How shrec/aiworkhub works
AIWorkHub is a local-first control plane for multi-model coding agents, surfaced through VS Code and MCP. Its principal loop is to capture repository work, plan dependency-aware tasks, route and launch isolated workers, validate and review their changes, promote accepted work, and retain outcomes and project context for future decisions. The graph preserves documented intake, source intelligence, durable memory, review, and learning capabilities. Relationship evidence is limited to the README and SOURCE INDEX references; unsampled implementation wiring is not asserted.
Main entry points: manager_loop.py Manager loop · process_launcher.py Process launcher · workforce_router.py Workforce routing · task_store.py Task DAG
Components 22
User surfaces
- VS Code extension
vscode-extension/extension.js - MCP interface
src/aiworkhub/server.py
Planning and context
- Manager loop
src/aiworkhub/manager_loop.py - NeedFix and roadmap
src/aiworkhub/needfix_store.py - Task decomposition
src/aiworkhub/task_decomposition.py - Task DAG
src/aiworkhub/task_store.py - Project context tools
src/aiworkhub/manager_ai_tools.py - Source Graph
src/aiworkhub/source_graph.py
Worker execution
- Repository policy
src/aiworkhub/repo_policy.py - Workforce catalog
src/aiworkhub/workforce_catalog.py - Workforce routing
src/aiworkhub/workforce_router.py - Process launcher
src/aiworkhub/process_launcher.py - Runtime adapters
src/aiworkhub/runtime_adapters.py - Isolated worktrees
src/aiworkhub/worker_workspace.py
Verification and promotion
- Validation runner
src/aiworkhub/validation_runner.py - Evidence-based review
src/aiworkhub/review_orchestrator.py - Manager promotion
src/aiworkhub/task_engine.py
Durable authorities
- Context graph
src/aiworkhub/context_graph.py - Learning commits
src/aiworkhub/learning_commit.py - Storage registry
src/aiworkhub/storage_registry.py
Other components
- Developer
- Coding model providers
Connections 24
- Developer → VS Code extension: uses
README.md - Developer → MCP interface: uses
README.md - VS Code extension → Manager loop: manages work
- MCP interface → Manager loop: exposes control
- Manager loop → NeedFix and roadmap: observes intake
README.md - Manager loop → Task decomposition: plans tasks
README.md - Manager loop → Task DAG: coordinates DAG
README.md - Manager loop → Workforce routing: routes work
README.md - Manager loop → Project context tools: uses context
src/aiworkhub/manager_loop.py - Project context tools → Context graph: queries memory
src/aiworkhub/manager_ai_tools.py - Manager loop → Source Graph: uses intelligence
README.md - Workforce routing → Workforce catalog: ranks workers
src/aiworkhub/workforce_catalog.py - Workforce routing → Process launcher: dispatches tasks
- Process launcher → Repository policy: checks launch policy
src/aiworkhub/repo_policy.py - Process launcher → Runtime adapters: selects adapter
- Runtime adapters → Coding model providers: invokes models
README.md - Process launcher → Isolated worktrees: launches isolated
README.md - Manager loop → Validation runner: verifies changes
README.md - Manager loop → Evidence-based review: requests review
README.md - Manager loop → Manager promotion: controls promotion
README.md - Manager promotion → Task DAG: records decisions
src/aiworkhub/task_engine.py - Manager loop → Learning commits: learns outcomes
README.md - Context graph → Storage registry: uses canonical storage
src/aiworkhub/context_graph.py - Task DAG → Storage registry: registers authority
src/aiworkhub/task_store.py