1. Backfill Null Metadata

Purpose: Patch facts that are missing notability scores, challenge content, or other NULL metadata fields.

Prerequisites:

  • Supabase credentials
  • AI model API key (for notability scoring and challenge generation)

Cost / Duration: ~$0.005/fact for notability, ~$0.006/fact for challenges | varies by volume

Prompt

Backfill null metadata on existing facts.

Option A — Backfill everything:

```bash
bun scripts/seed/backfill-fact-nulls.ts --all
```

Option B — Backfill specific fields only:

```bash
# Notability scores only
bun scripts/seed/backfill-fact-nulls.ts --notability

# Challenge content only
bun scripts/seed/backfill-fact-nulls.ts --challenge-content
```

Before running, check how many nulls exist:

```bash
bun scripts/seed/backfill-fact-nulls.ts --audit
```

This will report the count of facts needing backfill per field, so you
can estimate cost before committing. Monitor progress in the console
output — the script reports batch completion with running totals.

Verification

  • Audit run first to confirm scope and estimate cost
  • Backfill completes without errors
  • Re-run --audit after to confirm null counts decreased
  • Notability scores are in valid range (0-100)
  • Challenge content passes validateChallengeContent() checks
  • Cost reported matches pre-run estimate (within 20%)

Back to index