Responses and transcripts

Positional rows, timestamp units, aligned transcript arrays and detail JSON.

Rows, units and detail JSON

List responses are JSON arrays of positional rows. Each cell follows the request’s select order. [] is a successful query with no matching readable rows.
Top-level request start and end use Unix microseconds. List started_at and started_at filter values use Unix seconds. Durations and transcript offsets are seconds.
Combine turn_data.texts, turn_data.is_customers and turn_data.starts by index. true identifies customer speech; false identifies agent speech.
/session JSON uses camelCase keys and enum names; byte fields are base64. Default-valued fields may be omitted, and wordData is omitted. Without x-proto-as-json: true, the endpoint returns protobuf.
/get-audio-links returns one session object plus aligned audio and audio_links arrays. The URLs expire after one hour. Retention can leave both arrays empty.
Unset proto values can appear as 0, false, empty strings or empty arrays. They do not prove an event was measured. Detection flags are signals, not verified outcomes.

Map rows in code

Map list rows, transcript turns and grouped metrics

// Example only: select = ["session_id", "started_at", "customer_data.name"]
// Response: [["CALL_ID", 1791849600, "Example contact"]]
const calls = rows.map(row => Object.fromEntries(
  select.map((column, index) => [column, row[index]])
));
const startedAt = calls[0] ? new Date(calls[0].started_at * 1000) : null;

// Transcript columns are aligned arrays:
const turns = texts.map((text, index) => ({
  text,
  speaker: isCustomers[index] ? "customer" : "agent",
  startSeconds: starts[index]
}));

// /session with x-proto-as-json: true uses different JSON spelling:
// { "sessionId": "CALL_ID", "customerData": { "name": "Example contact" },
//   "turnData": { "texts": ["Hello"], "isCustomers": [true], "starts": [0] } }
// Default-valued proto fields may be omitted. wordData is omitted.

// Grouped metrics: selects = [{"cnf": []}, {"column_name": "duration"}]
// Example response: [["Interested", 12, 840]]
// Read as: disposition, count, summed duration in seconds.

Common fields

Common fields and their exact paths
What you need	Column	How to use it
Call ID	`session_id`	Copy the exact string into the x-session-id header for /session.
Call time	`started_at`	Unix seconds in list rows. Convert with new Date(value * 1000).
Customer phone/address	`customer_address`	For phone calls, the customer address; phone is not a column name.
Rep ID	`resource_id`	Match to users[].user_id from /visible.
Contact name	`customer_data.name`	Use this instead of prospect_name or name.
Contact title	`customer_data.title`	Use this instead of title.
Company	`customer_data.company_name`	Use this instead of company.
Company website	`customer_data.company_url`	May be empty when not enriched.
Contact email / custom data	`customer_data.custom_fields`	No standard email column. Custom fields are JSON text; inspect available keys rather than assuming email exists.
Transcript text	`turn_data.texts`	Array of utterances, not one transcript string.
Transcript speaker	`turn_data.is_customers`	Aligned with texts: true = customer, false = agent.
Transcript timing	`turn_data.starts`	Aligned with texts; seconds from the start of the session.
Call duration	`duration`	Seconds. dial_metric.live_duration measures the live portion.
Saved outcome	`user_log.disposition`	Use /bootstrap for your exact configured option strings.

The field reference lists every selectable path with its type and values.