Skip to content

Automatic translation — the Portuguese version is the reference. View in Portuguese

Updated on 2026-09-08

Troubleshooting: AI, cost and performance

Symptom → cause → how to check → fix → prevention catalog for AI features (text-to-SQL, insights, Q&A), budget/egress and query cost/performance.

On this page (5)

AI (text-to-SQL, insights, Q&A) has a budget with reservation→settlement, controlled egress and provenance in its answers. Data cost comes from processing in the datalake, with a cost guard and cache. See Privacy, egress and budget and Usage and quotas.

AI (text-to-SQL, insights, Q&A)#

SymptomLikely causeHow to checkFixPrevention
AI refuses the actionAI credits exhausted (fail-closed ledger)check the balance in Usagetop up creditskeep an eye on the monthly allowance
AI "doesn't respond" / unavailableLLM not configured or kill switch onsee Statuswait; confirm the configurationmonitor the capability's status
AI answer without numbers/explained refusalQ&A requires evidence/citationread the answer's provenancerephrase with a clear metric/periodask about published data (Gold)
Generated SQL blockedtext-to-SQL defenses (SELECT only, published marts)see the 403 messagefit the question to the allowed scopequery only published marts
AI usage higher than expectedbroad/repeated questionssee the Audit log (cost per action)narrow the scope; reuse answersencourage objective questions

Cost and performance#

SymptomLikely causeHow to checkFixPrevention
Query cut off due to costhit the per-query volume capsee the run's error/costfilter/aggregate; shorten the periodmodel a lean Gold; partition
High projected bill/costlarge table scansee bytes in the Audit logfilter/aggregate; use partitionsavoid paging through a whole table
Paging the API doesn't reduce costLIMIT/OFFSET don't cut bytesbilling is by actual bytesuse $select/$filter (OData)prefer pushdown over raw paging
Dashboard slow on refreshheavy query / no cachecompare the 1st run vs. a repeatidentical queries are served from cache (R$ 0)schedule refresh off-peak
Manual refresh blocked (402)overdue payment (not performance)see Suspensionthe owner settles itkeep an eye on billing
Query over billions of rows is slowvolume too large for a direct query\1aggregate in Gold and use the Import modelsize with headroom

BRL values are projections today

x-ingestia-cost-brl and the per-event cost are a projection, not a guaranteed billed price. The per-query volume cap limits the cost of each query.

Quick cost check#

  1. Open the Audit log and sort by cost.
  2. Expand the SQL of the expensive event; check the bytes processed.
  3. Rewrite with filtering/aggregation; confirm that the repeat is served from cache.

Security and privacy#

  • Controlled egress: AI doesn't leak data beyond what's necessary; answers carry provenance — see Privacy, egress and budget.
  • Anti-injection and kill switches protect the AI path.

Related#


Last reviewed: 2026-09-08. Cost values on screen are estimates; the amount charged is the one on the invoice.

Related links