pytest==8.4.1 && python -m pytest tests -q"
+

.forgejo/workflows/deploy.yml

@@ -7,9 +7,8 @@ name: Deploy backend to beta VPS
 # versa. thermograph-infra/deploy/deploy.sh persists each service's live tag
 # host-side, so a single-service roll never disturbs the other.
 #
-# The SSH_HOST/USER/KEY/PORT secrets point at beta. Prod is NOT deployed by a
+# The SSH_HOST/USER/KEY/PORT secrets point at beta. Prod deploys the same way
-# workflow -- it's deployed with `terraform apply` on the `release` branch, so
+# from deploy-prod.yml (release branch, PROD_SSH_* secrets). appleboy/ssh-action is
-# there is deliberately no release-triggered workflow. appleboy/ssh-action is
 # referenced by full GitHub URL because it isn't mirrored in Forgejo's default
 # action registry.
 #

.gitignore

@@ -1,4 +1,5 @@
 .venv/
+.venv-test/
 __pycache__/
 *.pyc
 data/cache/*.parquet

Makefile

@@ -0,0 +1,14 @@
+# thermograph-backend — local dev/test targets.
+.PHONY: help test smoke clean-venv
+
+help: ## List targets
+	@grep -hE '^[a-z][a-zA-Z0-9_-]*:.*## ' $(MAKEFILE_LIST) | awk -F':.*## ' '{printf "  %-14s %s\n", $$1, $$2}'
+
+test: ## Build a py3.12 dev venv and run the hermetic pytest suite (pass ARGS=...)
+	./scripts/test.sh $(ARGS)
+
+smoke: ## Boot the backend image + a throwaway db and assert /healthz + /api/version
+	./scripts/smoke.sh
+
+clean-venv: ## Remove the test venv
+	rm -rf .venv-test

docker-compose.test.yml

@@ -0,0 +1,40 @@
+# Boot-smoke harness: run the backend's OWN image against a throwaway TimescaleDB and
+# assert it actually boots + serves its contract (build alone doesn't prove boot). Used
+# by scripts/smoke.sh; not a deploy file. DB data is tmpfs (ephemeral) — no volume, no
+# residue. See Makefile `smoke` and README.
+services:
+  db:
+    image: timescale/timescaledb:${TIMESCALEDB_TAG:-latest-pg18}
+    environment:
+      POSTGRES_USER: thermograph
+      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-smoke}
+      POSTGRES_DB: thermograph
+    tmpfs:
+      # pg18 images place PGDATA in a major-version subdir under the mount, so mount at
+      # /var/lib/postgresql (matches the infra compose), not /var/lib/postgresql/data.
+      - /var/lib/postgresql
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U thermograph -d thermograph"]
+      interval: 3s
+      timeout: 5s
+      retries: 20
+
+  backend:
+    # Defaults to a locally-built `:local` image (scripts/smoke.sh builds it). In CI,
+    # set BACKEND_IMAGE_TAG=sha-<12hex> to smoke the exact published image.
+    image: ${REGISTRY_HOST:-git.thermograph.org}/${BACKEND_IMAGE_PATH:-emi/thermograph-backend/app}:${BACKEND_IMAGE_TAG:-local}
+    depends_on:
+      db:
+        condition: service_healthy
+    environment:
+      THERMOGRAPH_DATABASE_URL: postgresql+asyncpg://thermograph:${POSTGRES_PASSWORD:-smoke}@db:5432/thermograph
+      # Required at import (web/app.py). The smoke never exercises a proxied page, so an
+      # unreachable placeholder is fine — it only hits /healthz and {BASE}/api/version.
+      THERMOGRAPH_FRONTEND_BASE_INTERNAL: http://127.0.0.1:1
+      THERMOGRAPH_BASE: "/"
+      THERMOGRAPH_ENABLE_NOTIFIER: "0"
+      PORT: "8137"
+      WORKERS: "1"
+    ports:
+      # Host port defaults to 18137 to avoid colliding with a running dev server on 8137.
+      - "127.0.0.1:${SMOKE_HOST_PORT:-18137}:8137"

