Skip to content

DecisionGraph

Local-first decision evidence for agents and automation

DecisionGraph records immutable decision events, builds deterministic projections, verifies replay behavior, and ships a self-hosted BEAM runtime for investigation, approvals, and operational review.

Architecture

DecisionGraph has a deliberate split:

Layer Responsibility
Python reference core Frozen semantic authority for contracts, fixtures, validation, append semantics, and replay invariants
Contract boundary Parity tests, golden fixtures, and release gates that prove alignment
BEAM self-hosted runtime Event store, projector runtime, Phoenix API, LiveView console, workflows, and operational tooling

The canonical docs for that decision are:

At A Glance

Quickstart Get from clone to a useful local flow. Showcase See the intended operator journey. Self-Hosted Install Bring up the supported local-first topology. Release Review the current release and evidence trail.

What You Can Do

  • Append and query decision traces through the service layer.
  • Inspect traces, graphs, precedents, and projection health.
  • Run operator workflows for approvals, escalations, and incident review.
  • Self-host the runtime without SaaS assumptions.

Read By Goal

If you want to... Start here
understand the product quickly Showcase and Quickstart
run the supported deployment Self-Hosted Install
understand the runtime architecture Phoenix Platform Architecture and BEAM Supervision Tree
understand the semantic guarantees V1 Contracts, Reference Overview, and Semantic Parity Policy
operate workflows and review trails Workflow Runtime and Operator Console
evaluate release quality Release, Phase 10 Release Validation, and Self-Hosted Release Checklist

Documentation Map

Run

Build

Operate

Understand

Evaluate