Skip to content

How orch is built

Five diagrams of the Go binary, each drawn from the code as it is on main — the import graph, the transition table, the order of calls in the scheduler — not from the plan. Each opens as an interactive page you can zoom, search and trace. The diagrams are written in Spanish; their sources are in docs/architecture/.

Packages of the binary Architecture

What internal/cli wires together, what the engine may and may not import, and why internal/state is the one place runtime status lives.
Data through a project Data flow

From specs/*.md to the client page: tasks.json as static input, orch.db as the truth, agents and CI writing back, one snapshot for every reader.
Dispatching one task Sequence

What orch run does, in order: budget before capacity, worktree before prompt, the agent's report, the reap, and the pull request.
A task's lifecycle Lifecycle

The five statuses, the transitions orch allows, what triggers each one, and the moves it refuses.
Where the client page lives Architecture

orch publish to a folder, a gh-pages branch or your own orch-cloud Worker, and which secret protects each step.