AOA Toolkit

Architecture · LikeC4

AOA Toolkit

Runs an agent against real tasks mined from a repository, then scores retrieval locality, edit locality, invariant discoverability, and mutation surface

Open the interactive explorer → View on GitHub
built evolving planned research risk

The whole picture

System landscape

The system in context, then opened up into its containers.

AOA Toolkit — system landscape

AOA Toolkit — system landscape

index
AOA Toolkit — containers (read-only vs write-gated)

AOA Toolkit — containers (read-only vs write-gated)

aoaSystem

Runs an agent against real tasks mined from a repository, then scores retrieval locality, edit locality, invariant discoverability, and mutation surface

Risks & open questions

Risks & open questions

risks

Inside each box

Containers & components

Each part decomposed into the components that implement it. Every box links to its source in the interactive explorer.

Trace + symbol-graph substrate — components

Trace + symbol-graph substrate — components

substrateContainer

Normalizes every run, live or recorded, into one 8-span trace schema, and indexes the repository into a confidence-tiered symbol graph

Read-only measurement — components

Read-only measurement — components

measureContainer

observe / audit / lint-context / eval — zero-write against the repository under test

Trust gates (R0 / R5 / R7) — components

Trust gates (R0 / R5 / R7) — components

gatesContainer

The gates that decide whether migrate is worth trusting on a given repository, and whether a mutation may proceed at all

How it runs

Walkthrough flows

Dynamic views — the narrative spine of the system, step by step.

Walkthrough — observe + audit (zero-write)

Walkthrough — observe + audit (zero-write)

auditFlow

aoa observe installs telemetry; nothing beyond an ignored .aoa/ dir is visible to the agent. aoa audit is the read-only report.

Walkthrough — eval run against a codeprobe-mined task

Walkthrough — eval run against a codeprobe-mined task

evalFlow

aoa eval run --codeprobe-run <path> --tasks <path> turns a recorded transcript into per-task metrics.

Walkthrough — falsify: the deciding experiment

Walkthrough — falsify: the deciding experiment

falsifyFlow

Before migrate is trusted on a given repo: does migrating it beat simply swapping the harness, across at least five held-out repos? A tie defaults to pivot.

Walkthrough — a gated migration

Walkthrough — a gated migration

migrateFlow

migrate stays dry-run by default; --apply only runs once the operator, informed by falsify, decides to trust it on this repo.

Where it runs

Deployment

What runs where, and the process & data boundaries between the pieces.

Deployment — where each piece runs

Deployment — where each piece runs

deployment

Explore it live

The figures above are static exports. The interactive explorer lets you pan, zoom, follow relationships, and jump from any box to the source.

Open the interactive explorer →