Aggregate Conversation Insights

Fields to group by (can be comma-separated or multiple parameters). Prefix a field with 'metadata.' (e.g. 'metadata.assistant_id') to group by the conversation's metadata instead of the insight result.

Common fields used for over-time charts:

• score — Group by the insight's score value (e.g. for Agent Instruction Following, User Satisfaction).
• metadata.assistant_id — Group by the assistant that handled the conversation.
• metadata.assistant_version_id — Group by the assistant version, useful for comparing performance across versions in the portal's 'Insights Over Time' chart.
• metadata.telnyx_conversation_channel — Group by conversation channel (phone_call, web_chat, etc.).

Fields to include in the result (can be comma-separated or multiple parameters). Supports the same 'metadata.<key>' prefix as group_by. Each returned row will contain the grouped field values plus a record_count indicating how many conversation insights match that combination.

Optional insight ID to filter conversation insights. Only insights matching this ID will be included in the aggregation.

Filter by creation datetime to scope the aggregation window. Supports range operators (e.g., created_at=gte.2025-01-01T00:00:00Z for the start of the range, created_at=lt.2025-01-02T00:00:00Z for the end). To build per-day time series (as the portal does for the 'Insights Over Time' chart), issue one request per day bounded by created_at=gte.<day_start> and created_at=lt.<next_day_start>.

Filter by assistant ID (e.g., metadata.assistant_id=eq.<assistant_id>). When provided, only conversation insights for the specified assistant are aggregated. Used by the portal to scope the 'Insights Over Time' chart to a single assistant.