A daily read on your BankiPay. Generated by your AI.

AI Insight lives on the BankiPay dashboard. Every visit, it looks at your last 2h / 7d activity and tells you what changed — in a tone matched to what it found.

See it in action Read the docs

What you get

LLM-powered

Calls your active provider (default: MiniMax / MiniMax-Text-01). Returns a 1–2 sentence insight + tone tag (info / warning / success).

Provider tag visible

Every insight shows which model generated it. e.g. "MiniMax · MiniMax-Text-01". Full transparency.

Deterministic fallback

If the LLM call fails (rate limit, network), we fall back to a template-string insight. You always see something useful.

3 tone modes

Info (neutral), warning (something to look at), success (something good). The color and icon change accordingly.

Privacy-first

Only numeric metrics (count, volume, etc.) are sent to the LLM. No PII, no merchant names, no counterparty.

Refreshes on visit

Each dashboard load triggers a fresh insight. No stale data. Debounced 60s on rapid tab switches.

In practice

AI inside · Anya · checking BankiPay in the morning

Anya opens her BankiPay. The AI Insight card reads: "You have 3 pending transactions from yesterday — the $240 transfer to Maya is still settling. Your 7d volume is up 18% — looks like a busy week."

~3s

Avg generation time

At a glance

~200

Max output tokens

0.4

Temperature

Compatible providers

60s

Refresh debounce