PROTOTYPE 14:42 UTC
Back to Fleet Intelligence
AI Quality
AI quality drift over 7 days. No degradation detected. 0 hallucinations week-to-date.
Validator pass
99.6%
7d range 99.3–99.6
Citation rate
100%
7d range 99.8–100
Avg latency
2.4s
target ≤ 3.0s
Hallucinations
0
week-to-date
Drift over 7 days
3 metrics · no degradation detected
Validator pass
99.6%
7d range 99.3 — 99.6
Avg latency
2.4s
7d range 2.4 — 2.6s · target ≤ 3.0s
Citation rate
100%
7d range 99.8 — 100% · every response cited
Quality discipline

Concierge runs a dual-validator quality gate: Sonnet generates the response; Opus validates relevance and precision before sending. Every response resolves to a specific source record (PostgreSQL row, Pinecone vector, or both). Citation isn't a heuristic; it's an architectural floor. 7 days of operation, 0 hallucinations.

When validators flag a response (precision concern, off-topic content, or critical safety/PII issue), the system either rewrites silently, hedges with caveats, or pauses for human review. Detail breakdown lives at the validator decisions view.

View Validator Decisions →

Action confirmed.
PROTOTYPE · synthetic data · all guest records and entities are illustrative