pig/AGENTS.md

# Working on PIG

Onboarding for an agent or engineer joining this codebase cold. Read this
first, then [`docs/build-plan.md`](./docs/build-plan.md) for what to build.

Named `AGENTS.md` because that is the file coding agents look for by
convention. Everything here applies equally to humans.

---

## 1. What PIG is, in three sentences

A company that aggregates GPU capacity and resells it runs **two pipelines**,
and its business is the spread between them. Generic CRMs model one pipeline
against companies — they have no concept of inventory, no concept of a
commitment you already paid for, and so cannot answer the only question that
matters: *which contracted capacity is sold, to whom, at what margin, and what
is idle right now.*

PIG is built around that question, it is Apache 2.0, and it is designed to be
self-hosted by the customer.

**The load-bearing table is `allocations`**, which joins a `capacity_commitment`
(what we bought, at a known cost) to a `demand_deal` (what we sold, at a known
price). Margin, utilisation and idle capacity all fall out of that one join.
Everything else is plumbing that exists to keep that ledger honest.

---

## 2. Orientation

```
packages/core     Ontology (stages, tiers, enums) + permissions + margin + palette
packages/db       Drizzle schema (47 tables), migrations, seeds
packages/prime    Typed client for the Prime Intellect compute API
apps/api          Hono HTTP API, auth, capacity/contract/calendar services
apps/web          React + Vite + Tailwind + shadcn-idiom components
apps/piggy        The agent — a Prime Agent session over the CRM tools behind a
                  private chat server, plus a lease-based queue worker (§6)
apps/mcp          MCP server (stdio) — 9 tools
apps/cli          `pig`, the HTTP surface for scripts and agent kernels
docs/             ontology.md, build-plan.md, agents.md, seed-data.md,
                  screenshots.md, learn-scripts.md
deploy/           README.md (deployment), Caddyfile example, autodeploy units
```

~45,000 lines including tests. 403 unit tests across five packages
(core 62, prime 24, api 217, piggy 95, cli 5), plus E2E suites under
`apps/api/e2e` and `apps/piggy/e2e` that need a database — and, for one Piggy
case, a key. Node 22+.

| | |
|---|---|
| Repo | `PIG/pig` on git.karti.ai (Gitea) |
| Live | https://primeintellectgrowth.com |
| CI | Gitea Actions, `.gitea/workflows/ci.yml`, ~2 min, must stay green |
| Deploy | Push a `release-*` tag; CI publishes the image and the host's poller pulls it. A push to `main` deploys nothing. `bash scripts/deploy.sh` on the host is the manual path |

---

## 3. Running it

**PIG uses pnpm**, pinned by the `packageManager` field. Do not run `npm
install` — it will write a `package-lock.json` that nothing reads and resolve a
dependency tree that neither CI nor the image uses. Corepack ships with Node and
installs the pinned version for you:

```bash
corepack enable

pnpm install

# Postgres. PIG needs its own database — never point it at a shared one.
docker run -d --name pig-dev -p 5432:5432 \
  -e POSTGRES_USER=pig -e POSTGRES_PASSWORD=pig -e POSTGRES_DB=pig \
  postgres:16-alpine

export DATABASE_URL=postgres://pig:pig@localhost:5432/pig
pnpm run db:migrate
pnpm run db:seed     # sourced, cited people — optional
pnpm run db:demo     # a plausible demo book — optional, prefixed "DEMO — "

pnpm run dev:api     # :8920
pnpm run dev:web     # :5173, proxies /api to 8920
```

With no `SUPABASE_URL` set, **authentication is disabled in development** and
every request runs as the first user in the table. `loadConfig` refuses to start
in production without it, so this cannot leak.

Before pushing:

```bash
pnpm run typecheck && pnpm test
```

---

## 4. Rules that must not be broken

These are load-bearing. Breaking one produces a subtle failure, not an error.

**Intelligence never lives in the API.** Handlers validate, authorize, call a
service, serialise. Research, enrichment, scoring and matching heuristics live
in the service layer or in the agent. The API signals the agent by *writing a
row to `agent_tasks`*, never by calling it — so the queue survives the agent
being down and no request thread ever blocks on a model.

