Skip to main content
POST
Prepare the read of a turn

Authorizations

Authorization
string
header
required

nia_sk_...

Body

application/json

POST /v1/context/prefetch: the customer's turn so far, sent while they are still speaking, so the read that answers the turn finds its work done (memory v2). Same target, view and level as that read.

query
string
required

The partial transcript. The read with this same query is served from the prefetch; a longer transcript still finds the customer's memory already open.

Required string length: 1 - 2000
about
Handle · object | null

The account or partner the person acts for.

conversation_id
string | null
Required string length: 1 - 512
object
ObjectRef · object | null

A business object in a system of record, e.g. invoice / erp / 0823.

subject
Handle · object | null

An identifier of a subject in some channel or system: a phone, an e-mail, a CRM id.

task_id
string | null
Required string length: 1 - 512
verification
enum<string>
default:V0

Session verification levels. no_customer sits outside the V0-V4 scale.

Available options:
V0,
V1,
V2,
V3,
V4,
no_customer
view
string
default:voice
Pattern: ^(voice|chat|brief|full|custom|account|partner|task:[a-z0-9_]{1,40})$

Response

Successful Response

queued
boolean
required

False when there is nothing to prepare: the space does not have memory v2 on, or the server is at its bound of prefetches. The read answers the same either way.