Reformat the lint session's ruff check call in noxfile.py to a multi-line
form so ruff format --check passes (the single-line form exceeded 88 chars).
Add CHANGELOG.md entry under [Unreleased] ### Changed for issue #10848.
ISSUES CLOSED: #10848
- features/steps/tdd_plan_generation_validate_steps.py: drop duplicate
@given/@then registrations (live in plan_generation_langgraph_coverage_
steps.py; redefinition caused AmbiguousStep errors crashing all 8
behave-parallel workers); have the @when step feed the docstring as the
FakeListLLM response so each scenario tests _validate()'s parsing of a
specific PASS/FAIL signal rather than the input code.
- features/tdd_plan_generation_validate_logic.feature: add the required
@tdd_issue tag alongside @tdd_issue_10746 (enforced by
features/environment.py); tighten scenario 4 wording to remove the
reference to the obsolete length guard.
- robot/plan_generation_graph.robot: drop assertion for handle_retry node
(retry is a conditional edge, not a fifth node) and update node-count
check from 5 to 4; update Should Retry test to assert _should_retry does
NOT mutate state (read-only contract for LangGraph conditional edges).
- src/cleveragents/agents/graphs/plan_generation.py:
* _validate: persist retry_count increment in return dict (FAIL path and
exception path) so LangGraph propagates it through the state graph.
* _should_retry: remove state mutation (conditional-edge functions are
read-only in LangGraph; mutations were silently dropped, causing
retry_count to remain 0 forever and the graph to loop infinitely).
Adjust comparison to retry_count <= max_retries because _validate has
already incremented before _should_retry runs.
* __init__: add max_context_files parameter (default 5, validated > 0)
and wire it into _format_context_summary in place of the hardcoded 5,
implementing the configurable-limits contract tested by
features/agent_configurable_limits.feature.
ISSUES CLOSED: #10746
- Added Given step handler 'I have a langgraph PlanGenerationGraph instance'
to create a PlanGenerationGraph with FakeListLLM in the test setup.
- Added Then step handler 'the langgraph validation status should be "{status}"'
to assert PASS/FAIL results from _validate().
- Fixed TDD tag format: replaced '@tdd_issue @tdd_issue_10746' with '@tdd_issue_10746'
per project convention (single tdd tag, not two).
These changes resolve all three review blocking issues for PR #10867:
1. Missing Given step handler causing test execution failures.
2. Missing Then step handler causing Behave StepDefinitionNotFoundError.
3. TDD tag format violation preventing CI from properly tagging tests.
The core code fix (removing length-based bypass in _validate) was already
correctly implemented and does not need changes.
ISSUES CLOSED: #10867
The `agents actor run` and `actor_run` Typer commands wrapped their main
try/except around `_resolve_config_files` with a bare `except
click.exceptions.Exit: raise` clause to let resolver-raised exits
propagate cleanly. Modern Typer re-exports `typer.Exit` from its vendored
click (`typer._click.exceptions.Exit`), which is NOT a subclass of
`click.exceptions.Exit`. So `_resolve_config_files`'s
`raise typer.Exit(code=2)` for unknown-actor / no-config-data / bad-blob
cases fell through to the generic `except Exception` clause, which
re-raised as `typer.Exit(code=3)` and replaced the targeted stderr
("not found in registry") with a generic "Unexpected error" message.
Behave scenarios at `actor_run_signature.feature:38` and `:56` and three
Robot integration scenarios verified the original exit-code/message
contract and failed. Catch `(typer.Exit, click.exceptions.Exit)` instead.
The five errored scenarios in `actor_run_signature_resolve_steps.py` and
`actor_run_signature_security_steps.py` had the same root cause from the
test side: `except (SystemExit, click.exceptions.Exit)` could not catch
the raised `typer.Exit`. Widen the tuple to include `typer.Exit`.
memory_service.py's `SQLChatMessageHistory(connection_string=...)` call
broke against langchain-community 0.4.2: the keyword was renamed
`connection` (which now accepts a URL string OR an Engine). Update the
call site and refresh the local `typings/langchain_community/...`
pyright stub to match the upstream signature so typecheck stays clean.
