Loading…
Loading…
Describe your product requirements and Archiet generates a complete FastAPI + Next.js application: Pydantic v2 request/response models, SQLAlchemy 2 ORM with async support, Alembic migrations, JWT auth, workspace isolation, Celery workers, and auto-generated OpenAPI 3.1 documentation.
FastAPI is the fastest Python web framework by benchmark (comparable to Node.js and Go on async workloads) and the most developer-friendly: type hints drive automatic validation and OpenAPI documentation, async/await is native, and the ecosystem (Pydantic, SQLAlchemy, Celery, pytest) is mature. For AI-adjacent applications, FastAPI is the dominant choice — it pairs naturally with LangChain, Anthropic SDK, OpenAI SDK, and Hugging Face pipelines because it handles async LLM calls natively.
The generated application ships: FastAPI 0.115+ with lifespan context management, Pydantic v2 BaseModel for all request/response schemas with strict mode enabled, SQLAlchemy 2.0 ORM with async engine (asyncpg), Alembic migrations with auto-detection, JWT authentication via python-jose with httpOnly cookie transport, APIRouter-based modular structure (one router per entity), dependency injection for db sessions and current user, workspace-scoped multi-tenancy enforced via dependency, Celery 5 with Redis for async tasks, pytest-asyncio test suite with async fixtures, and auto-generated OpenAPI 3.1 JSON served at /openapi.json.
FastAPI is the natural home for AI-powered applications. If your requirements include ML workloads — RAG pipelines, LLM gateways, model inference endpoints, or embedding services — Archiet detects these signals from your PRD and generates the ML infrastructure alongside the standard application: a RAG service with pgvector integration, an LLM gateway that routes to OpenRouter/Anthropic/OpenAI via environment variable, and a Celery worker for async inference tasks that don't block the API. The ML components are governed by the same auth and tenancy controls as the rest of the application.
If your requirements mention approval workflows, claims routing, or decision-based automation, Archiet detects the governed-agent signal and generates a BPMN-backed approval workflow alongside your FastAPI application. The generated agent includes: a FastAPI router at /api/agents/{workflow_name}/decide, a Python runtime that evaluates a DMN policy table (fully deterministic, no LLM in the routing path), Pydantic request/response models for the decision inputs and outputs, and an audit log model that records every decision with the rule that matched. The same governed-agent architecture is used for invoice approval, claims triage, loan adjudication, and prior authorization workflows.
The generated test suite uses pytest-asyncio with an async test client, factory functions for entities, and seeded test data. The Docker Compose stack includes PostgreSQL, Redis, the FastAPI application, and Celery worker. The application is pre-configured for deployment on any container platform — Dockerfile included, multi-stage build for production image size, non-root user, and health check endpoint at /api/health.
Archiet targets Python 3.12+ for all generated Python backends. The Dockerfile uses python:3.12-slim as the base image. This is important: Python 3.12 introduced f-string improvements used in the generated code that cause SyntaxError on 3.11. The CI guard (python -m compileall) validates this at image build time.
Yes. The genome (your architecture model) is stack-agnostic. Generate FastAPI today, switch to Flask tomorrow and regenerate — the entities, relationships, and business rules are preserved. This is useful for teams evaluating both options or migrating between them.
All database operations use SQLAlchemy's AsyncSession with asyncpg. Route handlers are async def. Celery tasks run in separate workers (not in the async event loop) — this is the correct pattern for CPU-bound or IO-blocking work. The generated code does not mix sync and async incorrectly, which is the most common source of FastAPI production bugs.
Pydantic v2 with the new @model_validator and @field_validator APIs. The generated schemas use model_config = ConfigDict(from_attributes=True) for ORM mode (v2 equivalent of orm_mode=True in v1). All validators use the v2 signature — no deprecated v1 compatibility shims.
Describe your product once. Archiet generates a complete Flask + Next.js application — auth with httpOnly cookies, RBAC, multi-tenant workspace isolation, Alembic migrations, Celery workers, and a full OpenAPI 3.1 spec — without writing a single line of boilerplate.
Upload your product requirements document and Archiet generates a NestJS application with JWT authentication guards, TypeORM entities and migrations, workspace-scoped multi-tenancy, Swagger/OpenAPI 3.1 documentation, and a typed API client for your Next.js frontend — zero boilerplate written by hand.
Import your ArchiMate 3.2 model from Enterprise Architect, Archi, or draw.io. Archiet generates a production-ready application in your chosen stack — with auth, RBAC, multi-tenancy, migrations, and compliance documentation — directly from the formal architecture. No gap between design and code.
free plan. No credit card required. Generate your first compliant architecture blueprint in under 10 minutes.