Open the method.
Make the boundary clear.

LLM privacy · Kimi · Reference

BridgeTime Kimi Privacy

The BridgeTime merchant assistant uses local-first Private Intent. Deterministic BridgeTime workflows list staff and services; create staff, tier, service, duration, price, and assignments in one operation; build simple recurring schedules; and guide merchants through staff PINs, account invitations, the staff portal, and customer booking links. These common paths make no Kimi request and use zero provider tokens.

The interaction boundary is equally explicit: the assistant is text-first. A complete, low-risk single-item configuration may run directly. A batch containing multiple staff, services, or assignments is first parsed entirely inside BridgeTime and shown as one read-only list; no staff, service, or assignment record is created until the merchant confirms the whole batch once, after which one transaction applies it to the merchant's live data. Confirmation controls appear only below that batch list or for delete, disable, remove, overwrite, or another hard-to-recover action requiring the account holder's authorization—never once per staff member or service. If a required field is missing, the assistant asks in text. Enter inserts a line break; Run is the single persistent submit control; the “?” beside the input only opens a local guide to supported keywords and exact sentences, never submits, writes data, or calls Kimi. PINs and account invitations continue in dedicated secure screens, are not written into chat, and never enter the LLM.

Only an unclassified interface-routing fallback may send the five abstract enum fields schema · action · entity · source · stage. Raw chat, names, business and service labels, prices, IDs, counts, schedules, dates, times, PINs, passwords, and customer data do not enter the Private Intent payload; requests containing sensitive values are blocked locally. The public project verifies outbound fields and wire bodies—not deployment state, provider retention, compliance, or zero risk. Kimi naming identifies API compatibility only and does not imply affiliation, certification, or endorsement.