**There are two auth providers, behind one interface.** Supabase for the
hosted deployment, OIDC for on-premises — see `apps/api/src/lib/auth-provider.ts`.
Both reduce to "verify a bearer token, return a subject and an email", because
that is all PIG needs. Never reach for a provider SDK outside that file.

**Authentication is not authorization.** A verified JWT proves someone has an
account in an identity provider that PIG *shares with another application*. It
does not prove they belong here. Access requires a row in PIG's own `users`
table. A token without one gets `403 needs_profile`, which the front end turns
into the join flow — never a login screen they have already completed.

**Cost is charged against the full commitment, not the hours that sold.**
Unsold hours are already paid for. Any other treatment reports a healthy margin
on a block that is losing money, which is precisely the failure PIG exists to
prevent. There is a test pinning this; if you "fix" it, read the test first.

**Sold and held are different.** A live hold removes capacity from everyone
else's availability — otherwise two sellers promise the same GPUs — but it is
not revenue and must never count toward utilisation.

**Money is integer cents.** Never floats. Round, never truncate: 2.43 is
2.4299999 in binary and a lost cent compounds across millions of GPU-hours.

**Agent claims carry evidence.** Enrichment writes to `facts` with a score, a
band, a source URL and a status — never directly to the record. Only `verified`
self-applies; anything weaker waits for a human. An agent allowed to write
unattributed claims will eventually write a wrong one and nobody will be able
to tell which.

**Never invent data about real people.** Seed records carry a confidence grade
and a source URL, both shown in the UI. No email address is ever inferred.
Authorship is not employment — contributors, residency participants and alumni
are recorded as what the evidence shows. Demo data is prefixed `DEMO — ` and
its customers are fictional.

---

## 5. Traps that have already bitten

Each of these cost real time. None produced an error message.

**`onConflictDoNothing()` is a no-op without a matching unique constraint.**
It has silently duplicated seed data twice. If there is no unique index to
conflict on, do an existence check instead.

**`z.coerce.boolean()` turns the string `"false"` into `true`.** Every feature
flag set to false was silently on. Use the `envBoolean` helper in
`apps/api/src/lib/config.ts`.

**Mutations must confirm themselves.** `<Toaster />` is mounted in `App.tsx`
inside `ThemeProvider`; use `toast.success` / `toast.error` in every mutation's
`onSuccess` / `onError`. The shadcn Toaster ships wired to `next-themes`, which
PIG does not use — it was rewired to PIG's `useTheme`. Before that it was never
mounted, so toasts already written in RecordSheets fired into nothing and every
save completed in silence.

**shadcn's `accent` is a SUBTLE surface, not the brand.** shadcn uses
`bg-accent` for hover, focus and selected states — dropdown items, command
rows, ghost buttons. The brand is `primary`. In `tailwind.config.js`, `accent`
is therefore aliased to `--accent-subtle` and `primary` to `--accent`. Use
`bg-primary` for a solid brand fill; never `bg-accent`. Mapping them the other
way makes every hover state paint a full-strength brand block, which in dark
mode with the monochrome palette is a glaring white slab.

**Grid and flex children need `min-w-0`.** They default to
`min-width: auto`, meaning they refuse to shrink below their content — and a
`tabular-nums` figure, a `whitespace-nowrap` badge or a `truncate` title is all
it takes. The page then scrolls sideways on a phone and nothing reports an
error. This was fixed three separate times at individual call sites before
`Card` was given `min-w-0` on its base class; **any new container primitive
needs the same**. Check with:

```js
document.documentElement.scrollWidth - document.documentElement.clientWidth
```

It should be 0 on every route at 393px wide.

**pnpm needs `CI=true` in any non-interactive build.** When it decides a
modules directory is stale it asks before removing it; with no TTY it cannot
ask, so it aborts with `ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY`. This reads
like a pnpm bug and is not — it is pnpm refusing to delete files nobody
confirmed. Both the Dockerfile and CI set it. The usual trigger is a host
`node_modules` reaching the build context, which is why `.dockerignore` exists:
pnpm's tree is symlinks into a content-addressed store, so copying it into an
image yields dangling links and a modules directory pnpm considers corrupt.

