Parent: #163
Goal
Assemble a boot-loadable BootContext from durable stores: frontier cursor, memory IDs, in-flight task saga IDs, and latest event sequence.
This is the reset recovery contract. A maestro starting after context loss should not guess what matters; it should load a compact strategic context from explicit durable projections.
Scope
Expected owned surface:
src/forge_loop/control/boot.pysrc/forge_loop/frontier/src/forge_loop/eventlog/src/forge_loop/memory/src/forge_loop/tasks/- focused tests under
tests/test_boot_context.py or equivalent.
Prefer small adapter protocols/fakes over a broad runner integration. The goal is boot assembly, not dispatch.
Required Behavior
BootContext includes:
- frontier cursor summary;
- active memory IDs;
- rejected-path memory IDs if the memory store exposes them;
- in-flight task IDs/saga IDs;
- latest event sequence;
- projection cursor positions or lag when available.
- Boot assembly can run after reopening all durable stores.
- Missing optional stores fail diagnosably or return empty sections; missing required frontier state should be explicit.
BootContext.summary() remains compact enough to be placed near the edge of an LLM context.
Acceptance Tests
- Seed a frontier cursor, event log, memory records, and task saga records.
- Reopen the stores and assemble a
BootContext.
- Assert the summary contains the current problem, next expansion, active memory IDs, in-flight task IDs, and latest event sequence.
- Add a missing-frontier test that errors clearly.
- Add a no-active-tasks test proving empty task state is not confused with failure.
Non-goals
- Do not start workers from boot assembly.
- Do not call model providers.
- Do not add a dashboard.
- Do not make the runner depend on this until the contract is tested.
Verification
Run at minimum:
env -u VIRTUAL_ENV uv run --extra dev pytest tests/test_frontier_store.py tests/test_eventlog_sqlite.py -q- New boot-context tests.
env -u VIRTUAL_ENV uv run --extra dev ruff check <changed files>env -u VIRTUAL_ENV uv run --extra dev ruff format --check <changed files>
Customer Story
A forge-loop operator running autonomous work across multiple sessions benefits because maestro startup can reconstruct what matters without relying on model context.
Source
Expanded during Forge self-dogfood sprint planning on 2026-06-02.
Parent: #163
Goal
Assemble a boot-loadable
BootContextfrom durable stores: frontier cursor, memory IDs, in-flight task saga IDs, and latest event sequence.This is the reset recovery contract. A maestro starting after context loss should not guess what matters; it should load a compact strategic context from explicit durable projections.
Scope
Expected owned surface:
src/forge_loop/control/boot.pysrc/forge_loop/frontier/src/forge_loop/eventlog/src/forge_loop/memory/src/forge_loop/tasks/tests/test_boot_context.pyor equivalent.Prefer small adapter protocols/fakes over a broad runner integration. The goal is boot assembly, not dispatch.
Required Behavior
BootContextincludes:BootContext.summary()remains compact enough to be placed near the edge of an LLM context.Acceptance Tests
BootContext.Non-goals
Verification
Run at minimum:
env -u VIRTUAL_ENV uv run --extra dev pytest tests/test_frontier_store.py tests/test_eventlog_sqlite.py -qenv -u VIRTUAL_ENV uv run --extra dev ruff check <changed files>env -u VIRTUAL_ENV uv run --extra dev ruff format --check <changed files>Customer Story
A forge-loop operator running autonomous work across multiple sessions benefits because maestro startup can reconstruct what matters without relying on model context.
Source
Expanded during Forge self-dogfood sprint planning on 2026-06-02.