Rewrite CLAUDE.md to reflect full monorepo

Replaces the old art-dag-only docs with comprehensive documentation
covering all web platform services, shared library, art DAG subsystem,
architecture patterns, auth, inter-service communication, and dev/deploy.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-02-27 09:17:30 +00:00
parent 1a74d811f7
commit 22460db450

151
CLAUDE.md
View File

@@ -1,6 +1,6 @@
# Art DAG Monorepo
# Rose Ash Monorepo
Federated content-addressed DAG execution engine for distributed media processing with ActivityPub ownership and provenance tracking.
Cooperative web platform: federated content, commerce, events, and media processing. Each domain runs as an independent Quart microservice with its own database, communicating via HMAC-signed internal HTTP and ActivityPub events.
## Deployment
@@ -9,71 +9,134 @@ Federated content-addressed DAG execution engine for distributed media processin
## Project Structure
```
core/ # DAG engine (artdag package) - nodes, effects, analysis, planning
l1/ # L1 Celery rendering server (FastAPI + Celery + Redis + PostgreSQL)
l2/ # L2 ActivityPub registry (FastAPI + PostgreSQL)
common/ # Shared templates, middleware, models (artdag_common package)
client/ # CLI client
test/ # Integration & e2e tests
blog/ # Content management, Ghost CMS sync, navigation, WYSIWYG editor
market/ # Product catalog, marketplace pages, web scraping
cart/ # Shopping cart CRUD, checkout (delegates order creation to orders)
events/ # Calendar & event management, ticketing
federation/ # ActivityPub social hub, user profiles
account/ # OAuth2 authorization server, user dashboard, membership
orders/ # Order history, SumUp payment/webhook handling, reconciliation
relations/ # (internal) Cross-domain parent/child relationship tracking
likes/ # (internal) Unified like/favourite tracking across domains
shared/ # Shared library: models, infrastructure, templates, static assets
artdag/ # Art DAG — media processing engine (separate codebase, see below)
```
### Shared Library (`shared/`)
```
shared/
models/ # Canonical SQLAlchemy ORM models for all domains
db/ # Async session management, per-domain DB support, alembic helpers
infrastructure/ # App factory, OAuth, ActivityPub, fragments, internal auth, Jinja
services/ # Domain service implementations + DI registry
contracts/ # DTOs and service protocols
browser/ # Middleware, Redis caching, CSRF, error handlers
events/ # Activity bus + background processor (AP-shaped events)
config/ # YAML config loading (frozen/readonly)
static/ # Shared CSS, JS, images
templates/ # Base HTML layouts, partials (inherited by all apps)
```
### Art DAG (`artdag/`)
Federated content-addressed DAG execution engine for distributed media processing.
```
artdag/
core/ # DAG engine (artdag package) — nodes, effects, analysis, planning
l1/ # L1 Celery rendering server (FastAPI + Celery + Redis + PostgreSQL)
l2/ # L2 ActivityPub registry (FastAPI + PostgreSQL)
common/ # Shared templates, middleware, models (artdag_common package)
client/ # CLI client
test/ # Integration & e2e tests
```
## Tech Stack
Python 3.11+, FastAPI, Celery, Redis, PostgreSQL (asyncpg for L1), SQLAlchemy, Pydantic, JAX (CPU/GPU), IPFS/Kubo, Docker Swarm, HTMX + Jinja2 for web UI.
**Web platform:** Python 3.11+, Quart (async Flask), SQLAlchemy (asyncpg), Jinja2, HTMX, PostgreSQL, Redis, Docker Swarm, Hypercorn.
**Art DAG:** FastAPI, Celery, JAX (CPU/GPU), IPFS/Kubo, Pydantic.
## Key Commands
### Testing
### Development
```bash
cd l1 && pytest tests/ # L1 unit tests
cd core && pytest tests/ # Core unit tests
cd test && python run.py # Full integration pipeline
./dev.sh # Start all services + infra (db, redis, pgbouncer)
./dev.sh blog market # Start specific services + infra
./dev.sh --build blog # Rebuild image then start
./dev.sh down # Stop everything
./dev.sh logs blog # Tail service logs
```
- pytest uses `asyncio_mode = "auto"` for async tests
- Test files: `test_*.py`, fixtures in `conftest.py`
### Linting & Type Checking (L1)
### Deployment
```bash
cd l1 && ruff check . # Lint (E, F, I, UP rules)
cd l1 && mypy app/types.py app/routers/recipes.py tests/
./deploy.sh # Auto-detect changed apps, build + push + restart
./deploy.sh blog market # Deploy specific apps
./deploy.sh --all # Deploy everything
```
- Line length: 100 chars (E501 ignored)
- Mypy: strict on `app/types.py`, `app/routers/recipes.py`, `tests/`; gradual elsewhere
- Mypy ignores imports for: celery, redis, artdag, artdag_common, ipfs_client
### Docker
### Art DAG
```bash
docker build -f l1/Dockerfile -t celery-l1-server:latest .
docker build -f l1/Dockerfile.gpu -t celery-l1-gpu:latest .