**`tsx` is a production dependency, not a dev one.** The server runs TypeScript
directly — `pnpm exec tsx apps/api/src/server.ts` is the container's command —
so pruning it away breaks the image. It lives in `dependencies` deliberately;
moving it back to `devDependencies` because "it's a build tool" makes
`pnpm install --prod` produce an image that cannot start.

**Drizzle-generated migrations are not always valid SQL.** A `jsonb → integer`
cast was emitted without the `USING` clause Postgres requires. Always apply a
new migration to a real empty database before pushing — CI does this, but find
out before CI does.

**The proxy allows exactly one inline script, by hash.** `index.html` carries a
pre-paint theme script that prevents a white flash for dark-mode users. Editing
it changes its hash and the browser silently blocks it. CI asserts the hash;
if it fails, update the CSP in `deploy/Caddyfile.example`, on the server, and
the expected value in the workflow.

**`prices.onDemand` from the Prime Intellect API is the TOTAL FOR THE NODE.**
Verified: 1× A100 at 1.79, 2× A100 at 3.58. `gpuMemory` is likewise a node
total. `packages/prime/src/map.ts` now divides both by `gpuCount` at the
boundary and keeps the node totals in `raw` for reconciliation — this was a
real bug that made an 8-GPU node read eight times too expensive. Anything new
that reads an upstream price must normalise the same way.

**The SPA fallback must never answer an `/api/` path.** Without an explicit
guard, an unknown API route returns `200 text/html` — the app shell — and the
caller sees `response.ok === true` before failing on `JSON.parse` with
"Unexpected token '<'", a long way from the cause. Both the static-file
middleware and the SPA fallback in `apps/api/src/server.ts` carry the guard;
anything added after them needs it too.

**Prime Intellect has two API hosts.** `api.primeintellect.ai` is compute and
pods. Inference is `api.pinference.ai/api/v1`, OpenAI-compatible.

**Piggy's default model thinks aloud, and the harness makes it worse.** The
agent SDK defaults `thinkingLevel` to `medium`; on `nvidia/nemotron-3-nano-30b-a3b`
that produced 6,195 output tokens of reasoning and an *empty* answer. The fix is
two halves and both are needed — see [§6](#6-piggy-and-the-harness-it-runs-on).

**A route file with green tests can still be unmounted.** Every route module is
a factory returning a `Hono` app, and `createApp` has to call it. The tests
mount the factory themselves, so they pass whether or not `app.ts` ever does.
Four modules are in exactly that state right now — `read-guards.ts`,
`learn.ts`, `hubspot.ts`, `hubspot-webhook.ts` — which is why read
authorisation is unenforced and `/learn` answers 404 from a page that is in the
navigation. After adding a route file, curl the path against a running server;
the test suite cannot tell you.

**Deployment traps** live in `deploy/README.md` — chiefly that every Caddy site
block on that host needs `bind 10.0.0.2`, and that the CI runner uses
`container.network: host` so dependencies must be published on `127.0.0.1`.

---

## 6. Piggy, and the harness it runs on

Everything below was learned by running the thing. The product-level account is
in the README under *The agent surface*; this section is the engineering one,
and it exists because much of what follows either contradicts the SDK's own
documentation or is invisible in TypeScript.

### 6.1 The shape

`apps/piggy` embeds **Prime Agent** — Prime Intellect's harness,
`@earendil-works/pi-coding-agent@0.84.1`, MIT — as a Node library. Nothing is
shelled out to, and there is no second process.

```
src/agent/session.ts      Builds a turn: runtime, credential, model, prompt,
                          tools, and the assertions that make the tool set a
                          fact rather than a hope
src/agent/models.json     The provider document the harness reads: five models,
                          their prices, their context windows, their reasoning
                          maps. Copied verbatim into PIGGY_AGENT_DIR when the
                          runtime is first built (once per process)
src/agent/models.ts       Validates that file and turns it into the picker's
                          catalogue. One source for price and size
src/agent/prompt.ts       Piggy's system prompt, including the tool list the
                          harness stops writing (§6.4)
src/agent/tool-bridge.ts  PIG's zod `AgentTool`s → harness `ToolDefinition`s
src/chat-tools.ts         Read tools    ─┐
src/page-tools.ts         Page summaries ├─ the product; the harness swap did
src/lifecycle-tools.ts    Lifecycle     ─┘  not touch a line of them
src/write-tools.ts        The five write tools and the approval flow
src/chat-server.ts        The NDJSON server, the approval rendezvous, the ledger
src/provider.ts + worker.ts + queue.ts    The queue worker, which does NOT use
                                          the harness at all — it still speaks
                                          OpenAI-completions directly
```

The queue worker and the chat agent are different code paths that happen to
share a process. `PIGGY_MODEL` and `PIGGY_INFERENCE_BASE` belong to the worker;
`PIGGY_AGENT_*` and `models.json` belong to the agent. Changing one does not
change the other, which has already confused one person into "fixing" the model
in the wrong place.

### 6.2 No shell, and why the flag is not enough

The session is constructed with `noTools: 'all'` **plus** an explicit `tools`
allowlist (the `createAgentSession` call in `agent/session.ts`). Neither alone
would do:
`noTools: 'all'` removes the built-ins, and the allowlist is the positive
statement of what may exist. But both are *the harness's* configuration, and the
harness composes its tool set from several sources — built-ins, extensions,
skills, custom tools — so a future release that changes the precedence between
them would widen the set without changing a line of PIG. Three gates exist for
that reason:

1. `assertPigToolBoundary` (`src/chat.ts`) — a name must start `pig_` and must
   not read like a shell. PIG's own code, PIG's own rule.
2. `assertUniqueToolNames` (`agent/session.ts`) — the harness keeps its tools in
   a `Map` keyed by name and *sets* each one in turn
   (`dist/core/agent-session.js:1963-1968`), so a duplicate silently overwrites
   the other. That is how a read tool ends up answering for a write tool of the
   same name, with nothing anywhere saying so.
3. `assertExactToolSet` (`agent/session.ts`) — compares the live
   `session.agent.state.tools` against exactly what was handed in and throws at
   session construction if they differ. This is the one that would notice a
   harness upgrade.

`test/agent-session.test.ts` pins all three, including `pig_bash` and friends.
Extensions, skills, prompt templates, themes and context-file discovery are all
disabled on the `DefaultResourceLoader`, and `PIGGY_AGENT_DIR` is deliberately
not a checkout: the harness reads context files from its cwd, and the cwd is
also appended to the live system prompt verbatim as
`Current working directory: …`.

### 6.3 Four places the SDK's own docs are wrong

Each of these compiles, starts, and fails somewhere else.

**`apiKey` in `models.json` is not an environment variable name.** Writing
`"apiKey": "PRIME_API_KEY"` sends the literal string `PRIME_API_KEY` as the
bearer token, and the endpoint answers 401. The value is a *template*:
`$PRIME_API_KEY` or `${PRIME_API_KEY}` interpolate, a leading `!` executes the
rest as a shell command, and anything else is a literal
(`dist/core/resolve-config-value.js:116-128`). PIG uses none of those forms —
it calls
`modelRuntime.setRuntimeApiKey(PIGGY_PROVIDER_ID, config.PRIME_API_KEY)`
(`agent/session.ts`), which is the only line that authenticates Piggy and keeps
the key out of the file that gets written to disk.

**There is no built-in `prime-inference` provider in 0.84.1.** The published
docs describe a build that is not on npm; `KnownProvider` in
`@earendil-works/pi-ai/dist/types.d.ts:19` lists forty providers and none of
them is Prime Intellect's inference host. PIG registers one itself from
`models.json`, and the id `prime-inference` has to match in three places — the
JSON key, `setRuntimeApiKey`, and `modelRuntime.getModel`. A typo in any of them
surfaces as a 401 or an undefined model, never as "unknown provider".

