fix: @Bartok9 contributions (batch 2)

[maciejmajek]

Purpose

• Salvage the second round of open PRs from @Bartok9 (fix(tools): reject non-positive topic receive/image/TF timeouts #827 to fix(bench): reject negative TaskArgs.extra_tool_calls #869) into one change. Most of them add an imperative guard plus a helper module plus a test file each, around 1400 added lines in total, with five PRs shipping a byte-identical copy of the same helper.

Proposed Changes

• Tool argument validation moves to the existing pydantic args_schema as Field(min_length=1), Field(gt=0) and Field(allow_inf_nan=False). Enforced before _run, visible in the schema the model sees, and ToolRunner already turns ValidationError into a ToolMessage. Replaces fix(tools): reject non-positive topic receive/image/TF timeouts #827, fix(tools): reject non-positive CallROS2ServiceTool timeout_sec #828, fix(tools): validate configured image/TF tool timeouts #829, fix(tools): reject empty ROS topic names on publish/receive/image #831, fix(tools): reject empty ROS service name/type on CallROS2ServiceTool #832, fix(tools): reject empty ROS action name/type on StartROS2ActionTool #833, fix(tools): reject empty TF frame names on GetROS2TransformTool #834, fix(tools): reject empty topic/message_type on PublishROS2MessageTool #835, fix(nav2): reject non-finite NavigateToPose coordinates #836 in ~15 lines with no new modules.
• hri_connector.py: AI multimodal messages passed base64_images as audios, so images-only messages raised (fix(hri): pass base64_audios for AI multimodal to_langchain #847).
• configurator.py: ASR selectbox read transciption_model while the handler writes transcription_model (fix(frontend): read transcription_model key consistently in ASR config UI #859).
• artifacts.py: store_artifacts creates the parent directory for a new db_path (fix(messages): honor db_path in store_artifacts #848).
• tool_runner.py: unknown tool names report Unknown tool instead of a bare KeyError repr (fix(tool_runner): return error ToolMessage for unknown tool names #853).
• VADConfig and WWConfig validate thresholds in __post_init__, matching TranscribeConfig (fix(s2s): validate OpenWakeWord detection threshold #854, fix(s2s): validate SileroVAD detection threshold #856, fix(s2s): validate VADConfig and WWConfig thresholds and grace #866).
• Not taken: fix(ros2): reject non-positive waiters time_interval #855 (our own tests pass time_interval=0 with time.sleep patched), the rai_bench task guards (fix(bench): reject non-positive PlaceCubesTask threshold_distance #860, fix(bench): validate GroupObjectsTask threshold and obj_types #861, fix(bench): validate PlaceObjectAtCoordTask displacement and coordinates #863, fix(bench): validate RotateObjectTask obj_types and quaternion #864, fix(bench): reject empty MoveObjectsToLeftTask obj_types #865, fix(bench): validate BuildCubeTowerTask obj_types and displacement #867, fix(bench): reject empty Scenario scene_config_path #868, fix(bench): reject negative TaskArgs.extra_tool_calls #869) and the remaining rai_s2s model guards (fix(s2s): validate SoundDeviceConfig block_size and channels #857, fix(s2s): validate BaseTranscriptionModel sample_rate and names #858, fix(s2s): reject non-positive TTSModel sample_rate and channels #862), which validate parameters only we set, and the AGENTS.md hunk in fix(bench): validate PlaceObjectAtCoordTask displacement and coordinates #863.
• rai_core 2.12.3, rai_s2s 1.0.1, uv.lock regenerated.

Issues

• Supersedes fix(tools): reject non-positive topic receive/image/TF timeouts #827, fix(tools): reject non-positive CallROS2ServiceTool timeout_sec #828, fix(tools): validate configured image/TF tool timeouts #829, fix(tools): reject empty ROS topic names on publish/receive/image #831, fix(tools): reject empty ROS service name/type on CallROS2ServiceTool #832, fix(tools): reject empty ROS action name/type on StartROS2ActionTool #833, fix(tools): reject empty TF frame names on GetROS2TransformTool #834, fix(tools): reject empty topic/message_type on PublishROS2MessageTool #835, fix(nav2): reject non-finite NavigateToPose coordinates #836, fix(hri): pass base64_audios for AI multimodal to_langchain #847, fix(messages): honor db_path in store_artifacts #848, fix(tool_runner): return error ToolMessage for unknown tool names #853, fix(s2s): validate OpenWakeWord detection threshold #854, fix(s2s): validate SileroVAD detection threshold #856, fix(frontend): read transcription_model key consistently in ASR config UI #859, fix(s2s): validate VADConfig and WWConfig thresholds and grace #866
• Closes as already on main via fix: @Bartok9 contributions #824: fix(ros2): reject non-positive get_future_result timeout_sec #830, fix(communication): require positive callback_max_workers #849, fix(agents): reject empty agents list in runner helpers #850, fix(agents): require positive HRICallbackHandler max_buffer_size #852
• Declined: fix(ros2): reject non-positive waiters time_interval #855, fix(s2s): validate SoundDeviceConfig block_size and channels #857, fix(s2s): validate BaseTranscriptionModel sample_rate and names #858, fix(bench): reject non-positive PlaceCubesTask threshold_distance #860, fix(bench): validate GroupObjectsTask threshold and obj_types #861, fix(s2s): reject non-positive TTSModel sample_rate and channels #862, fix(bench): validate PlaceObjectAtCoordTask displacement and coordinates #863, fix(bench): validate RotateObjectTask obj_types and quaternion #864, fix(bench): reject empty MoveObjectsToLeftTask obj_types #865, fix(bench): validate BuildCubeTowerTask obj_types and displacement #867, fix(bench): reject empty Scenario scene_config_path #868, fix(bench): reject negative TaskArgs.extra_tool_calls #869
• fix(tool_runner): avoid leaking tool exception details to LLM (CWE-209) #814 touches the same except block in tool_runner.py and needs a rebase after this lands

Testing

• tests/tools/ros2/test_input_validation.py covers every schema constraint, tests/s2s/test_asr_config.py covers the config guards, and the bug fixes are asserted in the existing test files.
• pytest tests/tools tests/agents tests/messages tests/communication tests/s2s: 343 passed, 4 skipped, 11 xfailed.

Co-authored-by: Bartok danielrpike9@gmail.com

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.49%. Comparing base (54f2423) to head (84f1d03).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #872      +/-   ##
==========================================
+ Coverage   74.48%   74.49%   +0.01%     
==========================================
  Files          82       82              
  Lines        3621     3627       +6     
==========================================
+ Hits         2697     2702       +5     
- Misses        924      925       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

[boczekbartek]

LGTM! Thanks for refactoring these PRs