ISSUES CLOSED: #10921
- Remove misplaced pytest test files: tests/unit/a2a_test_http_transport.py,
tests/unit/__init__.py, and features/steps/test_a2a_http_transport_pytest.py.
Project layout uses Behave in features/ exclusively per CONTRIBUTING.md.
- Resolve AmbiguousStep crash in features/steps/a2a_facade_steps.py by
deduplicating step_transport_connect / step_transport_disconnect /
"the transport should not be connected" definitions left over from the
pre-implementation stub.
- Remove all `# type: ignore[arg-type]` comments (zero-tolerance policy).
- Fix ruff lint failures in src/cleveragents/a2a/transport.py: drop unused
imports (Any, map_domain_error, BaseHandler, OpenerDirector), wrap long
log lines (E501), and switch ssl.VerifyMode literal 0 to CERT_NONE for
pyright compliance.
- Update Robot helpers (robot/helper_a2a_facade.py,
robot/helper_m6_autonomy_acceptance.py) and the m6 / consolidated Behave
scenarios to verify the new server-mode lifecycle (connect succeeds with
valid URL, send-before-connect raises RuntimeError, invalid scheme raises
ValueError) instead of the obsolete "stub raises A2aNotAvailableError"
contract.
- Broaden the "I try to connect via the transport to ..." regex so the
invalid-URL scenario outline matches the empty-string / quoted / None
example cells; alias "I disconnect the transport" with @then so it is
reachable from `And` after a `Then` keyword.
Replace the A2aHttpTransport stub with a working HTTP(S) transport that:
- Connects to remote A2A servers via HTTPS (with configurable TLS verification)
- Sends JSON-RPC 2.0 requests over HTTP POST
- Parses JSON-RPC responses into A2aResponse objects
- Handles HTTP errors (4xx/5xx) with structured error mapping
- Supports JWT Bearer token authentication
- Validates connection state before send operations
Added comprehensive test coverage:
- Behave BDD scenarios for validation and lifecycle testing
- Pytest unit tests with mocked HTTP responses covering success,
errors, network failures, auth tokens, and roundtrip serialization
Three artifacts of the bad merge resolution on this branch are now corrected:
- [tool.coverage.report] section restored. The Robot integration test
``Coverage Threshold :: Noxfile Contains Coverage Threshold Constant``
asserts pyproject.toml contains ``fail_under = 96.5`` as the single
source for the coverage floor. The section was lost during merge-conflict
resolution; ``noxfile._read_coverage_fail_under`` was falling back to its
hard-coded default and the robot test was failing as a result.
- ``fastapi>=0.115.0`` (was 0.100.0). Master pins 0.115.0; the older floor
on this branch came in with the auto-scratch fix and is now bumped to
match.
- Duplicate ``langchain-anthropic>=0.2.0`` entry removed (line 40). Master
declares it once; the duplicate is a stray from the same bad merge.
Refs: #9055
Three step patterns in pyyaml_security_steps.py clashed with existing
step files, causing all Behave features to error at load time:
- "I call load_yaml_text with YAML text" clashed with actor_config_coverage_boost_steps.py:103
- "the load_yaml_text result should have key" clashed with actor_config_coverage_boost_steps.py:90
- "a ValueError should be raised" clashed with lsp_registry_steps.py:475
Rename all three to unique patterns and update pyyaml_security.feature
to match. Also fix typings/behave/runner.pyi ruff format (.pyi convention:
single blank line before class, no blank lines between stub methods) and
add missing fastapi>=0.100.0 to pyproject.toml (asgi_app.py imports
fastapi but it was absent from declared dependencies, causing typecheck
and integration test failures).
Refs: #9055
Remove ``# type: ignore[import-untyped]`` comments from features/steps/pyyaml_security_steps.py, replacing them with proper .pyi stubs for behave.runner.Context in typings/behave/runner.pyi.
Refs: #9055
Scenario 2 in features/pyyaml_security.feature referenced three step
definitions that were not implemented in the step file, causing
StepDefinitionNotFoundError and CI unit_tests failure:
- When I call load_yaml_text with YAML text "..."
- Then the load_yaml_text result should have key "..." equal to "..."
Added both missing step definitions with proper type annotations.
Also fixed dead code (except (ValueError, Exception) -> except Exception,
ruff B014) and toned down the alarmist assertion message per reviewer
feedback.
Added explicit pyyaml>=6.0.3 constraint to pyproject.toml to address
CVE-2017-18342 and related advisories. A codebase-wide audit confirmed
all YAML loading uses yaml.safe_load() exclusively via
cleveragents.actor.yaml_loader. Added BDD regression scenarios in
features/pyyaml_security.feature to verify the version constraint and
safe-load enforcement are maintained. Updated CHANGELOG.md with a
security entry.
ISSUES CLOSED: #9055
Address CI lint + unit_tests failures on the namespace-lock test suite:
- Step file: drop unused imports (yaml, ExitStack, ValidationError) and
unused noqa directives flagged by ruff F401/RUF100; rewrite the
collected-errors list to use iterable unpacking (RUF005); register the
"I create an ActorLoader with initial actors from multiple namespaces"
step under both @given and @when so scenario 3 is no longer reported as
undefined; remove the dead _error_collector inner function the reviewer
flagged.
- Step regex fix: the concurrent-modifications step pattern ended with
'' (two single quotes) instead of `` (two backticks), so it never
matched the feature file's `(triggering ``discover()``)` literal.
- Test fixtures: add the required `description` field to _make_actor_yaml
in both the BDD step file and tests/actor/test_loader_list_actors_thread_safety.py
so ActorConfigSchema validation passes (previously every scenario
errored at discover() with "description: Field required").
- Concurrent worker names: collapse three-slash actor names like
"conc/ns{i % 2}/concurrent_{i}" to the single-slash form
"conc/concurrent_{i}" required by the schema's namespaced-name rule.
- Test file: drop unused `yaml` import and three unused F841 assignments
in _list_worker; apply ruff format.
- Feature file: switch the @issue_8660 TDD tag to @tdd_issue_8588 to
match the CONTRIBUTING.md tag convention for the bug issue this PR
closes.
ISSUES CLOSED: #8588
Fixes a race condition where the namespace filter was applied outside
the threading.RLock, allowing concurrent mutations (discover(), clear())
to corrupt the iteration state. The filter now runs inside the locked
section, matching the locking discipline of all other public methods.
- Moved namespace filtering inside lock in list_actors()
- Added BDD concurrency regression test
- Added unit test for thread-safety under concurrent discover/clear
- Updated CHANGELOG.md with fix description
- Updated CONTRIBUTORS.md
ISSUES CLOSED: #8588
The `_format_table` helper documented itself as rendering an "ASCII
table" but built a Rich `Table` with the default `HEAVY_HEAD` box,
emitting Unicode box-drawing chars (│ ─ ┌). The new
@format_flag scenario "Tell table format outputs ASCII table"
asserts the output contains `|` or `+`, so the rendered table did
not match either the docstring contract or the BDD expectation.
Pass `box=box.ASCII` so the table actually uses `|` and `+` borders.
Also apply `ruff format` to the new step definitions
(four split-string concatenations the formatter wants collapsed
onto one line each).
Add BDD/Behave test scenarios for the existing --format/-f flag on `agents session tell`,
and update CHANGELOG.md and CONTRIBUTORS.md.