**A `ResourceLoader` you pass in is never reloaded for you.**
`createAgentSession` constructs and reloads one *only when you do not supply
one* (`dist/core/sdk.js:75-78`). Pass your own and forget `await loader.reload()`
and the session runs on the stock coding-assistant preamble — no error, no
warning, and an agent that offers to read your files.

**The stock prompt is a coding-assistant prompt and must be replaced, not
appended to.** It opens "You are an expert coding assistant operating inside pi"
and cites the SDK's own README paths (`dist/core/system-prompt.js:73`).
Appending does not help: a CRM agent told it edits code reaches for tools it
does not have and apologises for not having them. The replacement goes through
the loader's `systemPromptOverride`, which takes the literal text — the
`systemPrompt` option is a *file source*, and handing it a prompt loads nothing
and says nothing.

### 6.4 Replacing the prompt silently removes the tool list

`buildSystemPrompt` returns early on the `customPrompt` branch
(`dist/core/system-prompt.js:13-33`); the "Available tools" section is only ever
built further down, on the branch where no custom prompt was supplied
(`:40`, `:75`). So the moment the preamble is replaced — which is not optional
here — every tool becomes invisible to the model, `promptSnippet` or not.

`agent/prompt.ts` therefore renders the list itself, in `toolSection`. A 30B
model that cannot see a tool in its prompt answers from the page title instead
of calling it, and that failure is completely silent: the tool is registered,
callable, and never called. If you add a tool, give it a `promptSnippet`, and
check it appears in `session.systemPrompt`.

### 6.5 The thinking-level trap

The one that cost real money.

The harness defaults `thinkingLevel` to `medium`. On the default model that
produced **6,195 output tokens of reasoning and an empty answer**, stopping at
`finish_reason: length` — the budget was gone before a word of the reply was
written, and reasoning bills as output. `low` was worse. After the fix the same
question answered correctly in **149 output tokens**.

The fix is two halves and either alone is silent:

- `PIGGY_AGENT_THINKING` defaults to `off` (`src/config.ts`), and
- the model entry carries a `thinkingLevelMap` mapping `off` → `"none"`
  (`src/agent/models.json`).

Why the second is needed: a thinking level of `off` becomes
`reasoningEffort: undefined` in the provider
(`@earendil-works/pi-ai/dist/api/openai-completions.js:473-474`), and the
request builder then emits `reasoning_effort` **only if the model has a map**:

```js
else if (options?.reasoningEffort && model.reasoning && compat.supportsReasoningEffort) {
    params.reasoning_effort = model.thinkingLevelMap?.[options.reasoningEffort] ?? options.reasoningEffort;
}
else if (!options?.reasoningEffort && model.reasoning && compat.supportsReasoningEffort) {
    const offValue = model.thinkingLevelMap?.off;
    if (typeof offValue === "string") { params.reasoning_effort = offValue; }
}
                                    — dist/api/openai-completions.js:657-666
```

Without the map, `off` sends **no reasoning parameter at all** and the
endpoint's own default — thinking on, verbosely — wins. This is per model. The
two nemotron entries have a map; deepseek, opus and gpt-5.6 do not, and were
left to their own defaults deliberately. **If you change `PIGGY_AGENT_MODEL` and
answers start coming back empty or truncated, this is why.**
`test/agent-thinking.test.ts` fails if the default model has no map, and
`e2e/prime-agent.test.ts` counts the tokens against the live endpoint.

### 6.6 Modes, and the one function that decides

`PiggyMode` is `read_only` | `confirm` | `auto`.

- **`read_only`** offers no write tool at all. Not offered-and-refused: absent
  (`createPigWriteTools` returns `[]`). A model that can see a capability
  narrates using it.
- **`confirm`** — the shipped default — turns every write into a proposal. The
  tool emits an `approval_required` card, the turn stays open, the decision
  arrives on a separate `POST /internal/approve`, and only then does the
  mutation run.
- **`auto`** writes immediately, as the calling user, under their permissions.

