pipeline-doctor

name	pipeline-doctor
description	Use when checking whether the data + model pipeline is healthy, when returning to the project after time away, or when a workflow looks stuck. Runs the read-only health snapshot and triages any breach.
argument-hint	[--refresh]

/pipeline-doctor — One-screen pipeline health + triage

Strictly read-only on the money path — never places bets, never writes the ledger. It composes pipeline_health() (fit/odds freshness, persisted Stan diagnostic drift, orphaned bets, bankroll/drawdown) into a single status table and helps triage anything that is WARN or FAIL.

1. Refresh the snapshot

cd /Users/brynjolfurjonsson/sports && Rscript scripts/07_healthcheck.R

This prints the per-check table, lists the breaches, and writes data/health/status.json (overall status OK < WARN < FAIL, plus PAUSED for an off-season cell). It always exits 0 — it is observability, not a gate.

2. Add operational context

# Recent data commits — did the crons actually run?
git -C /Users/brynjolfurjonsson/sports log --oneline -8 -- data/

# Recent CI runs and their conclusions (needs gh + a GitHub remote)
gh run list --repo metill-is/sports --limit 12

3. Triage each breach

For every WARN/ row, map the to its runbook in and follow the symptom → diagnose → fix → verify steps: