Endpoint reference
Inspect persisted voice sessions.
Use this endpoint to review transcripts, tool calls, model and voice context, QA state, tags, summaries, and launch-governance evidence.
Endpoint contract
| Property | Value |
|---|
| Method | GET |
| Path | /api/v1/voice-sessions |
| Authentication | Authorization: Bearer ${STATESET_VOICE_API_KEY} |
| Primary use | Inspect persisted conversation state, transcript artifacts, function calls, tags, and QA metadata. |
Query parameters
| Parameter | Description |
|---|
limit | Maximum session records to return. |
offset | Offset for paginated review and export jobs. |
status | Filter active, completed, failed, or reviewed session states where available. |
stream_id | Use when reconciling a realtime media stream to a persisted session. |
agent_key | Use for prompt, tool, and rollout analysis by agent. |
Example request
curl -sS "https://voice.stateset.app/api/v1/voice-sessions?limit=25&offset=0" \
-H "Authorization: Bearer ${STATESET_VOICE_API_KEY}"
Response fields
| Field | Use |
|---|
stream_id | Join realtime media stream events to the persisted session. |
call_sid | Join the session to Twilio call state and call logs. |
transcript | Review what the caller and agent said for QA and prompt iteration. |
function_calls | Inspect tools used by the agent, arguments, results, and failures. |
summary | Display concise operational context in CRM, ticketing, or callback queues. |
tags | Drive routing, review workflows, analytics, and launch-governance filters. |
QA workflow
- Start from call logs or webhook events, then open the matching voice session.
- Review transcript turns, model output, tool calls, and escalation decisions.
- Record evaluation notes and tags for launch governance and prompt regression tracking.
- Feed recurring failure patterns into the agent contract, tool boundaries, and production readiness checklist.
Data governance
- Limit access to transcript and tool-call data to teams that need operational review.
- Use tenant-scoped API keys and separate admin credentials from runtime readers.
- Define retention, redaction, and export practices before routing regulated traffic.
- Use tags and evaluations to support audit trails without overexposing raw conversation data.