fix: frontend port exposition

This commit is contained in:
Ivan Rezchikov
2026-07-11 20:02:33 +03:00
parent 7ad9cff390
commit 1046929e2c
3 changed files with 1 additions and 256 deletions
-176
View File
@@ -1,176 +0,0 @@
# Kotosoft ID - AI System Documentation
This document is written for AI assistants (and developers) that need a fast, practical understanding of this repository.
## What this system is
Kotosoft ID is an identity and OAuth service with:
- a Go backend (`kotosoft-id-backend`) exposing OAuth and user/profile APIs,
- a Next.js frontend (`kotosoft-id-frontend`) handling login/consent UI and token cookies,
- PostgreSQL + Redis as dependencies.
Main use cases:
- OAuth2 Authorization Code + PKCE login flow,
- access/refresh token issuance and refresh,
- user profile management,
- consent management,
- role/permission ("positions") administration.
## Repository structure
- `kotosoft-id-backend`: Go API service (Gin, pgx, redis, JWT logic)
- `kotosoft-id-frontend`: Next.js app (App Router, server actions)
- `docker-compose.yml`: production-style compose wiring backend + frontend images
- `.drone.yml`: CI/CD pipeline (backend tests, frontend lint, image build/push, deploy)
## Runtime architecture (high-level)
1. Frontend starts OAuth initiation (`/api/auth/initiate`) and creates PKCE verifier/state cookies.
2. Browser is redirected to backend `/oauth/authorize`.
3. Backend creates OAuth session and redirects to frontend `/oauth?session=...`.
4. Frontend login/consent UI calls backend session endpoints (`/oauth/session/*`).
5. Backend returns redirect URL with authorization code.
6. Frontend callback exchanges code at backend `/oauth/token`.
7. Frontend stores access/refresh tokens in cookies and calls protected APIs.
8. On 401, frontend attempts refresh token flow.
## Backend API surface
Key groups from router:
- OAuth routes (`/oauth`)
- `GET /authorize`
- `GET /session/:token`
- `POST /session/login`
- `POST /session/attach`
- `POST /session/approve`
- `POST /token`
- `POST /revoke`
- `POST /introspect`
- User routes (`/api/v1`)
- `POST /auth/register`
- `POST /auth/refresh`
- `GET /me`
- `PATCH /me`
- `DELETE /me`
- `GET /me/consents`
- `DELETE /me/consent/:clientId`
- `GET /me/positions`
- Admin routes (`/api/v1/admin`)
- CRUD for permissions
- CRUD for positions
- assign/revoke positions to users
Security notes:
- protected routes require Bearer access token,
- scopes are checked in middleware,
- admin group additionally requires admin middleware,
- register/login are rate-limited.
## Required environment variables
### Backend
- `SERVER_ADDR` (default `:8080`)
- `FRONTEND_URL` (required)
- `POSTGRES_URL` (required)
- `POSTGRES_MIGRATION_FILE_PATH` (required)
- `REDIS_URL` (required)
- `JWT_SECRET` (required)
- `JWT_TTL` (default `24h`)
- `CLIENT_ID` (required)
- `CLIENT_SECRET` (required)
- `CLIENT_NAME` (optional, default `Basic Client`)
- `CLIENT_REDIRECT_URIS` (comma-separated)
- optional: `CLIENT_GRANT_TYPES`, `CLIENT_SCOPES`
### Frontend
- `PUBLIC_URL` (used for callback URI generation)
- `KOTOSOFT_ID_API_URL` (backend base URL)
- `CLIENT_ID`
- `CLIENT_SECRET`
## How to run locally (development)
Prerequisites:
- Go `1.25.x`
- Node `20+`
- PostgreSQL
- Redis
### Backend
1. `cd kotosoft-id-backend`
2. Set required env vars (or create `.env`)
3. Run:
- `go mod tidy`
- `go run ./cmd`
On startup backend:
- connects to Postgres/Redis,
- runs SQL migrations from `migrations/migrations.sql`,
- seeds/upserts OAuth client from env vars.
### Frontend
1. `cd kotosoft-id-frontend`
2. Install deps: `npm install`
3. Set env vars (`PUBLIC_URL`, `KOTOSOFT_ID_API_URL`, `CLIENT_ID`, `CLIENT_SECRET`)
4. Run dev server: `npm run dev`
## How to run in containers
The provided `docker-compose.yml` expects prebuilt images:
- `localhost:5000/kotosoft_id_backend:latest`
- `localhost:5000/kotosoft_id_frontend:latest`
Typical deployment flow is in `.drone.yml`:
- test backend,
- lint frontend,
- build and push both images,
- deploy with `docker compose up -d`.
## Is it functional?
Short answer: **yes, likely functional as a working auth system**, with normal production caveats.
Why this is likely:
- backend has organized routing/handlers/services/repositories,
- OAuth code flow + refresh + revoke + introspect are implemented,
- user profile and consent endpoints are implemented,
- admin permission/position features are implemented,
- CI runs `go test ./...` for backend and `npm run lint` for frontend.