Contracts, commitments, allocations and compliance require a human in **every**
mode. That rule is one function — `requiresApproval` in
`packages/core/src/piggy-protocol.ts` — and it is the single source of truth:
the write tools read it, the tests assert against it, and nothing restates it.
If you add a guarded kind, add it to `PIGGY_ALWAYS_CONFIRM_KINDS` and everything
downstream follows.

Two properties of the write path are not negotiable. Every write goes through
`executeMutation` with the caller's own `Principal`, so Piggy holds no privilege
of its own — there is no elevated principal anywhere in `write-tools.ts` and
there must never be one. And a refusal is an *answer*: a missing capability, a
declined card and a rejected input all come back as ordinary tool results whose
first line says `NOT SAVED`. Thrown into the stream they would end the turn on
the user's own permissions, which reads to them as Piggy being broken.

The rendezvous itself (`ApprovalRegistry` in `src/chat-server.ts`) is single-use
— an id is deleted the instant it settles, so a replayed decision cannot apply a
change twice — deadlined at five minutes, and turn-owned: an abandoned turn
rejects every approval it opened, because a pending promise there holds a billed
inference connection open.

### 6.7 The browser never learns which harness this is

Prime Agent emits twenty-three event types. PIG's own protocol
(`PiggyChatEvent` in `packages/core/src/piggy-protocol.ts`) has nine, and
`translateSessionEvent` in `src/chat-server.ts` maps exactly four of the
harness's — `message_update`, `tool_execution_start`, `tool_execution_end`,
`turn_end` — and drops the rest on the server. That is deliberate: a harness
upgrade is then a server change and never a client one.

The risk in a `default: return` is the upgrade that *adds* an event — a
delegated sub-agent, a permission request — which would be dropped in silence
for as long as it took somebody to notice a missing feature.
`test/chat-server.test.ts` therefore writes out both lists and asserts, at
compile time, that they are mutually assignable with `AgentSessionEvent['type']`.
Bump the SDK and `tsc` tells you what is new before anything runs.

### 6.8 Working on Piggy without spending credit

Almost all of it is free, and only one path is not.

- **The unit suite never makes a request.** `createPiggySession` resolves the
  model, builds the prompt and registers the tools entirely offline with a fake
  key, so the tool set, the prompt, the thinking level and the model's own
  ceiling are all inspectable without inference. That is what
  `test/agent-session.test.ts` and `test/agent-thinking.test.ts` do.
- **The whole chat protocol is drivable with no model at all.**
  `startPiggyChatServer` takes `createSession`, `createReadTools` and
  `createWriteTools` as options; the tests hand it a fake harness that emits
  real `AgentSessionEvent`s. `e2e/approval-rendezvous.test.ts` does this against
  a real database, which is how the approval flow is tested end to end for free.
- **`src/dev/mock-inference.ts`** (`pnpm -F @pig/piggy run dev:mock`, port 8945)
  speaks the OpenAI-compatible wire protocol with steering directives —
  `/mock error`, `/mock ratelimit`, `/mock cut`, `/mock badtool`. Note what it
  serves: the **queue worker**, through `PIGGY_INFERENCE_BASE`. The agent reads
  its base URL from `models.json`, so pointing the chat path at the mock means
  editing that file.
- **`src/dev/verify-prime-agent.ts`**
  (`pnpm -F @pig/piggy exec tsx src/dev/verify-prime-agent.ts [modelId]`) is the
  live probe: it asks the real endpoint one question with a seeded tool and
  prints the model, the tool set, whether anything shell-shaped survived, the
  first 200 characters of the system prompt and the answer. It spends a few
  hundred tokens. Nothing in CI runs it.
- **The database.** Anything that writes runs against a scratch database, never
  the development book — an activity appearing in somebody's feed because a test
  ran is exactly what a CRM must not do. `e2e/write-tools.test.ts` and
  `e2e/approval-rendezvous.test.ts` take `PIGGY_WRITE_DATABASE_URL` and refuse
  `pig_combined` by name.
- **The one paid test** is `e2e/prime-agent.test.ts`, gated on
  `PIGGY_E2E_LIVE=1` *and* a key, because a suite that spends money whenever the
  environment happens to be loaded spends money by accident. One turn is about
  $0.0003.

```bash
# Unit suite: no database, no key, no network.
pnpm -F @pig/piggy run typecheck && pnpm -F @pig/piggy run test

# E2E: a scratch database of its own. `pig_combined` is refused by name.
docker exec pig-ux-db psql -U pig -d postgres -c "CREATE DATABASE pig_scratch"
DATABASE_URL=postgres://pig:pig@localhost:54330/pig_scratch pnpm -F @pig/db run migrate
DATABASE_URL=postgres://pig:pig@localhost:54330/pig_scratch \
  PIGGY_WRITE_DATABASE_URL=postgres://pig:pig@localhost:54330/pig_scratch \
  pnpm -F @pig/piggy run test:e2e          # the live case skips, and says so

# Add the paid one deliberately, never by default.
PIGGY_E2E_LIVE=1 PRIME_API_KEY=... pnpm -F @pig/piggy run test:e2e
```

