Module engine

Expand description

Engine — the core orchestration loop.

The engine is the central coordinator. It owns the tick loop, backend connections, channel registry, transport layer, and tmux session manager.

Submodules handle the heavy lifting:

This file contains struct definitions, project initialization, and the main event loop (serve()).

Modules§

auto_merge: PR auto-merge and review-changes handling.
channel_handler: Channel message handlers — routes incoming channel messages to the appropriate action.
cleanup: Worktree cleanup and merged-PR detection.
commands: Owner slash commands — parse, validate, and execute /commands in issue comments.
cooldown: Per-agent and per-model cooldown tracking.
dependencies: Sync GitHub blocked-by links into SQLite and hold tasks until those links are done.
dispatch_guard: RAII guard for the per-task dispatching DashMap.
events: Task event bus — broadcast channel for status transitions.
jobs: Job scheduler — replaces jobs_tick.sh.
review: PR review pipeline — review agent orchestration.
review_poll: PR review polling — review_open_prs.
router: Agent router — selects the best agent and model for each task.
runner: Task runner — executes tasks using AI agents.
subscribers
suspend: Host suspend/resume detection.
sync: Sync tick — periodic operations that run less frequently than the core tick.
tasks
tick: Core tick loop phases.

RATE_LIMITED_UNTIL_KEY: KV key holding the epoch second the GitHub core window resets, so orch task status can show the limit from a separate process.

attribution_footer: Build a standard attribution footer for GitHub comments posted by orch bots.
configured_agents: Return the configured agent list from top-level agents in config.yml, falling back to router::config::DEFAULT_AGENTS when not set.
format_rate_limited_until: Format an epoch second as a local HH:MM:SS for operator-facing output.
orch_footer: Standard Orch attribution footer for issue bodies, PR bodies, and comments.
serve: Start the orch service.