apflow¶
Distributed Task Orchestration Framework
Start standalone in 30 seconds with DuckDB. Scale to distributed clusters with PostgreSQL -- no code changes required.
Deployment Modes¶
-
Standalone
Single process, DuckDB storage, zero configuration. Ideal for development, testing, and small-scale automation.
-
Distributed Cluster
PostgreSQL-backed, leader/worker topology with automatic leader election, task leasing, and horizontal scaling. Add worker nodes at any time.
Same application code in both modes
Write your tasks once. Switch from standalone to distributed by changing only the runtime configuration -- set APFLOW_CLUSTER_ENABLED=true and point to PostgreSQL.
Documentation¶
-
Getting Started
Installation, core concepts, and step-by-step tutorials to get up and running.
-
Guides
In-depth guides including distributed cluster deployment, task orchestration, custom executors, and best practices.
-
API Reference
Complete Python API, HTTP API, GraphQL API, and quick-reference cheat sheet.
-
Architecture
System design, task lifecycle, extension registry, and configuration internals.
-
Protocol Specification
The language-agnostic AI Partner Up Flow Protocol for cross-implementation interoperability.
-
Development
Environment setup, contribution guidelines, and project roadmap.
Key Capabilities¶
- 12+ Built-in Executors -- REST, SSH, Docker, gRPC, WebSocket, MCP, CrewAI, LLM, and more. Executor Selection Guide
- A2A Protocol -- Real-time streaming, Server-Sent Events, and WebSocket support for live task monitoring. HTTP API
- CLI Tools -- Create, run, and monitor tasks from the command line. CLI Reference
- Real-time Streaming -- Subscribe to task progress and intermediate results as they happen. API Server Guide
- Scheduler -- Cron-based and interval scheduling for recurring tasks. Scheduler Guide
- Extensible Architecture -- Add custom executors and extend the framework via the extension registry. Extensions Guide