Tasks: a manager instance over many instances of the-loop

The last spec artifact. All eight tasks landed on PR #436 (commits aaf3ebb…d4a89a2). A DAG derived from the design and testing plan; each task names the testing-plan row that proves it. TDD: the test first, red, then green. The owner approved the spec set on PR #436 ("go ahead with implementation").

Task list

• [x] 1. The block — cli/the_loop/instance.py: role, Member, ManagerConfig, InstanceConfig.role/manager, InstanceConfigError under strict=True, the registry validation by index; the schema (authored + packaged); core.config.RESTART_REQUIRED gains instance.role; the template and this repository's cli-config.yaml
  • Depends on: none
  • Requirements: R1.1, R1.2, R1.4, R1.5, R1.6
  • Test: T1 — test_instance.py::test_the_role_*, ::test_the_registry_*, ::test_a_manager_refuses_to_boot_*; T10 — test_config_schema_parity.py, test_migrations.py, make validate
• [x] 2. The facade seam — api/facade.py (Facade, CORE), routes.build_router(…, facade=), mcp.build_server(…, facade=), the instance parameter on every keyed operation, errors.Conflict → 409, the The-Loop-Instances-Unreachable header; core.instance.assert_self; the OpenAPI contract
  • Depends on: 1
  • Requirements: R2.1, R2.6, R2.9
  • Test: T3 — test_api_contract_parity.py (both roles); T1 — test_instance.py::test_a_worker_refuses_a_foreign_instance; T8 — A6
• [x] 3. The registry surface — core/instances.py (list_instances, register_instance, unregister_instance), the three routes, the MCP read tool, loop.instances() on the SDK, status_all and the status lines, the the-loop instances command
  • Depends on: 2
  • Requirements: R3.1, R3.2, R3.3, R4.1–R4.6
  • Test: T1 — test_instances_core.py, test_instances_cmd.py, test_lifecycle_cmd.py::test_status_prints_the_fleet; T3 — the tool list; T10 — test_sdk_docs_parity.py
• [x] 4. The fleet — manager/fleet.py: Fleet, Probe, the local member, the transport, the probe cache and its transitions, the bounded fan-out, the resolvers, error translation
  • Depends on: 1
  • Requirements: R2.3, R2.4, R2.9, R3.4, R6.1–R6.4
  • Test: T1 — test_manager_fleet.py; T7 — -k "bounded or upstream_count"; T8 — A1, A2, A3, A4, A5, A9
• [x] 5. The facade — manager/facade.py: one module per core module, the § 4 table row by row; create_app picking the facade by role; serve.py refusing an unknown role or an unnamed manager
  • Depends on: 2, 4
  • Requirements: R1.3, R1.5, R2.2–R2.7, R2.9
  • Test: T1 — test_manager_facade.py; T2 — test_manager_integration.py; T8 — A7, A8
• [x] 6. The stream — manager/stream.py: FleetTail, the composite cursor, per-member resume, stamping, desync, backoff; StreamBroker(tail=)
  • Depends on: 4
  • Requirements: R2.8
  • Test: T1 — test_manager_stream.py; T2 — test_manager_stream_integration.py
• [x] 7. The dashboard — types, client (instance on keyed calls, the instances family), board keyed by instance@ref, routes (#/item/<ref>@<instance>, #/instances, #/instances/<name>), the sidebar chip and filter, healthTone, the Instances view, the instance pane, the demo fixture
  • Depends on: 2 (the contract)
  • Requirements: R5.1–R5.6
  • Test: T5 — cd ui && bun run test; T9; T12 — lint, typecheck, build
• [x] 8. Docs, capability docs, decision — docs/config/cli/instance-options.md, docs/cli/instances.md, docs/cli/state.md, docs/cli/commands/status.md, docs/cli/commands/instances.md + index + sidebar, docs/sdk/reference.md, docs/capabilities/instances.md, control-plane.md, cli.md, capabilities.md, ui/README.md, decision-138 accepted
  • Depends on: 3, 5, 6, 7
  • Requirements: R7.1–R7.3
  • Test: T10 — test_docs_parity.py, test_sdk_docs_parity.py; T12 — make check

Dependency graph (DAG)

Checkpoints

After 1 and 2: test_instance.py, test_api_contract_parity.py, the schema parity tests. After 3–6: the manager suites and the whole cli/tests. After 7: cd ui && bun run lint && bun run typecheck && bun run test && bun run build. After 8: make check and the docs parity tests; then the verification node executes testing-plan.md, and the review phases run the self/critic rounds and the security review gate (evidence/security-review.md, with the owner's named sign-off at tier 4).

Review comments

Appended by the-loop's record-feedback hook when a human gate approves with comments (issue-109). Append-only and attributed: an approval never silently discards a reviewer's suggestions, and the feedback travels with the document it concerns rather than living in a side-channel tracker.