| name | health-check |
| description | Self-diagnostic for the aifolimizer harness wiring. Use when something is silently broken or before relying on the system - "is everything working?", "health check", "why is the MCP not responding?", "check my setup", "diagnose aifolimizer". Audits MCP server, services, WS token freshness, hooks. Catches stale-token / unregistered-MCP / broken-hook silent failures. |
Health Check (Harness Self-Diagnostic)
Goal
Catch the silent-failure class a solo operator hits with no second pair of eyes:
stale WS token, MCP server that won't import, a service import break, missing
hooks. Output is a PASS/WARN/FAIL report with the one thing to fix.
How to run
Step 1 - Run the deterministic diagnostic:
backend/.venv/Scripts/python.exe backend/scripts/health_check.py
Reports: python version, mcp_server import + tool count, core service imports,
WS session token freshness (vs WS_TOKEN_TTL_HOURS), settings.json hook events.
Step 2 - Live data-source check (only if Step 1 PASS):
mcp__aifolimizer__get_data_source_reliability - per-source success rate +
latency over the trailing window. Flags providers degrading (yfinance / FRED /
CoinGecko / SEC).
Step 3 - Optional backend reachability:
curl -s http://127.0.0.1:8000/health if the FastAPI backend is expected up.
Interpreting results
- ws_session WARN (stale) → re-auth: delete
~/.aifolimizer/ws_session.json
and re-run WS login, or run mcp_login.
- mcp_server_import FAIL → a recent edit broke import; run
py_compile mcp_server.py and read the traceback.
- core_services FAIL → a service has a syntax/dependency error; the detail
names the module + exception.
- settings_hooks missing event → a hook didn't register; check
~/.claude/settings.json.
- data_source low success rate → provider outage or rate-limit; the cached
layer will serve stale data until it recovers.
Output
Relay the script's OVERALL line + any non-PASS rows, then the single
highest-priority fix. Don't paste the full report unless asked.
