Integrating ArcFlow

What this track is

The integrating track is for teams that have run a local stub workflow and now need to choose how ArcFlow fits their stack. The lessons compare embedded SDK runs with server-backed HTTP execution, walk through a first server API call, and introduce the two pause patterns that require Postgres: external callbacks and human-in-the-loop (HITL) approval.

Nothing here replaces the full guides under documentation/guides/. These pages are short introductions with one runnable example each and pointers to normative detail.

How to use it

Read the lessons in order if you are deciding deployment shape for the first time. Skip lesson 1 if you already know you need the HTTP server. Each page includes a Verify section so you can confirm behavior before moving on.

OrderLessonWhat you learn
1Embedded SDK vs serverWhen to embed the runtime vs call arcflow-server
2Server API first runDocker stack, POST /v1/runs, poll, trace
3External callbacks introExternalOutcome.report / report_outcome lifecycle
4HITL approval introInterrupt, approve, resume on the server

Before you start

Complete at least one path from Getting started paths:

PathDocument
Fastest local runFirst workflow in five minutes
Python SDK detailPython quickstart
HTTP without SDKServer API quickstart

For server lessons (2 through 4), you need Docker with Compose v2 and ports 8080 and 5432 available on your machine.

After this track

GoalDocument
Full external callback guideExternal callbacks
Webhook HMAC verificationWebhook security
HITL configuration detailConfiguring interrupts
HITL approve APIApprove and reject
Guided server tutorialTrack B: Server API
HITL and external tutorialTrack E: HITL and external
Static site productStatic site chatbot