How kortix-ai/suna works
Project Type and Purpose
• This is a full-stack web application called Suna, an open-source AI assistant that automates real-world tasks through conversation.
• It has four main parts:
• A Frontend (Next.js/React) for the user interface
• A Backend API (Python/FastAPI) that manages threads, users, and talks to LLMs
• An Agent Docker environment (Daytona) where each AI agent runs securely
• A Supabase Database for persistence, authentication, and real-time subscriptions
File Structure
• Top-level directories:
• .github / docker-compose.yaml – CI/CD and container orchestration
• backend/ – FastAPI service, agent code, sandbox runner, service wrappers, scripts
• frontend/ – Next.js app with UI components, pages, APIs, and styling/config files
• Key backend sub-directories:
• backend/agent/ – core agent logic, tool integrations (browser, file system, data providers)
• backend/agentpress/ – thread and tool registry, context and response processing
• backend/services/ – wrappers for Redis, LLMs, Supabase, billing
• backend/sandbox/ – isolated sandbox for secure tool execution
• Deployment files:
• Dockerfiles (root, frontend/Dockerfile, backend/Dockerfile, sandbox/docker)
• Fly.io and docker-compose configurations
README for Additional Insights
• Confirms the four components and their roles.
• Mentions LLM providers (Anthropic, OpenAI, Groq), web search/scraping tools (Tavily, Firecrawl), Playwright for browser automation.