Skip to content

Introducing Shakti Python Framework

Most Python web projects today start the same way: pick a routing framework, then spend the next few days wiring up an ORM, an auth system, an admin panel, and — increasingly — an AI provider. Each piece comes from a different library, with its own conventions, its own version-compatibility surface, and its own documentation to learn.

Shakti Python Framework exists to close that gap. It's an open-source, AI-first Python web framework built on async ASGI from day one, with the pieces most APIs eventually need shipped in a single install instead of assembled by hand.

What "AI-first" actually means here

A lot of frameworks bolt AI support on as an afterthought — a community package that wraps an HTTP client around a chat API. Shakti treats AI as a first-class subsystem, the same way it treats routing or the ORM:

from shakti import AI

ai = AI(config)
ai.init_app(app)

That gives you a working POST /ai/chat endpoint, plus a Python API for chat, streaming, retrieval-augmented generation (RAG), and tool-calling agents — backed by Anthropic or OpenAI, switchable via config. See AI Overview.

What's in the box

  • Routing — typed path converters, middleware groups, WebSocket routes, static file serving with production-sane cache headers.
  • Async ORM — SQLAlchemy 2.x under the hood, a repository pattern for common CRUD, and one-command migration generation via Alembic.
  • Auth — JWT access/refresh tokens, role-based access control, and API keys, all backed by the same User model.
  • Admin panel — register a model, get a working dark/light-mode CRUD UI with search, CSV export, and an activity log. No template to write.
  • Background jobs — an async job queue with retries, exponential backoff, and interval scheduling.
  • Monitoring — a live dashboard, health checks, and request metrics with zero configuration.

Every piece is designed to compose with the others — the admin panel reads the same User model your auth system issues tokens for; the AI module's routes go through the same request/response pipeline as everything else.

Who it's for

If you're building an API, a full web app, or an AI agent in Python and want a shorter path from pip install to a running, database-backed, authenticated service, Shakti is built for that. If you want to see how it stacks up against what you're already using, read Shakti vs FastAPI or Shakti vs Django.

Try it

pip install "shakti-framework[all]"
shakti new myapp
cd myapp
shakti run --reload

Start with Your First App for a full walkthrough, or the Quick Start for the condensed version.