---

## 7. Conventions

**Comments explain *why*, never *what*.** The code says what it does. Comments
carry the reasoning that would otherwise be lost — why this treatment and not
the obvious one, what breaks if it changes, what was tried and rejected. Match
the density already in the file; do not add narration.

**Match the surrounding style.** British spelling in prose and comments
(`utilisation`, `normalise`). Types are explicit at module boundaries. No
default exports.

**Enums come from `@pig/core`.** Never retype a stage list or a tier into a zod
schema — import it, so removing a value stops validating rather than silently
persisting.

**Tests pin decisions, not implementations.** The valuable cases are the ones
that would pass under a plausible-but-wrong version. Look at
`packages/core/test/margin.test.ts` for the register.

**Commits explain the reasoning**, including what was tried and rejected, and
say plainly when something was found by running the code rather than reading
it. Read `git log` for the register.

**Verify by running, not by assuming.** Several of the worst defects here were
found only by fetching a URL from another machine or applying a migration to a
real database. "It should work" has been wrong repeatedly.

---

## 8. Where to start

**Every task in the original three-wave plan has shipped.**
[`docs/build-plan.md`](./docs/build-plan.md) is now an audited record of that
rather than a queue, and it carries the remaining work at the bottom. The two
interfaces everything else codes against — `packages/core/src/permissions.ts`
(the RBAC model) and `apps/api/src/lib/mutation.ts` (the write path) — are
settled; read them before adding any write.

**The highest-value work now, in order:**

1. **Mount `createReadGuardRoutes`.** The read half of the permission model is
   written, tabulated and tested, and does nothing, because `app.ts` never
   mounts it. Until it does, every authenticated member can read supplier cost
   and margin. It is one line, and it must be registered *before* the handlers
   it guards — Hono runs matched handlers in registration order.
2. **Mount `learn.ts`.** `/learn` is in the navigation and its API answers 404.
3. **Enqueue the other six agent task kinds.** The worker is complete; only
   `enrich_account` and `enrich_contact` are ever written to `agent_tasks`, so
   Piggy does far less than the ontology implies.
4. **Mount the HubSpot routes, or delete them.** Seven tables, OAuth, sync jobs
   and webhook verification, all written, tested and unreachable.
5. **Give Piggy's writes their notification.** A stage change made through the
   API raises a Slack notification; the same change made in chat does not,
   because `write-tools.ts` passes no `NotificationOutbox` — it runs in the
   Piggy process and the outbox is wired in the API server. The other open
   Piggy items are listed under *Left to do* in the build plan.

`app.ts` is the one shared file. If your change needs a route mounted, a public
path allowlisted or a schema widened there, say so rather than racing another
agent for it.

---

## 9. What not to do

- Do not copy component files out of other people's repositories. Where a
  primitive is a shadcn/ui original, take it from upstream, where it is
  canonical and current. Compositions we write ourselves.
- Do not open self-registration on the identity provider. It is shared with
  another application. PIG mints accounts itself, gated on an invite.
- Do not put a production SSH key on the CI runner. Deployment is manual on
  purpose.
- Do not weaken the production guard that refuses to start without an identity
  provider.
- Do not seed or infer email addresses for real people.