dagron/ docs
Documentation

Operate dagron

dagron is an open-source DAG workflow engine: plain-YAML workflows, retries and backoff, a live web UI, SQLite or Postgres, and Kubernetes execution — from one static Rust binary, with a database as its only state. These pages cover running it: the CLI, day-to-day administration, upgrades, and scaling.

Install

Three ways in, smallest first. The workflow YAML is the same in all of them.

One binary — local development
# engine + management API + Swagger on localhost
$ dagron dev workflow.yaml

# one-shot: run a workflow, drain, exit
$ dagron examples/simple_dag.yaml
Docker Compose — full stack with the web UI
$ git clone https://github.com/lucheeseng827/dagron
$ cd dagron
$ docker compose -f compose.quickstart.yaml up -d

# console at /, API at /api — one port
http://localhost:8080
Helm — production on Kubernetes
# multi-arch images + OCI chart on Docker Hub
$ helm install dagron \
    oci://registry-1.docker.io/mancube/dagron

# point at your own Postgres
$ helm install dagron oci://registry-1.docker.io/mancube/dagron \
    --set postgres.enabled=false \
    --set externalDatabaseSecret.name=dagron-db

Prebuilt binaries for Linux, macOS and Windows are on GitHub Releases; the chart is also listed on Artifact Hub.

The pages

Start

Get started

From an empty directory to a workflow you can watch run — and a tour of every screen in the console, region by region, so you know what you are looking at before it matters.

 Reference

HTTP API

Every endpoint of both surfaces — the authenticated gateway and the engine's cluster-private ops API — with the permission model, the server-side log filter, the SSE contract, and the SDK and MCP clients that speak them.

 Reference

CLI & configuration

The dagron binaries and how they are invoked, the environment variables that configure them, compile-time features, and the MCP tools an agent can call.

 Operate

Admin & maintenance

What you are operating and what its state is. Monitoring and health, backups and restore, dead letters, API tokens, secrets, and a symptom-first troubleshooting table.

 Operate

Update & upgrade

How schema migrations actually work, the production upgrade procedure, rolling restarts across multiple schedulers, and rollback.

 Operate

Scaling & HA

How the data plane scales out, the deployment configurations from a single binary to a replicated cluster, the knobs that govern concurrency, and the limits.

Going deeper

These pages are the operator's cut, and HTTP API covers both API surfaces in full. The rest of the reference — workflow YAML fields, GitOps, datasets, streaming, AI workloads, the constrained-host profile and the Raspberry Pi measurements — lives with the source: