3 Commits

Author	SHA1	Message	Date
hurui200320	0c5724c2f6	feat(actor-run): wire ToolCallingRuntime into actor run for skill-based tool calling ... Implements the full tool-calling path for `agents actor run --skill` so that LLM actors can actually invoke tools through the ToolCallingRuntime loop when a skill is attached. Core changes: - reactive/tool_caller.py (new): ToolCallingLLMCaller implements the LLMCaller protocol; binds tool schemas via bind_tools(), threads SystemMessage+HumanMessage on first call and AIMessage+ToolMessages on subsequent calls, extracts tool calls from LangChain responses following the LangChainSessionCaller pattern. - reactive/tool_caller.py: bidirectional tool name encoding via uppercase sentinels (_encode_tool_name / _decode_tool_name) to make CleverAgents namespaced tool names ("builtin/file-read", "server:local/tool") compatible with Anthropic's tool name pattern ^[a-zA-Z0-9_-]{1,128}$. Uses _C_ for ":" and _S_ for "/" — uppercase sentinels are safe because valid CleverAgents tool names forbid uppercase letters. Encoding applied in _resolve_llm() before bind_tools(); decoding applied in invoke() when extracting tool calls from the LLM response. - reactive/tool_agent.py (new): ToolCallingAgent builds a per-run local ToolRegistry from resolved skill tool entries by looking up names in the shared builtin_registry; drives ToolCallingRuntime.run_tool_loop(); exposes last_result for tool_calls surfacing. - reactive/application.py: (ST-1) _make_agent_instance() now always merges skill tools instead of silently dropping them when actor has no base tools list; routes tools+llm→ToolCallingAgent, tools+non-llm→SimpleToolAgent, no-tools+llm→SimpleLLMAgent; (ST-4) _builtin_registry created at startup with register_file_tools/git/subplan; (ST-6) _tally_tool_calls() + last_run_tool_calls property. - reactive/graph_executor.py: (ST-5) ToolCallingAgent added to isinstance check in _invoke_agent() so context dict is forwarded for Jinja2 rendering. - cli/commands/actor_run.py: prints "Tool Calls: {n}" when > 0. Test fixes: - features/steps/actor_cli_run_steps.py: _make_app() sets last_run_tool_calls=0 to avoid MagicMock>int TypeError in Python 3.13. - features/steps/actor_run_signature_resolve_steps.py: same fix. - robot/helper_actor_run_signature.py: same fix. - features/reactive_application_coverage_boost.feature: updated scenario to verify new correct behavior (LLM+skills → ToolCallingAgent, not silently kept as SimpleLLMAgent). BDD coverage: 34 scenarios in features/actor_run_tool_calling.feature covering tool call success, multi-turn loop, no-skill regression, silent-drop fix, LLMCaller internals, _build_tool_registry edge cases, last_run_tool_calls tallying, tool name encoding/decoding, and LLM response decoding. ISSUES CLOSED: #11211	2026-05-18 06:05:54 +00:00
hurui200320	554d6889cc	fix(cli): add --skill flag to actor run command (#971) ... ## Summary Add the missing `--skill` repeatable flag to `actor run` and `actor-run` CLI commands, aligning the implementation with the specification (CLI Synopsis line 277). The flag enables ad-hoc skill injection at runtime without modifying YAML configuration. Closes #887 ## Changes ### DI Container - **`container.py`**: Added `_build_skill_service()` factory and `skill_service` Singleton provider, following the established `_build_*` pattern. Falls back to in-memory `SkillService()` when the database is unavailable. Exception handling narrowed to `(ImportError, OperationalError, DatabaseError, OSError)` with `exc_info=True` for traceability. ### CLI Layer - **`actor.py`**: Added `--skill` Typer option (`list[str] | None`, repeatable, `metavar="NAME"`). Help text notes that skills only augment tool-bearing agents. Wrapped constructor in the existing `try/except` block so `CleverAgentsException` from skill resolution is properly caught. - **`actor_run.py`**: Same `--skill` option with `metavar="NAME"`. Exception handler catches `CleverAgentsException` (matching master — not broadened to `CleverAgentsError`). - **`skill.py`**: Removed module-level `_service` cache. `_get_skill_service()` now always delegates to `get_container().skill_service()` so that `reset_container()` correctly invalidates the cached instance. `_reset_skill_service()` now overrides the container's provider via `providers.Object()`. Removed dead `validate_skill_names()` function. ### Runtime Layer - **`application.py`** (438 lines, down from 625): `ReactiveCleverAgentsApp` gains `skill_names` parameter with automatic deduplication via `dict.fromkeys`. `_resolve_skills()` obtains `SkillService` from the DI container (no CLI layer import). Separate `except KeyError` and `except ValueError` produce distinct error messages (`"not found in registry"` vs `"resolution failed: {exc}"`). Skill tools are only injected into agents that already have tools (`if self._resolved_skill_tools and tools:`), preventing LLM agents from being converted to pass-through `SimpleToolAgent` instances. When skill tools are skipped for tool-less agents, `logger.debug` emits a diagnostic message. `_sanitize_skill_name()` validates skill name format with tightened regex: `^[\w.-]{1,127}/[\w.-]{1,127}$` with `re.ASCII` flag. Zero-tool skill warning now uses `logger.warning` (not `print(stderr)`), ensuring structured log output and proper log-level filtering. - **`graph_executor.py`** (334 lines): Extracted graph execution logic. Type annotations improved. ### Tests - 24+ Behave scenarios across feature files covering: single/multiple/unknown skill flags, skill+context combined, duplicate deduplication, skill resolution, ValueError path, zero-tool resolution, error handling, tool merging, default behavior, overrides, LLM agent guard, `_sanitize_skill_name` edge cases (empty string, too-long name, ANSI escape codes, disallowed characters), `_build_skill_service` happy+fallback paths, `_get_skill_service` container delegation. - CLI "unknown skill" tests for **both** `actor.py` and `actor_run.py` exercise the real error chain (mock only `get_container()`, not the entire `ReactiveCleverAgentsApp`), testing `_resolve_skills()` → `CleverAgentsException` → `except CleverAgentsException` → exit code 2 end-to-end. - Combined skill+context tests assert `ContextManager` was instantiated and `exists()` was called in dedicated **Then** steps. - `@coverage` tags added to all new scenarios. - **Robot Framework smoke tests** added (`robot/skill_actor_run.robot` + `robot/helper_skill_actor_run.py`): unknown-skill error path and valid-skill acceptance path. ### Changelog - Added entry under `## Unreleased` in `CHANGELOG.md`. ## Review Fixes Applied (Brent Edwards, Rounds 1 & 2) | # | Finding | Resolution | |---|---------|------------| | **P1-1** | `print(stderr)` for zero-tool skill warning | **Fixed** — replaced with `logger.warning("Skill '%s' resolved to zero tools", name)`, removed unused `import sys` | | **P2-2** | Skill tools silently skipped for tool-less agents | **Fixed** — added `logger.debug` when skipped; updated `--skill` help text to note "only augments tool-bearing agents" | | **P2-3** | `container.py` at 739 lines | **Acknowledged** — pre-existing growth (+59 lines for `_build_skill_service`); extracting factories is a separate refactoring task | | **P2-4↑** | `CleverAgentsException` → `CleverAgentsError` broadens catch scope | **Fixed** — reverted `actor_run.py` to `except CleverAgentsException` matching master | | **P3-5** | No Robot Framework smoke test for `--skill` | **Fixed** — added `skill_actor_run.robot` with 2 test cases (unknown-skill error, valid-skill acceptance) | | **P3-6** | `GraphExecutor._follow_chained_edges` static-calling-static | **Acknowledged** — cosmetic pattern that doesn't affect correctness; can address in a follow-up | ## Known Limitations / Deferred Items | Item | Reason | |------|--------| | `actor.py` at 679 lines (500-line guideline) | Pre-existing (670 on master), +9 lines for `--skill`. Refactoring the shared `_execute()` closure is a separate task. | | `container.py` at 739 lines (500-line guideline) | Was 680 lines on master, +59 lines for `_build_skill_service()` and `skill_service` provider. Refactoring into sub-modules is a separate task. | | Code duplication between `actor.py` and `actor_run.py` `run()` | ~47 lines identical code. Coupled with the line-count issue above — both require extracting shared execution logic into a helper module. | | `SimpleToolAgent` only executes `tools[0]` | Deferred to #974. Pre-existing architectural limitation, not introduced by this PR. | | `GraphExecutor._follow_chained_edges` static-calling-static pattern | Cosmetic, doesn't affect behavior. | ## Quality Gates - `nox -s lint`: ✅ PASS - `nox -s typecheck`: ✅ PASS (0 errors) - `nox -s unit_tests`: ✅ PASS (11,130 scenarios, 0 failures) - `nox -s integration_tests`: ✅ PASS (1,559 tests, 0 failures) - `nox -s coverage_report`: ✅ 97% (meets threshold) - Branch rebased onto latest `master` (`ab1fd19b`) Reviewed-on: #971 Co-authored-by: Rui Hu <rui.hu@cleverthis.com> Co-committed-by: Rui Hu <rui.hu@cleverthis.com>	2026-03-19 09:30:35 +00:00
freemo	a808c395f9	test(coverage): add Behave BDD tests to improve unit test coverage across 53 source modules ... Add 53 new .feature files and corresponding step definition files targeting uncovered lines identified in build/coverage.xml. Fix AmbiguousStep conflicts in 7 pre-existing step files by disambiguating step text. New tests cover: ACP clients/facade, actor CLI/config, application container, ACMS service/strategies, async worker, automation profile CLI, autonomy guardrail, bridge, change model, config CLI/service, context service, cross-plan correction, database models, decision service, decomposition clustering/service, discovery handler, langchain chat provider, langgraph nodes, materializers, multi-project service, plan apply/CLI/lifecycle/model/ preflight/resume/service, PostgreSQL analyzer, project CLI/context CLI, provider registry, reactive application/route, repositories, resolver handler, resource registry service, resume model, retry patterns, sandbox protocol, server CLI, skill CLI/service, skills registry, subplan execution/service, system CLI, UKO loader, UoW, and YAML template engine. Closes #645	2026-03-09 13:01:58 -04:00