The implementation of --format on session tell exists in the codebase (commit 87a7ce35d),
but lacks dedicated BDD test coverage. This PR adds:
- 6 new Behave scenarios in features/session_cli.feature testing JSON, YAML, plain, table,
short flag (-f), and Rich output regression paths
- 6 corresponding step definitions in features/steps/session_cli_steps.py verifying
spec-compliant JSON envelopes, valid YAML/JSON output, ASCII table output, and Rich console
content preservation
- CHANGELOG.md entry under [Unreleased] documenting the --format flag feature
- CONTRIBUTORS.md entry crediting Jeffrey Phillips Freeman
Quality gates: lint ✓, typecheck ✓ (only pre-existing warnings about optional provider imports)
ISSUES CLOSED: #10466
- Return state update dict from _analyze_error using iterable unpacking
so existing messages are preserved (state.get + [new_message]) and the
RUF005 concatenation lint rule is satisfied
- Remove @tdd_expected_fail from tdd_auto_debug_analyze_error_mutation
feature now that bug #10494 is resolved
- Add BDD node-contract tests for _generate_fix, _validate_fix, _finalize
verifying each returns only the changed keys, not the full state
- Fix typer.Exit propagation in actor_run.py and actor.py: widen the
passthrough except clause from click.exceptions.Exit to
(click.exceptions.Exit, typer.Exit) so _resolve_actor's typer.Exit(2)
is not swallowed and re-raised as Exit(3)
- Add typer.Exit to Behave step except clauses in
actor_run_signature_resolve_steps.py and actor_run_signature_security_steps.py
so test scenarios capture the exit code instead of erroring
- Fix SQLChatMessageHistory call in memory_service.py: rename kwarg
connection_string to connection per langchain_community 0.4.x API change
ISSUES CLOSED: #10496
All four LangGraph node functions (_analyze_error, _generate_fix, _validate_fix, _finalize)
were violating the node contract by mutating input state and returning the full state object.
They now correctly return dicts containing only the changed keys.
Three CI gates were failing on this PR; this commit addresses the root
causes for each:
* lint (ruff format): drop the blank line between the docstring close
and first statement in step_pr_create_with_error, and add the missing
second blank line between step_pr_check_remove_link_persisted and the
"Data integrity BDD step extensions" section comment block.
* unit_tests: two scenarios were inverted by `@tdd_expected_fail` on
post-fix assertions, masking unrelated test-logic problems.
- Remove `@tdd_expected_fail` from both `@tdd_issue_8179` scenarios in
project_repository.feature - they describe post-fix behaviour and
must report PASS as PASS, not as inverted-FAIL.
- Drop the "Given project exists" precondition from the Update-non-
existent scenario; the Background already initialises the in-memory
DB and creating the same project being "updated as non-existent" is
self-contradictory (caused the prior scenario to silently report
inverted-PASS while actually never raising).
- Update the OperationalError scenario in database_repository_coverage
to assert the post-fix invariant: the repository no longer calls
session.rollback() itself; that responsibility is delegated to the
outer UnitOfWork. Step text + assertion both flipped.
ISSUES CLOSED: #8179
Removed unconditional session.rollback() calls within exception handlers in:
- ProjectRepository.create()
- NamespacedProjectRepository.create() (IntegrityError handler)
- NamespacedProjectRepository.create() (OperationalError handler)
- NamespacedProjectRepository.update()
- NamespacedProjectRepository.delete()
The Unit of Work pattern already handles transaction rollback at the outer layer
via its except Exception: session.rollback() handler, making these inner rollbacks
redundant. SQLAlchemy automatically invalidates the transaction state when exceptions
occur after flush(), preventing partial data from being committed.
Removing the redundant rollbacks improves clarity, eliminates potential issues related
to exception chaining across retry boundary layers, and aligns repository implementations
with explicit transaction boundaries.
ISSUES CLOSED: #8179
Add missing [Unreleased] CHANGELOG entries for the A2A stdio transport
feature and the .py path routing fix, both referencing the correct
issue #691 (not #264 which was already closed in v3.0.0).
ISSUES CLOSED: #691
step_mock_popen_success stored mock_proc (Popen's return value) as
context.popen_mock, but call_args is recorded on the mock *replacing*
subprocess.Popen (what patcher.start() returns). Reading call_args from
mock_proc returns None, causing TypeError in the three command-construction
scenarios — behave reports these as "errored" not "failed".
Fix: assign patcher.start() to context.popen_mock so the assertion steps
read call_args from the correct mock. Also remove the redundant
patcher.stop() calls from the assertion Then steps (context.add_cleanup
already handles teardown). Add the required @tdd_issue and @tdd_issue_691
tags to the Connect with .py file path scenario per the TDD bug fix workflow.
ISSUES CLOSED: #691
controller-ci-rerun
HAL9000
CleverThis
cleveragents-auto
freemo