Calls and dispositions
Every call carries two dispositions:
dialer_disposition: the code the agent set on the dialer (SALE,NI,DNC,NA…).standard_dispositionis the same code normalised through the dialer’s mapping (seeGET /directory/disposition-codes).ai_disposition: the code the AI evaluator chose after listening to the call, from the same code set.disposition_matchistruewhen the two agree,falsewhen they differ andnullwhen either is missing.
A mismatch usually means the agent coded the call wrongly: a sale that was really “not interested”, or an answered call marked as an answering machine.
Search calls
Section titled “Search calls”GET /calls lists calls newest first. Give a window of at most 31 days with
ended_from/ended_to, then combine any filters. Repeat a list filter to match any of its
values: ?dialer_disposition=SALE&dialer_disposition=XFER.
| You want | Filters |
|---|---|
| Calls the AI dispositioned differently | disposition_match=false |
| Sales the AI did not hear as sales | dialer_disposition=SALE&disposition_match=false |
| Calls the AI judged “not interested” | ai_disposition=NI |
| One agent’s fatal compliance failures | employee_id=…&quality_band=FATAL |
| Low-scoring sales calls | is_sales_call=true&max_score=60 |
| Calls a reviewer should look at | needs_review=true |
| A team’s transferred calls | team_id=…&transfer_connected=true |
| Calls where the customer hung up | hangup_reason=CALLER |
| Answering-machine calls too | include_voice_machine=true |
Look up a lead
Section titled “Look up a lead”Lookups need no window:
GET /calls?dialer_lead_id=…: every call on a lead, for matching against your CRM.GET /calls?dialer_call_id=…: one dialer call by itsuniqueid.GET /calls?phone=…: any format, e.g.+1 (555) 010-2030. Needsleads:read.
GET /leads/details?dialer_lead_id=… returns what the agent captured on those calls (name,
contact, address, age), also under leads:read.
Drill into one call
Section titled “Drill into one call”| Endpoint | Returns |
|---|---|
GET /calls/{call_id} |
the call plus AI summary, coaching note, sentiment, the AI’s disposition rationale |
GET /calls/{call_id}/evaluation |
score, band and every rule result with transcript quotes as evidence |
GET /calls/{call_id}/review |
QA reviewers’ notes and overrides |
GET /calls/{call_id}/transcript |
speaker-labelled segments with timestamps |
GET /calls/{call_id}/recording |
the audio (supports Range) |
score and quality_band on a call already include any override by a QA reviewer. The
evaluation also shows ai_score and ai_quality_band, what the AI gave before review.
Counts instead of calls
Section titled “Counts instead of calls”To count rather than list:
GET /performance/dispositions: calls per dialer disposition, by agent, team, campaign, dialer or day.GET /performance/disposition-accuracy: a matrix of dialer disposition × AI disposition.mismatches_only=truekeeps the disagreements.GET /performance?group_by=agent&sort=ai_disposition_mismatch_rate: agents ranked by how often the AI disagrees with them.