Dialer

Backfill Grading

CONTACTS NOBODY - but it SPENDS, and it can spend a lot.

With dry_run=false this runs AI grading over EVERY ungraded call with a transcript in the window - a model invocation each, one call every two seconds, with no ceiling other than how many calls exist. On a busy organisation over 365 days that is an unbounded bill from one request.

dry_run defaults to TRUE and genuinely short-circuits before any grading (parallel_calls.py:4602), returning the eligible count and a sample. Call it that way first and read the count.

Tiered billable, not write: it reaches a paid model per record.

post/dialer/parallel-calls/grading/backfill

Request

  • Base URL: https://be.graph8.com/api/v1
  • URL: https://be.graph8.com/api/v1/dialer/parallel-calls/grading/backfill
  • Auth: HTTP bearer

Query parameters

daysinteger

How far back to look.

dry_runboolean

true counts and samples; false GRADES AND SPENDS.

Headers

Authorizationstring nullable
X-Target-Org-Idstring nullable

Response

Successful Response

dataDialerPayload required

An open envelope for one dialer result.

Open because voice owns the shape of a session, a call stub and a queue stat, and it grows them without a contract bump here. A dropped field would otherwise turn an added upstream field into a removed public one.

Changes

No recorded changes to this operation across all 2 revisions of this API.