docker build -f l2/Dockerfile -t l2-server:latest .
./deploy.sh # Build, push, deploy stacks
cd artdag/l1 && pytest tests/ # L1 unit tests
cd artdag/core && pytest tests/ # Core unit tests
cd artdag/test && python run.py # Full integration pipeline
cd artdag/l1 && ruff check . # Lint
cd artdag/l1 && mypy app/types.py app/routers/recipes.py tests/
```
## Architecture Patterns
- **3-Phase Execution**: Analyze -> Plan -> Execute (tasks in `l1/tasks/`)
- **Content-Addressed**: All data identified by SHA3-256 hashes or IPFS CIDs
- **Services Pattern**: Business logic in `app/services/`, API endpoints in `app/routers/`
- **Types Module**: Pydantic models and TypedDicts in `app/types.py`
- **Celery Tasks**: In `l1/tasks/`, decorated with `@app.task`
- **S-Expression Effects**: Composable effect language in `l1/sexp_effects/`
- **Storage**: Local filesystem, S3, or IPFS backends (`storage_providers.py`)
- **Inter-Service Reads**: `fetch_data()` → GET `/internal/data/{query}` (HMAC-signed)
- **Inter-Service Actions**: `call_action()` → POST `/internal/actions/{name}` (HMAC-signed)
- **Inter-Service AP Inbox**: `send_internal_activity()` → POST `/internal/inbox` (HMAC-signed, AP-shaped activities for cross-service writes with denormalized data)
### Web Platform
## Auth
- **App factory:** `create_base_app(name, context_fn, before_request_fns, domain_services_fn)` in `shared/infrastructure/factory.py` — creates Quart app with DB, Redis, CSRF, OAuth, AP, session management
- **Blueprint pattern:** Each blueprint exposes `register() -> Blueprint`, handlers stored in `_handlers` dict
- **Per-service database:** Each service has own PostgreSQL DB via PgBouncer; cross-domain data fetched via HTTP
- **Alembic per-service:** Each service declares `MODELS` and `TABLES` in `alembic/env.py`, delegates to `shared.db.alembic_env.run_alembic()`
- **Inter-service reads:** `fetch_data(service, query, params)` → GET `/internal/data/{query}` (HMAC-signed, 3s timeout)
- **Inter-service writes:** `call_action(service, action, payload)` → POST `/internal/actions/{action}` (HMAC-signed, 5s timeout)
- **Inter-service AP inbox:** `send_internal_activity()` → POST `/internal/inbox` (HMAC-signed, AP-shaped activities for cross-service writes)
- **Fragments:** HTML fragments fetched cross-service via `fetch_fragments()` for composing shared UI (nav, cart mini, auth menu)
- **Soft deletes:** Models use `deleted_at` column pattern
- **Context processors:** Each app provides its own `context_fn` that assembles template context from local DB + cross-service fragments
- L1 <-> L2: scoped JWT tokens (no shared secrets)
- L2: password + OAuth SSO, token revocation in Redis (30-day expiry)
- Federation: ActivityPub RSA signatures (`core/artdag/activitypub/`)
### Auth
- **Account** is the OAuth2 authorization server; all other apps are OAuth clients
- Per-app first-party session cookies (Safari ITP compatible), synchronized via device ID
- Grant verification: apps check grant validity against account DB (cached in Redis)
- Silent SSO: `prompt=none` OAuth flow for automatic cross-app login
- ActivityPub: RSA signatures, per-app virtual actor projections sharing same keypair
### Art DAG
- **3-Phase Execution:** Analyze → Plan → Execute (tasks in `artdag/l1/tasks/`)
- **Content-Addressed:** All data identified by SHA3-256 hashes or IPFS CIDs
- **S-Expression Effects:** Composable effect language in `artdag/l1/sexp_effects/`
- **Storage:** Local filesystem, S3, or IPFS backends
- L1 ↔ L2: scoped JWT tokens; L2: password + OAuth SSO
## Domains
| Service | Public URL | Dev Port |
|---------|-----------|----------|
| blog | blog.rose-ash.com | 8001 |
| market | market.rose-ash.com | 8002 |
| cart | cart.rose-ash.com | 8003 |
| events | events.rose-ash.com | 8004 |
| federation | federation.rose-ash.com | 8005 |
| account | account.rose-ash.com | 8006 |
| relations | (internal only) | 8008 |
| likes | (internal only) | 8009 |
| orders | orders.rose-ash.com | 8010 |
## Key Config Files
- `l1/pyproject.toml` - mypy, pytest, ruff config for L1
- `l1/celery_app.py` - Celery initialization
- `l1/database.py` / `l2/db.py` - SQLAlchemy models
- `l1/docker-compose.yml` / `l2/docker-compose.yml` - Swarm stacks
- `docker-compose.yml` / `docker-compose.dev.yml` — service definitions, env vars, volumes
- `deploy.sh` / `dev.sh` — deployment and development scripts
- `shared/infrastructure/factory.py` — app factory (all services use this)
- `{service}/alembic/env.py` — per-service migration config
- `_config/app-config.yaml` — runtime YAML config (mounted into containers)
## Tools