Align backend plan with codebase

This commit is contained in:
adilallo
2026-04-04 22:20:02 -06:00
parent fe54390849
commit c8e930552b
36 changed files with 2216 additions and 2 deletions
+239
View File
@@ -0,0 +1,239 @@
# Backend roadmap (reference)
Temporary working notes for building the backend. Safe to delete once the stack is stable.
---
## 1. Where we are
- **Next.js 16** single repo ([`package.json`](package.json)).
- **PostgreSQL + Prisma**: schema and migrations under `prisma/`; product APIs under `app/api/*` (health, auth/OTP, session, drafts, rules, templates, web-vitals).
- **Server modules** in `lib/server/` (db, session, mail, rate limiting, etc.).
- **Create flow** persists in the browser (`localStorage`); optional **server draft sync** when `NEXT_PUBLIC_ENABLE_BACKEND_SYNC=true` and the user is signed in ([`app/create/context/CreateFlowBackendSync.tsx`](app/create/context/CreateFlowBackendSync.tsx)).
- **Web vitals** [`app/api/web-vitals/route.ts`](app/api/web-vitals/route.ts) still use **file-based** storage under `.next` (not suitable for multi-instance production).
- **CI:** [`.gitea/workflows/ci.yaml`](.gitea/workflows/ci.yaml) (build, test, lint, `prisma validate`); no in-repo production deploy definition.
### HTTP API (implemented in repo)
Mirrors [CONTRIBUTING.md](../CONTRIBUTING.md) **API routes** table; handlers live under `app/api/*/route.ts`.
| Method | Path | Purpose |
| ---------- | ----------------------- | --------------------------------------------- |
| GET | `/api/health` | Liveness / DB check |
| GET | `/api/auth/session` | Current user or null |
| POST | `/api/auth/otp/request` | Send email OTP |
| POST | `/api/auth/otp/verify` | Verify OTP, set session cookie |
| POST | `/api/auth/logout` | Clear session |
| GET / PUT | `/api/drafts/me` | Load or save create-flow JSON (authenticated) |
| GET / POST | `/api/rules` | List or publish rules |
| GET | `/api/templates` | List curated templates |
**Also present (not in CONTRIBUTING table):** `POST` / `GET` [`/api/web-vitals`](../app/api/web-vitals/route.ts) — file-based store today; production path TBD (§7).
---
## 2. What were building
**Step 1.** Treat this as **greenfield**: new **PostgreSQL** database and new tables. Do **not** migrate data from the old Community Rule backend.
**Step 2.** Keep the backend **inside this Next app**:
- HTTP handlers under `app/api/…`
- Shared server code under `lib/server/…`
**Step 3.** Use the old backend only as a **product hint** (email OTP, saving rules, listing rules). Do **not** copy its Express layout or MySQL schema.
---
## 3. Stack choices
**Step 1.** Use **PostgreSQL** everywhere (local Docker, staging, production).
**Step 2.** Use **Prisma**`schema.prisma`, `npx prisma migrate dev` / `migrate deploy`.
**Step 3.** Add **SMTP** (or Mailhog locally) for email OTP in deployed environments; dev can log OTP to the console when `SMTP_URL` is unset.
**Step 4.** **Redis / queues / Kubernetes** — not required for v1. **Exception:** before running **multiple app instances**, plan a **shared rate-limit store** (often Redis) for OTP endpoints; the current limiter is in-memory per process ([`lib/server/rateLimit.ts`](lib/server/rateLimit.ts)).
---
## 4. Data to model (first pass)
Plain-English entities (names can evolve):
| Area | Purpose |
| ----------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
| **User** | Identified by email after OTP verification. |
| **Session** | **Custom v1:** HttpOnly cookie; opaque token; **hash** stored in DB ([`lib/server/session.ts`](lib/server/session.ts)). Not NextAuth/Lucia. |
| **OtpChallenge** | Short-lived email codes (hashed). |
| **RuleDraft** | **One** JSON blob per user (create-flow state). Schema already has **`updatedAt`**; no draft **versioning** or **multiple named drafts** in v1. |
| **PublishedRule** | Saved rule after publish (title, summary, document JSON). |
| **RuleTemplate** | Curated templates (slug, category, ordering). |
**Session follow-ups to implement or decide:** token **rotation** on sensitive events, whether **new login invalidates other sessions**, and **cleanup** of expired `Session` rows (job or lazy delete). Revisit a small auth library (e.g. Auth.js, Lucia) only if maintaining custom code becomes costly.
