SaaS Alternative
AWS Step FunctionsCategories
Self-Hosted Developer Tools Automation Orchestration Distributed Systems Task Queue Workflow Engine Background JobsStars
Forks
Watchers
Developer links
Hatchet
Processing over one billion tasks per month on Hatchet Cloud and holding 7,600+ GitHub stars, Hatchet is the open-source orchestration engine that replaces fragile legacy queues with durable, fault-tolerant task execution built entirely on PostgreSQL — no Redis, RabbitMQ, or Kafka required. Born from the creators' experience scaling Uber's Cadence-inspired patterns, the v1 engine achieves 10,000 tasks per second sustained throughput with 20ms average queue latency through dynamic buffer flushing and batch insert optimization. Define tasks, durable workflows, and DAG pipelines as code using native SDKs for Python, TypeScript, Go, and Ruby — every function gets automatic retries with configurable backoff, concurrency control with group round robin or cancellation policies, priority queuing, and dynamic rate limiting for third-party API protection. Durable execution persists the complete history of every task and state transition, enabling replay from failure, debugging via full event timelines, and complex pause/resume conditions using durable sleep and event waits. The real-time web dashboard provides workflow run visualization with DAG timeline rendering, worker health monitoring with slot utilization, queue depth metrics, task throughput charts, and error rate tracking — all filterable by status, workflow, or time window. OpenTelemetry integration and Prometheus metrics export enable advanced observability. Multi-tenant by default with users, roles, and namespace isolation. Self-host via Docker Compose with PostgreSQL and optional RabbitMQ, or use the single-container Hatchet Lite image for development. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Benefits
- Postgres Only Durable Task Queue
- Built entirely on PostgreSQL as the durability and observability layer. No Redis, RabbitMQ, or Kafka dependencies — a single database handles task state, execution history, and queue management.
- Multi Language Native SDK Support
- Write tasks, workflows, and AI agents in Python, TypeScript, Go, or Ruby using idiomatic native SDKs. Define everything as code with full IDE support and type safety.
- Ten Thousand Tasks Per Second
- The v1 engine achieves 10,000 tasks per second sustained throughput with 20ms average queue latency through dynamic buffer flushing and PostgreSQL batch insert optimization.
- Full Execution History and Replay
- Every task and state transition is durably persisted. Debug failures with complete event timelines, replay from any checkpoint, and inspect workflow runs via the real-time web dashboard.
Features
- Durable Workflow Execution
- Fault-tolerant long-running workflows with automatic state persistence, configurable retry policies, durable sleep, event waits, and checkpoint-based recovery from failures.
- DAG Pipeline Orchestration
- Build directed acyclic graph pipelines with task dependencies, parallel execution branches, conditional logic, and visual DAG timeline rendering in the web dashboard.
- Concurrency and Rate Limiting
- Group round robin, cancel-in-progress, and cancel-newest concurrency strategies. Static and dynamic rate limits protect third-party APIs with per-user or per-resource granularity.
- Real Time Web Dashboard
- Monitor task runs, workflow executions, worker health, queue depth, and error rates in real time. Filter, inspect, cancel, and replay runs directly from the browser.
- OpenTelemetry and Prometheus
- Built-in OpenTelemetry collector for distributed tracing across workflow steps. Prometheus metrics endpoint exposes queue size, throughput, and rate limiting counters.
- Docker Compose Deployment
- Self-host via Docker Compose with PostgreSQL and optional RabbitMQ, or use the single-container Hatchet Lite image for development and low-throughput production workloads.