Caveats:
- no explicit end-to-end test workflow is defined in CI,
- local functionality still depends on correct Postgres/Redis/env configuration,
- deployment compose relies on external image registry and external `proxy` network.
## Important operational notes for AI agents
- Do not assume public API docs exist; use backend router/handlers as source of truth.
- Frontend auth relies on server-side environment variables and secure cookie handling.
- Many failures are configuration failures (wrong redirect URI, client secret, DB/Redis URLs).
- If debugging auth, verify in this order:
1. env vars, 2) OAuth session lifecycle, 3) callback exchange, 4) cookie presence, 5) token refresh.
- For permission issues, check both access token scopes and user admin flag.
## Suggested next docs to add
If you want this project to be easier for both humans and AIs, add:
- a root `README.md` with quick-start commands,
- an `.env.example` for backend and frontend,
- a simple API reference table with request/response examples,
- an end-to-end "happy path" auth walkthrough.
-79
View File
@@ -1,79 +0,0 @@
# Code Style Profile (Go + JSX/Next.js)
This document captures your current coding style as seen in the repository and lists questions to refine your preferred style rules.
## Go Style (Observed)
### What is consistent
- Clear layered architecture: `handlers -> services -> repositories`.
- Short, readable functions with early returns over deep nesting.
- Explicit request/response DTO structs in handlers with Gin binding tags.
- Business logic mostly lives in `services`, SQL mostly in `repositories`.
- Error mapping style is pragmatic (`models.ErrBadRequest`, `ErrForbidden`, etc.).
- Naming is mostly clear and domain-driven (`AssignPosition`, `RecordUserConsent`, `GetUserPositions`).
### Style characteristics
- You prefer practical over abstract code (few interfaces beyond service boundaries).
- You accept some inline structs for one-off request payloads.
- You keep SQL as raw strings in repository methods (not hidden behind ORM abstractions).
- You usually avoid over-commenting; code is expected to be self-explanatory.
- You follow `gofmt` defaults and avoid stylistic tricks.
## JSX / Next.js Style (Observed)
### What is consistent
- Functional components with hooks (`useState`, `useForm`, `useTranslations`).
- Strong TypeScript typing for props and form data.
- UI composition via reusable primitives (`Dialog`, `Table`, `Button`, custom fields).
- Local optimistic state updates after server actions.
- Short handlers with readable names (`handleCreated`, `handleUpdated`, `handleDelete`).
- Frequent early exits on failure (`if (!result.success) { ... return }`).
### Style characteristics
- You favor compact code and inlined JSX in places where intent is obvious.
- You rely on server actions for API calls and keep transport details in `actions/*`.
- You use translation keys heavily (good i18n discipline).
- You keep component files focused on one domain workflow.
## Your Answers (Captured)
### Go
- Error philosophy: undecided for now.
- Strict types over `any`: undecided for now.
- Authorization model: should be based on explicit permissions; admin role is a temporary service mechanism.
- Zero affected rows on update/delete: return `not_found` by default.
- SQL approach: keep inline raw SQL in this project.
### JSX / Next.js
- Component formatting: compact by default; multiline only when complexity justifies it.
- Mutation strategy today: optimistic updates currently used.
- Validation style: prefer schema-based validation over ad-hoc inline checks.
## Pending Clarifications
1. **Optimistic vs refetch default (Q7):**
Recommendation: keep optimistic update as default for snappy UX, and use refetch for high-risk operations (complex server-side side effects, computed fields, permission-dependent payloads).
2. **Display naming policy (Q8, clarified):**
Proposed rule: in UI, show user-facing label first (`name` or `description`) and fallback to `code` only when label is empty.
Please confirm if you want this as a global rule.
## Final Summary (Current)
### Confirmed style rules
- Go authorization should move toward explicit permissions, not role-only checks.
- Go repositories should return `not_found` for update/delete when `0 rows affected`.
- Go SQL stays inline/raw in this project.
- JSX/Next components should be compact by default; multiline formatting is allowed when readability needs it.
- Frontend validation direction is schema-first.
### Recommended defaults
- Mutation UX: optimistic updates by default; refetch only for risky mutations with possible server-side divergence.
- Action results: migrate from bare `{ success }` to richer result objects:
- success path: `{ success: true, data?: T }`
- error path: `{ success: false, errorCode: string, message?: string }`
### Still open
- Go error-wrapping strictness (`fmt.Errorf("...: %w", err)` everywhere vs selective).
- Strict ban of `any` in all service/repository boundaries.
- Global UI naming fallback rule (`name/description` first, `code` as fallback).
+1 -1
View File
@@ -14,6 +14,6 @@ COPY --from=builder /app/public ./public
COPY --from=builder /app/messages ./messages
RUN npm prune --omit=dev
EXPOSE 80
CMD ["npm", "start", "--", "-p", "80", "-H", "0.0.0.0"]
CMD ["npm", "start", "--", "-p", "3000", "-H", "0.0.0.0"]
HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
CMD wget -q --spider http://localhost:80/ || exit 1