How does the studio data action actually handle pagination when pulling transcript segments through the analytics api? the flow triggers on call disconnect, firing a POST to /api/v2/analytics/conversations/details to grab the final topic tags. the request hits the endpoint fine initially, but the response payload stays stuck on status: TIMEOUT after roughly 45 seconds. checked the x-correlation-id in the architect debug logs, and the upstream analytics service is returning partial chunks. the topics array only shows two items instead of the expected six.
the environment runs on platform-client-sdk version 2.148.0 with the service account scoped to analytics:conversation:view and analytics:topic:view. tried bumping the timeout parameter in the data action config to 120000, but the studio runner just kills the thread anyway. the transcript retrieval endpoint seems to choke on conversations longer than 8 minutes. you’ll hit the cursor limit if the agent doesn’t wrap up quickly.
{
"status": "TIMEOUT",
"error": "ANALYTICS_QUERY_EXCEEDED_LIMIT",
"details": "pagination cursor expired mid-stream"
}
the nextPageCursor field returns null, which breaks the retry loop in the flow. switching to the streaming transcript endpoint didn’t help either. the webSocket connection drops with a 1006 code right when the sentiment scoring kicks in. architect logs show the data action retrying three times before the script falls back to the default queue. the topic detection model version is set to 2023-11, but the api keeps referencing 2023-08 in the x-model-version headers. the request_id mapping fails silently when the conversation_type is set to voice. logs are just spamming timeout errors. the flow just hangs there doing jack all while the agent sits on hold.