**RuleDraft future (not v1):** versioning, multiple drafts per user, easier corruption recovery—only if product needs them.
Align JSON shapes with `app/create/types.ts` as it matures.
---
## 5. Session and authentication (v1)
- **Decision:** **Custom** database-backed sessions + email OTP; cookies are **httpOnly**; tokens are hashed at rest.
- **OTP rate limiting:** **In-memory** is acceptable for a **single Node process**. It does **not** coordinate across instances—**add a shared limiter (e.g. Redis)** before horizontal scaling or serious abuse exposure.
- Do **not** treat “switch to NextAuth/Lucia” as required for v1; document the custom lifecycle above instead.
---
## 6. Authorization (v1)
Match the current API behavior; tighten as product evolves:
- **`GET /api/drafts/me` / `PUT /api/drafts/me`:** Authenticated user only; draft is **scoped to that user** (`userId`).
- **`POST /api/rules`:** Authenticated user only; rule is stored with **`userId`** (owner).
- **`GET /api/rules`:** **Public list** of published rules (metadata: id, title, summary, timestamps)—no auth required today. **Not** a private “my rules” feed unless you add a separate route later.
- **v1:** No **editing** or **deleting** published rules via API in the shipped handlers; no **sharing** or **collaborative ownership**—treat each rule as **owned by one user** until product defines more.
---
## 7. API responses, errors, and observability
**Error JSON (target):** Prefer a stable shape, e.g. `{ "error": { "code": "string", "message": "string" }, "details"?: ... }` for 4xx/5xx, instead of only `{ "error": "string" }`. Validation errors can map into `details`. Implement gradually in route handlers.
**Logging:** Use the shared [`lib/logger.ts`](../lib/logger.ts) where possible. Include a **request correlation id** (reuse `x-request-id` if present, else generate) on API routes and log it with errors so support can tie logs together.
**Metrics:** No vendor required for v1; optional later: request duration, error counts.
**Web vitals:** **Default** is **external** RUM or log drain (e.g. host analytics, Vercel Analytics, OpenTelemetry, SaaS APM)—keep **product Postgres** focused on product entities. Storing vitals in Postgres is an **explicit tradeoff** only if ops strongly wants a single datastore.
---
## 8. Prisma migrations policy
- **Never edit** migration files that have **already been applied** to **staging or production** (or any shared database). Fixing schema drift = **add a new migration**.
- **Local dev:** `prisma migrate dev` creates migrations; **deployed envs:** `prisma migrate deploy` before serving new code that depends on the schema.
---
## 9. Build order (implementation steps)
**Operator / local (always manual):** Steps 14 — env file, Docker Postgres, `npm ci`, `prisma migrate dev`, `npm run dev`.
**Backend behavior already in the repo:** Steps **510** match implemented Route Handlers and middleware (`lib/server/*`). **Step 11** (web vitals) is **not** production-ready (files under `.next`); treat as follow-up work aligned with §7.
**Product / frontend still open (not only “backend exists”):** Sign-in UI, wiring publish from the create flow, template seed + UI consumption — see §12 and [docs/backend-linear-tickets.md](backend-linear-tickets.md).
---
**Step 1.** Copy `.env.example` to `.env.local`. Set `DATABASE_URL` and secrets (see file comments).
**Step 2.** Start Postgres locally:
```bash
docker compose up -d postgres
```
**Step 3.** Install dependencies and apply migrations:
```bash
npm ci
npx prisma migrate dev
```
**Step 4.** Run the app:
```bash
npm run dev
```
**Step 5.** Confirm **health**: `GET /api/health` should return JSON.
**Step 6.** **OTP login** (happy path):
1. `POST /api/auth/otp/request` with `{ "email": "you@example.com" }`
2. Read the code from your mail catcher or server logs (dev).
3. `POST /api/auth/otp/verify` with `{ "email": "...", "code": "..." }`
4. `GET /api/auth/session` should show your user.
