Enrich data via web search
web_search_enricherskillsetup L2★0
TrevorMann/AIDataCleansing ↗What it does
Resolve missing or ambiguous data fields via web search
Best for
Filling postal code → municipality gaps in data records when deterministic database lookups fail and confidence is low.
Inputs
- · Record with gaps: postal_code, city, address, state_province, _triage_route, _triage_data_confidence, _unknown_fsa, _municipality_confidence, _gap_hints
Outputs
- · Enriched record with resolved fields: municipality, _web_search_evidence (audit trail), _decisions
Requires
- · Tavily API (web search, rate-limited and billed)
Preconditions
- · _triage_route == 'needs_review'
- · _triage_data_confidence < trigger_below (default 0.70)
- · At least one identifiable gap
- · Per-batch budget not exhausted
- · Deterministic skills run first
Failure modes
- · Tavily API rate limit or budget exhausted → no search performed
- · Parser plugin returns no useful data → low-confidence signal only
- · Web search results incorrect → output as low-confidence signal, not ground truth
- · Gap hints ambiguous → skill may over-search
Trust signals
- · Gating logic prevents wasteful searches (route + confidence + gap checks)
- · Audit trail logged in _web_search_evidence
- · Domain-agnostic core with per-domain parser plugins
- · Web cache reuses results (TTL managed)