modular-monolith-architecture-fastapi（modular monolith 架构 FastAPI）

FastAPI Modular Monolith Architecture — Project Scaffolder

Generate production-ready Modular Monolith projects with FastAPI, featuring proper module boundaries, async-first design, dependency injection, shared kernel, and infrastructure following industry best practices.

Overview

A Modular Monolith is a single deployable application organized into loosely coupled, highly cohesive modules — each representing a bounded context. FastAPI's speed, async capabilities, dependency injection system, and automatic API documentation make it an ideal framework for this architecture.

This skill scaffolds complete FastAPI projects with:

• Module isolation: Each module owns its models, repositories, services, routes, schemas, and dependencies
• Async-first design: All I/O operations use async/await (SQLAlchemy async, aiosmtplib, aiocache)
• FastAPI dependency injection: Loose coupling via Depends() and Annotated types
• Generic repository pattern: Base CRUD with pagination, filtering, sorting, and soft deletes
• Inter-module communication: Via gateway contracts and domain events (fastapi-events)
• Shared kernel (core): Cross-cutting concerns — base models, config, services, exception handling
• Database-per-module schema: Logical isolation within a single PostgreSQL database via Alembic
• Production-ready: Docker, Redis caching, task queues (Taskiq), structured logging, rate limiting
• Migration-ready boundaries: Modules can be extracted to microservices later

Workflow

Step 1: Gather Requirements

If the user hasn't specified, ask for:

1. Project name (e.g., conduit, saas-platform, ecommerce-api)
2. Business modules (e.g., Auth, Articles, Comments, Payments, Notifications)
3. Database preference (PostgreSQL recommended, MySQL supported)
4. Additional features: Event bus, caching (Redis), task queue, email, Docker, CI/CD
5. Authentication method: JWT (default), OAuth2, API keys

If the user provides $ARGUMENTS, parse them: $ARGUMENTS[0] = project name, remaining = module names.

Step 2: Generate Project Structure

Read the architecture references for FastAPI:

• For detailed project structure, see references/project-structure.md
• For module design patterns, see references/module-design.md
• For architecture decisions and comparisons, see references/architecture-guide.md
• For event-driven patterns, see references/event-driven-patterns.md
• For database patterns (SQLAlchemy async, Alembic), see references/database-patterns.md
• For testing strategies (pytest, TestClient), see references/testing-strategies.md
• For authentication and security patterns, see references/authentication-security.md
• For deployment and scaling, see references/deployment-scaling.md
• For a concrete Conduit (Medium clone) scaffold, see examples/conduit-scaffold.md
• For the generated README template, see examples/README-template.md

Generate the project following these critical rules:

Module Rules

1. Each module gets its own directory with internal layers: models/, schemas/, repositories/, services/, routes/, dependencies/
2. Modules communicate ONLY through gateway contracts — never import another module's internal types
3. Each module has its own Alembic migration files
4. Each module registers its own dependencies via Depends() functions
5. No circular dependencies between modules
6. Each module has its own routers.py that aggregates its versioned route files

Shared Kernel (Core) Rules

1. Contains ONLY cross-cutting concerns: base models, configuration, database session, generic repository, API response schemas, service interfaces (cache, mail, queue, log, events), exception handlers, middlewares
2. Must be thin — if it grows large, something belongs in a module
3. Never contains business logic specific to any module

Infrastructure Rules

1. Single entry point: app/main.py (FastAPI app initialization)
2. Composition root: app/core/routers.py wires all module routers together
3. Database session is shared but models/schemas are isolated per module
4. Domain events via fastapi-events for async inter-module communication (upgradeable to Kafka/RabbitMQ)
5. Background tasks via Taskiq with Redis backend

Step 3: Generate Code

For each module, generate:

• Models layer (models/): SQLAlchemy ORM models with soft delete support
• Schemas layer (schemas/): Pydantic v2 request/response/DTO schemas
• Repository layer (repositories/): Data access extending BaseRepository with custom queries
• Service layer (services/): Business logic with event dispatch and cross-cutting service calls
• Routes layer (routes/v1/): Versioned API endpoints with FastAPI routers
• Dependencies layer (dependencies/): DI setup for repositories, services, and auth guards
• Gateway (gateway.py): Public interface exposing module functionality to other modules
• Events (events.py): Domain event definitions
• Exceptions (exceptions.py): Module-specific custom exceptions
• Config (config.py): Module-specific configuration

Also generate:

• Shared kernel (app/core/): Base classes, database setup, generic repository, API schemas, service interfaces, exception handlers, middlewares, dependency injection
• Entry point (app/main.py): FastAPI app with lifespan, middleware registration, exception handlers
• Tests: Unit tests, integration tests, factories, and architecture boundary tests — see references/testing-strategies.md
• Docker (if requested): Dockerfile + docker-compose with PostgreSQL, Redis, Taskiq worker, and Mailhog — see references/deployment-scaling.md
• Alembic: Migration configuration with alembic.ini and migrations/env.py
• pyproject.toml: Dependencies managed with UV or pip
• README.md: Architecture overview, how to run, how to add modules (use examples/README-template.md)

Step 4: Validate

After generation:

1. Verify no module directly imports another module's internal types (only gateways)
2. Confirm each module has its own models and migration support
3. Check that the shared kernel (app/core/) contains no business logic
4. Ensure the project runs successfully with uvicorn app.main:app --reload
5. Run any generated tests with pytest

Validation scripts are available in scripts/ for CI integration:

• scripts/validate-boundaries.sh <app-dir> — detects cross-module boundary violations in Python imports
• scripts/validate-shared-kernel.sh <core-dir> <modules-dir> — ensures core doesn't reference modules
• scripts/check-circular-deps.sh <app-dir> — detects circular dependencies between modules

Step 5: Migration Guidance

If the user asks about extracting modules to microservices:

• Replace gateway contracts (in-process function calls) with HTTP/gRPC clients
• Swap fastapi-events for Kafka/RabbitMQ for that module's events
• Migrate module's database tables to a separate database
• Deploy the extracted module as a standalone FastAPI service
• Keep remaining modules as a monolith (no need to extract everything)

Key Principles to Enforce

Example Invocations

/modular-monolith-architecture-FastAPI conduit Auth Articles Comments
/modular-monolith-architecture-FastAPI ecommerce-api Auth Products Orders Payments
/modular-monolith-architecture-FastAPI saas-platform Auth Tenants Billing Notifications
/modular-monolith-architecture-FastAPI social-api Auth Users Posts Messages

When NOT to Use This

Suggest microservices instead if the user describes:

• Teams needing completely independent deployment cadences
• Requirements for polyglot tech stacks (Python ML + Go APIs + Java enterprise)
• Extreme per-service scaling requirements
• Already having Kubernetes infrastructure and DevOps maturity

Suggest a simple FastAPI monolith if:

• Solo developer or very small team (1-3 devs)
• Prototype/MVP with unclear domain boundaries
• Application with fewer than 3 distinct business domains
• Simple CRUD API without complex business logic

Technology Stack