Long-running workflows need retries, queues, error handling, and monitoring — that's months of plumbing before you ship a single feature. Durable execution shouldn't require you to manage workers or learn a custom runtime.
GraphIngest lets you define workflows in plain code, deploy with one call, and run them on managed infrastructure — in Python, TypeScript, Go, JavaScript, or Java. Need AI agents? Built-in ReAct and tool calling are included.
from graphingest import graph, node
@node(retries=3, timeout=30)
def extract(url: str):
return fetch_data(url)
@node
def transform(raw_data):
return clean(raw_data)
@graph(name="etl-pipeline")
def pipeline(url):
data = extract(url)
return transform(data)Features
Production-grade primitives for building reliable data pipelines and AI agent workflows.
Define pipelines as directed graphs with @node and @graph decorators. Automatic dependency resolution and parallel execution.
Configurable retry policies with exponential backoff. Never lose a task to transient failures.
Concurrency limits, throttling, and priority queuing built into the SDK. No external config needed.
Per-node and per-graph timeouts with graceful cancellation. Keep pipelines on schedule.
Monitor runs, inspect task states, view logs, and track metrics from a single control plane.
Built-in ReAct agents and LangGraph support. Run multi-step AI workflows with tool calling, memory, and automatic tool routing.
How It Works
Use decorators to define nodes, edges, and retry policies. The SDK handles the rest.
Call deploy() and the platform handles everything. Auto-scales from zero to handle any load.
Watch runs in real-time on the dashboard. Debug failures with structured logs and traces.
SDKs
First-class SDKs with the same decorator-based API across all languages.
@node / @agent
node() / agent()
gi.Node() / gi.Agent()
node() / agent()
Node.create() / React.agent()
Why GraphIngest
Traditional orchestrators force you to manage workers and infrastructure. Serverless orchestrators lock you into one language and their cloud. GraphIngest gives you both simplicity and control.
One function call. No Dockerfiles, no CLI tools, no YAML.
Python, TypeScript, Go, JavaScript, Java — same API everywhere.
ReAct loops, tool calling, and LLM routing in the SDK — not bolted on.
Pay nothing at idle. Auto-scale to thousands of parallel tasks.
| Capability | Worker-based Orchestrators | Serverless Orchestrators | GraphIngest |
|---|---|---|---|
| Deployment | Workers + config | Add route to app | deploy() — zero config |
| Language SDKs | 1 (Python) | 1 (TypeScript) | 5 languages |
| Infra management | You manage workers | You manage app | Platform manages all |
| Scale to zero | yes | ||
| Parallel fan-out | Via external executor | No native support | .map() with auto-scale |
| AI agent support | Built-in ReAct + LLM | ||
| Dashboard latency | Polling (seconds) | Polling (seconds) | Real-time (instant) |
| Retry granularity | Re-run entire flow | Step-level | Resume from failure |
| Env variable mgmt | External blocks | You manage | Dashboard UI + env_path |
| Idle cost | Workers run 24/7 | Your infra cost | $0 when idle |
Pricing
Get started with generous free limits.
Managed infrastructure. Zero ops.
Join teams using GraphIngest to ship reliable data pipelines and AI agent workflows.