Search Conversations

conversation-search

Semantically search customer conversations using natural language. Powered by vector embeddings — finds conversations by meaning, not just keyword matches.

Parameters

Name	Type	Required	Description	Default
query	string	Yes	Natural language search query (3–500 chars)	—
category	string	No	Filter by insight category: blocker, post_purchase, product	—
tag	string	No	Filter by tag (e.g., price, shipping, returns, discount)	—
threshold	number	No	Minimum similarity score (0–1)	0.25
limit	number	No	Maximum results (1–50)	20

Example Prompt

Find conversations where customers were worried about shipping time

Example Response

{
  "query": "customers worried about shipping time",
  "filters": { "category": null, "tag": null },
  "results": [
    {
      "conversationId": "conv-456",
      "similarity": 0.82,
      "textContent": "Customer asked about 2-day shipping options and whether order would arrive before the weekend",
      "category": "blocker",
      "tag": "shipping",
      "customer": "Jane Doe"
    }
  ],
  "totalFound": 15
}

Notes

• This tool uses semantic search — it understands meaning, not just keywords. “customers worried about delivery” will match conversations about shipping delays
• Use category and tag filters to narrow results to specific types of conversations
• For structured filtering by date, status, or workflow, use list_conversations instead
• Use get_conversation_details to read the full message thread for any result