AsyncGraph

Published docs: https://artyomb.github.io/async-graph/

AsyncGraph is a Ruby runtime for graph-style workflows that suspend on external work, store jobs outside the graph, and resume on later passes. It supports:

• single-step graph execution
• runner helpers for opaque persisted run state and fan-out/join bookkeeping
• barrier joins such as edge %i[left right], :merge
• library-owned join processing for persisted branch tokens
• await.call(...) for one external job
• await.all(...) for multiple parallel jobs in one node
• graph validation before execution

gem install async-graph

require 'async-graph'

graph = AsyncGraph::Graph.new do
  node :fetch_user do |state, await|
    user = await.call("user", :fetch_user, user_id: state[:user_id])
    { user: user }
  end

  set_entry_point :fetch_user
  set_finish_point :fetch_user
end

step = graph.step(state: { user_id: 7 }, node: graph.entry)
request = step.requests.first

resumed = graph.step(
  state: step.state,
  node: step.node,
  resolved: { request.key => { id: 7, name: "Ada" } }
)

resumed.state
# => { user_id: 7, user: { id: 7, name: "Ada" } }

For persisted multi-pass execution, AsyncGraph::Runner can create and advance a run snapshot while your application still owns persistence and external jobs.

The repository includes three runnable examples in examples/:

• persisted multi-pass flow with external job persistence:

bash examples/run.sh

• self-contained runner loop with inline :add / :subtract request handling through resolve_request::

ruby examples/all_in_one_inline_runner.rb

• self-contained runner loop that still suspends, binds requests to in-memory job IDs, and resumes in the same process:

ruby examples/all_in_one_jobs_runner.rb

The repository also includes the Starlight source site in docs-site/.

cd docs-site
npm install
npm run dev