Intelligence Queries

After a transcript completes, ask follow-up questions to extract summaries, action items, sentiment, and more.

Submit a Question

POST https://app.resemble.ai/api/v2/speech-to-text/{uuid}/ask

$
curl --request POST 'https://app.resemble.ai/api/v2/speech-to-text/550e8400-e29b-41d4-a716-446655440000/ask' \
>
  -H 'Authorization: Bearer YOUR_API_TOKEN' \
>
  -H 'Content-Type: application/json' \
>
  --data '{"query": "What action items were mentioned?"}'

Response:

1
{
2
  "success": true,
3
  "item": {
4
    "uuid": "650e8400-e29b-41d4-a716-446655440001",
5
    "transcript_uuid": "550e8400-e29b-41d4-a716-446655440000",
6
    "query": "What action items were mentioned?",
7
    "status": "pending",
8
    "created_at": "2024-01-15T10:35:00Z"
9
  }
10
}

Check Question Status

GET https://app.resemble.ai/api/v2/speech-to-text/{uuid}/questions/{question_uuid}

Response:

1
{
2
  "success": true,
3
  "item": {
4
    "uuid": "650e8400-e29b-41d4-a716-446655440001",
5
    "transcript_uuid": "550e8400-e29b-41d4-a716-446655440000",
6
    "query": "What action items were mentioned?",
7
    "answer": "The following action items were identified...",
8
    "status": "completed",
9
    "created_at": "2024-01-15T10:35:00Z",
10
    "updated_at": "2024-01-15T10:35:30Z"
11
  }
12
}

List Questions

GET https://app.resemble.ai/api/v2/speech-to-text/{uuid}/questions

Supports page and per_page query parameters (default 25, max 50).

1
{
2
  "success": true,
3
  "items": [
4
    {
5
      "uuid": "650e8400-e29b-41d4-a716-446655440001",
6
      "query": "What action items were mentioned?",
7
      "answer": "The following action items were identified...",
8
      "status": "completed",
9
      "created_at": "2024-01-15T10:35:00Z"
10
    }
11
  ],
12
  "meta": {
13
    "page": 1,
14
    "per_page": 25,
15
    "total_pages": 1,
16
    "total_items": 3
17
  }
18
}

Common Questions

• Action items – “What action items were mentioned and who owns them?”
• Decisions – “What decisions were made?”
• Summaries – “Provide a 3-sentence executive summary.”
• Sentiment – “What is the overall sentiment of this conversation?”
• Entities – “List all people and companies referenced.”
• Data extraction – “Extract all statistics and numerical data mentioned.”

Workflow

1. Ensure the transcript status is completed.
2. Submit a question using the ask endpoint.
3. Store the returned question UUID.
4. Poll the status endpoint until completed.
5. Retrieve the answer (and cache it if you will reuse it).

Use exponential backoff when polling to avoid hitting rate limits.

Error Handling

Failures return standard error objects, for example:

1
{
2
  "success": false,
3
  "error": "Transcript must be completed before asking questions"
4
}

Other common errors include missing query values, invalid transcript UUIDs, or authorization failures.