Create DecisionService application-layer service that wraps
DecisionRepository with structured logging and UnitOfWork transaction
management. Wire DecisionService and PlanLifecycleService into the DI
container as Factory providers.
Inject DecisionService into PlanLifecycleService so that phase
transitions automatically record decisions: start_strategize records a
strategy_choice decision and start_execute records an
implementation_choice decision. Decision recording is optional and
never blocks lifecycle transitions.
Add Behave feature (5 scenarios), Robot Framework smoke tests (2 test
cases), ASV benchmarks (3 benchmark classes), and DI reference
documentation.
ISSUES CLOSED: #173
Tighten ToolRuntime._enforce_capabilities() to block ANY tool with
writes=True when plan_read_only is set, removing the not-cap.read_only
loophole that allowed certain write tools through. Tool name is now
always included in the ToolAccessDeniedError message.
Add read_only flag to ChangeSetCapture with ReadOnlyViolationError
raised when write-capable tools are wrapped on a read-only plan.
Add CLI fail-fast guards on plan execute and plan apply commands that
abort before calling the service layer if plan.read_only is True.
SkillContext.enforce_write_guard() already included tool name
correctly and required no changes.
Includes 18 Behave scenarios (90 steps), Robot integration tests,
ASV benchmarks, and docs/reference/read_only_actions.md.
ISSUES CLOSED: #322
Add AsyncResourceTracker (core/async_cleanup.py) providing a central
registry for async resources with timeout-bounded close_all(), async
context manager support, and a __del__ finalizer that logs leaked
resources by name.
Enhance LangGraphBridge with cleanup_tasks_async() that awaits
in-flight tasks with a deadline instead of fire-and-forget cancel().
Add cancellation_reasons dict to trace why tasks were cancelled.
Add StateManager.close() to properly release checkpoint file handles
and complete the RxPY BehaviorSubject. Add AcpEventQueue.close() to
dispose all subscriptions.
Includes 14 Behave scenarios (67 steps), Robot integration tests,
ASV benchmarks, and docs/reference/async_safety.md.
ISSUES CLOSED: #321
Add decision persistence layer with DecisionRepository, DecisionModel,
and Alembic migration. Includes tree queries (BFS traversal via deque,
path-to-root), superseded lookup, ordered decision path retrieval,
concrete Decision type annotations (via TYPE_CHECKING), and comprehensive
test coverage (Behave BDD, Robot Framework, ASV benchmarks). Updated
database_schema.md, CHANGELOG.md, and CONTRIBUTORS.md.
ISSUES CLOSED: #171
- decision-tree-view, decision-explain, and decision-tree-persistence
now invoke 'plan status --format plain' via mocked lifecycle service
so regressions in CLI rendering/serialization are caught
- plan-generates-decisions now asserts use_action was called by the CLI
and verifies plan status renders the strategize phase after creation
- Updated robot test case documentation to reflect CLI integration
Replace hard-coded '11 decision types' with dynamic len(DecisionType) in
the assertion and remove the literal from scenario names, step patterns,
docstrings, and docs so the tests stay correct when new DecisionType
values are added.
- Replace dead after_scenario() hook with context.add_cleanup() calls
so patchers are properly stopped (Behave only runs hooks from
environment.py)
- Fix invalid ULID constants: pad DECISION/CORRECTION to 26 chars,
replace invalid 'I' with 'J' in INVARIANT_ULID
- Add temp file cleanup for NamedTemporaryFile(delete=False) via
context.add_cleanup() in behave steps and try/finally in robot helper
- Remove decision-persistence doc sections from testing.md that belong
in PR #438, not this PR
- Patch CorrectionService at its definition site instead of the lazy
import location in plan.py (same fix as M3 helper)
- Replace hardcoded fake subplan/correction IDs with real ULIDs
generated via ulid.ULID() to satisfy Pydantic pattern validation
- Add missing --guidance flag to correction dry-run CLI invocation
- Fix subplan-status-sequential assertion to check for plan_id
(subplan_count is not part of the status JSON output)
The CorrectionService import in plan.py is a lazy import inside the
correct() function body, so it does not exist as a module-level attribute.
Patch the class at its definition site instead:
cleveragents.application.services.correction_service.CorrectionService
Fixes CI integration_tests failure for PR #439.
Robot Framework E2E test suite for M3 milestone verification covering:
- Plan execution generating decisions during Strategize phase
- Decision tree viewing with parent-child relationships and BFS traversal
- Decision explanation with full context snapshot verification
- Invariant add/list via CLI and InvariantService with scope filtering
- Dry-run correction via CorrectionService with impact analysis
- Live revert correction execution with decision re-creation
- Context snapshot round-trip serialisation assertions
- Decision tree persistence via model_dump/model_validate
- Correction revert re-execution from decision point
- Invariant enforcement during strategize with merge precedence
ISSUES CLOSED: #404
The 'Test Concurrent Retries With Jitter' test asserted that 5 concurrent
retry timestamps spread across >10ms. On busy CI runners thread scheduling
can compress wakeups into a narrower window, causing spurious failures.
Lowered the threshold from 10ms to 1ms — still validates that jitter produces
non-identical delays while tolerating CI scheduling variance.
Implemented plan-level and project-level locking with configurable timeouts.
Added locks table via Alembic migration storing owner_id, resource_type,
resource_id, acquired_at, and expires_at. Locks enforced in
PlanLifecycleService transitions. Support for re-entrant acquisition,
lock renewal, graceful shutdown release, and startup cleanup of expired
locks. Added diagnostics check for stale lock reporting.
ISSUES CLOSED: #327
- Extract MockMCPTransport to features/mocks/mock_mcp_transport.py
- Use ToolRegistry type with TYPE_CHECKING in register_tools()
- Move transport config validation into __init__ via _validate_config()
Implement MCPToolAdapter to connect to external MCP servers, enumerate
tools, and register them in ToolRegistry with source="mcp". Includes
connect/reconnect/disconnect lifecycle with timeout enforcement, input
validation on invoke, capability inference, Behave/Robot/ASV tests, and
docs/reference/mcp_adapter.md.
ISSUES CLOSED: #159
Extend actor YAML schema to support hierarchical graphs with explicit
node types (agent, tool, conditional, subgraph), per-node LSP bindings
(lsp_binding with server, languages, auto, capabilities), and tool-source
references (skills, mcp_servers, agent_skills).
Add schema validation for namespaced actor references, duplicate node IDs,
edge target existence, and graph reachability — all nodes must be reachable
from entry_node via explicit edges or conditional node routing targets.
Update loader to report YAML parse errors with precise line/column positions
and schema validation errors with dotted field paths and remediation hints
pointing to docs/reference/actor_config.md.
Add docs/reference/actor_config.md as the practical configuration reference
covering hierarchical graph examples, node type table, topology rules, and
common error cases with fix guidance.
Refresh examples/actors/graph_workflow.yaml to replace deprecated actor_path
with actor_ref. Add benchmarks/actor_yaml_bench.py for schema load overhead.
Tests: 95 Behave scenarios, 10 Robot smoke tests (including hierarchical
loader smoke test), security scan clean, coverage 99% (threshold 97%).
ISSUES CLOSED: #157
aditya
freemo
khyari hamza
brent.edwards
CoreRasurae
