The monorepo was on npm workspaces. pnpm gives it a content-addressed store
shared between the eight packages, a lockfile that records the whole graph
rather than a flattened view of it, and — the reason this mattered in practice —
`workspace:*`, which makes an internal dependency unambiguous instead of a
version range that npm may satisfy from the registry.
Mechanics:
- `packageManager: pnpm@11.21.0` pins the version; corepack installs it in CI
and in the image, so all three environments resolve identically.
- The npm `workspaces` array is replaced by `pnpm-workspace.yaml`. pnpm
ignores the former, and keeping both would leave two sources of truth.
- All six internal dependencies moved to `workspace:*`.
- Root scripts use `pnpm -r --if-present` and `pnpm -F <pkg>`.
Two findings worth recording, both from running it rather than reading it:
`tsx` was a devDependency, but the server runs TypeScript directly in
production — the container's command is `pnpm exec tsx apps/api/src/server.ts`.
Under npm this was concealed by the runtime stage re-installing tsx by hand
after pruning dev dependencies. Under `pnpm install --prod` that sleight of
hand stops working and the image simply fails to start. tsx is now declared in
`dependencies`, which is what it has always actually been.
The first image build failed with ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY.
That is not a pnpm bug: it had decided the modules directory was stale and
wanted confirmation before deleting it, which a non-interactive build cannot
give. The trigger was the host's `node_modules` reaching the build context —
there was no `.dockerignore` at all. pnpm's tree is symlinks into a
content-addressed store, so copying it into an image produces dangling links
and a directory pnpm rightly considers corrupt. Fixed by adding
`.dockerignore` and setting `CI=true`, which is required in any non-interactive
pnpm build.
`esbuild` is denied install scripts via `allowBuilds`. Its platform binary
arrives through the optional dependency `@esbuild/linux-x64` and the postinstall
only verifies it; confirmed by running the binary directly, which reports
0.25.12.
Verified under pnpm: typecheck clean, 150 tests / 0 failures, e2e passes, web
builds. The image was built and booted against a real Postgres — health ok,
`/api/dashboard` 401 with an issuer configured, `/` and `/capacity` serve the
SPA, `/og.png` serves as image/png, and the migrator runs from the pruned
runtime stage.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`npm test` did nothing until now. CI that runs no tests is theatre, so the
tests came first — 39 of them, over the two places where an error would be
silent and expensive.
packages/core: the margin arithmetic. Every dashboard figure, idle-capacity
alert and agent answer resolves through it, and wrong numbers still look like
numbers. The cases pin decisions rather than implementation: cost is charged
against the full commitment (a naive version reports the opposite sign on a
loss-making block), aggregation sums cents rather than averaging percentages
(averaging reports +22% on a book that is losing money), break-even prices the
remaining hours and returns null rather than Infinity when there are none, and
internal research burn counts as cost with no revenue.
packages/prime: the upstream mapping. Rounding rather than truncating cents,
because 2.43 is 2.4299999 in binary and a lost cent compounds across millions
of GPU-hours. And interconnect normalisation, where an unrecognised fabric maps
to Unknown rather than Ethernet — guessing low loses a deal, guessing high
sells a training customer a cluster that cannot train.
CI runs on push and pull request: typecheck all six packages, unit tests,
migrations applied twice to a real Postgres, a seed-idempotency assertion that
fails the build if row counts move on a second run, a server boot, the front-end
build, and a Docker build.
It also asserts the inline theme script's hash still matches the CSP the proxy
allows. That script prevents a white flash for dark-mode users; if it changes
without the CSP being updated, the browser silently blocks it and nothing
anywhere reports an error.
Deployment stays a script rather than push-to-deploy. Automating it would put
an SSH key with production write access on the CI runner — a real escalation
for a project this size. The script takes a database dump before migrating and
refuses to finish if an unauthenticated request returns anything but 401.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
packages/prime — a hand-written typed client, because the first-party SDK is
Python only. Deliberately narrow: PIG reads availability and nothing else, and
the key it holds should be scoped so it could not provision even if the code
tried. Rate limits are undocumented upstream, so it backs off empirically with
full jitter and honours Retry-After. Unknown fields survive in `raw` rather
than being dropped.
apps/api — Hono, with authentication and authorization kept firmly apart. A
verified JWT proves someone has an account in the identity project, which may
be shared with other applications; it does NOT prove they belong here. Access
requires a row in PIG's own users table, and a token without one gets 403
needs_profile rather than entry.
The capacity service is the business logic: availability counts sold and held
separately, so a live hold removes inventory from everyone else's availability
without inflating utilisation. Expired holds are ignored at read time, so the
numbers stay right even when the sweeper is behind. Matching treats
interconnect as a hard filter and excludes Unknown as well as Ethernet —
unverified is not the same as adequate.
apps/mcp — nine tools over stdio, so a team member drives PIG from Claude
Code, Codex, prime-agent, or a Buzz agent. It holds an API key and calls the
same HTTP API the browser does, with no database credentials, so an agent can
never reach further than the person it acts for. Results are formatted as
prose rather than raw JSON.
Theme preferences live in the database rather than localStorage, so a chosen
accent follows someone from laptop to phone. Status colours stay independent
of the accent: if "at risk" re-tinted to whatever a user picked, the signal
would be gone.
Note on the SDK import: its package exports use a `./*` wildcard whose types
entry resolves server/mcp.js to server/mcp.js.d.ts, which does not exist. The
runtime specifier must keep the .js suffix, so the types are mapped via
tsconfig paths rather than by writing an import that would fail at runtime.
Verified: all five packages typecheck; the MCP server constructs and registers
its tools.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
karti