**Step 7.** **Drafts**: With a session, `GET /api/drafts/me` and `PUT /api/drafts/me` with `{ "payload": { ... } }` (create flow state object).
**Step 8.** **Publish**: `POST /api/rules` with `{ "title", "summary?", "document" }`.
**Step 9.** **Templates** (when ready): seed `RuleTemplate` rows; `GET /api/templates` is implemented.
**Step 10.** **Frontend sync**: Set `NEXT_PUBLIC_ENABLE_BACKEND_SYNC=true` in `.env.local` for server drafts when logged in; `localStorage` remains fallback when off or anonymous.
**Step 11.** **Web vitals:** Move off `.next` files—**prefer an external analytics or logging pipeline** (see §7). Use Postgres for vitals only as a deliberate ops choice.
---
## 10. Security checklist
- **HTTPS** in staging/production; session cookie **Secure**.
- **Rate-limit** OTP (in-memory OK for one instance; **shared store before multi-instance**—see §5).
- **Hash** OTP codes and session tokens before storing; short OTP expiry.
- **Secrets** only in env / secret store — never commit `.env.local`.
- **CORS:** prefer **same-origin** `/api/*`; if cross-origin, configure CORS and CSRF carefully.
---
## 11. Environments
| Environment | Purpose | Notes |
| -------------- | ------------------------------- | --------------------------------------------------------------------- |
| **Local** | Daily development | Docker Compose: Postgres + optional Mailhog (`docker compose up -d`). |
| **Staging** | Rehearse deploys and migrations | Match prod as closely as possible; test SMTP. |
| **Production** | Users | Backups, monitoring, migration job before traffic. |
**Optional QA:** Run automated tests against an **ephemeral** database in CI instead of maintaining a fourth long-lived server.
**Admin / infra (coordinate with whoever runs the server):**
1. TLS certificates and hostnames.
2. PostgreSQL backups and restore drill.
3. SMTP DNS (SPF, DKIM).
4. Health check URL for reverse proxy (`/api/health`).
5. Log retention and alerts for 5xx errors.
---
## 12. Frontend hook-up
**Step 1.** Keep default behavior: **no env flag** → create flow uses **only** `localStorage` (current behavior).
**Step 2.** Set `NEXT_PUBLIC_ENABLE_BACKEND_SYNC=true` to opt in to server drafts when logged in.
**Step 3.** Implement sign-in UI when you are ready: call the OTP routes, then rely on the browser cookie for `/api/drafts/me`.
**Step 4.** On publish, call `POST /api/rules` from the completed step when the backend is required (wire when the final review UI is ready).
---
## 13. Optional later
- **Session library** spike (Auth.js, Lucia) if custom lifecycle cost grows.
- **Redis** (or similar) for **shared OTP rate limits** and horizontal scale.
- **RuleDraft** versioning or multiple drafts per user.
- Standalone **API service** (Fastify/Hono) if scaling or workers demand it.
- **OpenAPI** if external API clients appear.
- **Fourth environment** or stricter rate limiting at the edge.
---
## 14. Useful commands
| Command | When |
| --------------------------------------- | ------------------------------------------------------------------------------------------ |
| `npx prisma studio` | Inspect/edit DB locally. |
| `npx prisma migrate dev` | After changing `schema.prisma` in development. |
| `npx prisma migrate deploy` | Apply migrations in staging/production. |
| `docker compose up -d postgres mailhog` | Local DB + mail UI (http://localhost:8025). |
| `docker build -t community-rule .` | Optional production image (Next **standalone** + `node server.js`; see repo `Dockerfile`). |
---
## External reading
- [Docker: development best practices](https://docs.docker.com/develop/dev-best-practices/)
- [Docker Compose in production](https://docs.docker.com/compose/how-tos/production/)