Skip to content

Latest commit

 

History

History
14 lines (11 loc) · 1.86 KB

File metadata and controls

14 lines (11 loc) · 1.86 KB

Script Helpers

This folder contains local developer helpers and repository-maintenance scripts. They are not imported by the add-on at runtime.

  • assess.ps1 runs focused assessment checks against a live environment.
  • api_surface_smoke.py runs the HTTP API in-process with FastAPI TestClient, seeded SQLite state, and local stubs so the API surface can be regression-tested without deploying to Home Assistant. It covers health, dev status, topology/history, partitions, routing, stale links, network data, assessment, chat telemetry, and the prompt corpus through /v1/chat/turn.

Run it locally with PYTHONPATH=addons/thread-observability/app/src python scripts/api_surface_smoke.py.

  • chat-smoke.ps1 exercises chat flows against the add-on, including persisted transcript inspection through /v1/chat/transcript/{conversation_id} when transcript persistence is enabled. Semantic pass/fail now comes from chat_smoke_evaluator.py, which calls an OpenAI-compatible reviewer model using THREAD_OBS_SMOKE_EVAL_PROVIDER, THREAD_OBS_SMOKE_EVAL_BASE_URL, THREAD_OBS_SMOKE_EVAL_MODEL, THREAD_OBS_SMOKE_EVAL_API_KEY, and optional THREAD_OBS_SMOKE_EVAL_TEMPERATURE.
  • direct_chat_prompt_smoke.py is the prompt-focused Python smoke path: it takes a list of pre-authored likely user queries, runs them through the real direct_chat_turn orchestration in-process, uses AI to judge each final answer, and when a case fails asks AI to review the full transcript including the embedded reviewer/audit step. It also supports model comparison with --targets <json>.
  • dashboard-loop.ps1 repeats dashboard-oriented checks during live validation.
  • generate_mcp_reference.py regenerates the MCP reference documentation from the live tool registry.
  • test_real_logs.py is an ad hoc OTBR parser smoke helper for quickly checking a few real log lines outside the automated test suite.