scripts/smoke.sh

@@ -0,0 +1,45 @@
+#!/usr/bin/env bash
+# Image boot-smoke: prove the backend IMAGE actually boots and serves its contract
+# (a `docker build` succeeding does not prove the container starts). Builds the image
+# locally (or uses a provided BACKEND_IMAGE_TAG), runs it + a throwaway TimescaleDB via
+# docker-compose.test.yml, waits for /healthz, then asserts /healthz and the version
+# endpoint. Always tears the stack down.
+#
+#   ./scripts/smoke.sh                       # build :local and smoke it
+#   BACKEND_IMAGE_TAG=sha-abc123 ./scripts/smoke.sh   # smoke a published image (CI)
+set -euo pipefail
+cd "$(dirname "$0")/.."
+
+export POSTGRES_PASSWORD="${POSTGRES_PASSWORD:-smoke}"
+# The image to smoke, pulled from the registry (no local build). CI passes the just-
+# pushed sha (BACKEND_IMAGE_TAG=sha-<12hex>); locally it defaults to a published tag.
+export BACKEND_IMAGE_TAG="${BACKEND_IMAGE_TAG:-v0.0.2-split-ci}"
+IMG="${REGISTRY_HOST:-git.thermograph.org}/${BACKEND_IMAGE_PATH:-emi/thermograph-backend/app}:${BACKEND_IMAGE_TAG}"
+export SMOKE_HOST_PORT="${SMOKE_HOST_PORT:-18137}"
+COMPOSE=(docker compose -f docker-compose.test.yml)
+PORT_URL="http://127.0.0.1:${SMOKE_HOST_PORT}"
+
+cleanup() { echo "==> tearing down"; "${COMPOSE[@]}" down -v --remove-orphans >/dev/null 2>&1 || true; }
+trap cleanup EXIT
+
+echo "==> pulling $IMG"
+"${COMPOSE[@]}" pull backend
+
+echo "==> starting backend + db ($IMG)"
+"${COMPOSE[@]}" up -d
+
+echo "==> waiting for /healthz (up to 90s)"
+for i in $(seq 1 45); do
+  if curl -fsS -o /dev/null "$PORT_URL/healthz"; then ok=1; break; fi
+  sleep 2
+done
+[ "${ok:-}" = 1 ] || { echo "!! backend never became healthy"; "${COMPOSE[@]}" logs --tail=40 backend; exit 1; }
+
+echo "==> asserting contract"
+curl -fsS "$PORT_URL/healthz" >/dev/null && echo "   /healthz 200 ok"
+ver=$(curl -fsS "$PORT_URL/api/version")
+echo "   /api/version -> $ver"
+echo "$ver" | grep -q '"backend_version"[[:space:]]*:[[:space:]]*"2"' \
+  || { echo "!! /api/version did not report backend_version=2"; exit 1; }
+
+echo "==> SMOKE PASSED"

scripts/test.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+# Reproducible local test runner: build a Python 3.12 dev venv, install the pinned
+# dev deps, and run the hermetic pytest suite. Matches what CI should run.
+#
+# Why 3.12 explicitly: the runtime deps pin versions that need 3.12, and a common box
+# python (pyenv 3.10) ships without the `_sqlite3` extension, which the suite's
+# conftest imports — so we don't just use whatever `python3` is on PATH.
+#
+#   ./scripts/test.sh            # full suite
+#   ./scripts/test.sh tests/web  # pass pytest args through
+set -euo pipefail
+cd "$(dirname "$0")/.."
+
+VENV="${VENV:-.venv-test}"
+
+if [ ! -x "$VENV/bin/pytest" ]; then
+  echo "==> Building dev venv ($VENV) on Python 3.12"
+  if command -v uv >/dev/null 2>&1; then
+    uv venv --python 3.12 "$VENV"
+    VIRTUAL_ENV="$VENV" uv pip install -q -r requirements-dev.txt
+  else
+    PY="$(command -v python3.12 || echo "$HOME/.local/bin/python3.12")"
+    "$PY" -c 'import sqlite3' 2>/dev/null || { echo "!! $PY lacks the sqlite3 module — install a 3.12 with sqlite (e.g. via uv)"; exit 1; }
+    "$PY" -m venv "$VENV"
+    "$VENV/bin/pip" install -q --upgrade pip
+    "$VENV/bin/pip" install -q -r requirements-dev.txt
+  fi
+fi
+
+echo "==> pytest ($("$VENV/bin/python" --version))"
+exec "$VENV/bin/python" -m pytest "$@"