Vector_Store_Content_Retrieval_Guide

Purpose

Use this guide to retrieve file content semantically with GovTribe vector-store tools when metadata search is not enough. Use it when:

• users ask content-level questions about attached files.

• keyword matching over metadata is weak.

• you need semantic chunk retrieval from file text.

Workflow

1. Resolve input scope for vector ingestion.

   • Use only govtribe_type values allowed by the current Add_To_Vector_Store schema.

   • If a target entity type is unsupported, fall back to Search_* metadata tools or request a supported govtribe_id/govtribe_type.

2. Call Add_To_Vector_Store when no reusable store ID is available.

   • Capture and persist the returned govtribe_vector_store_id.

3. Reuse an existing govtribe_vector_store_id when one is already available for the same corpus.

4. Call Search_Vector_Store with a focused content question.

   • Keep max_num_results as small as needed to answer the request.

   • Use rewrite_query only when the user intent benefits from query rephrasing.

5. Treat returned chunks as semantic evidence, not full-document exact-match output.

Examples

Create and query a vector store sequence: Ingest one external file into a vector store: Step 1 Tool: Add_To_Vector_Store

{
  "items": [
    {
      "govtribe_type": "external_file_model",
      "govtribe_id": "<EXTERNAL_FILE_ID>"
    }
  ]
}

Search the resulting vector store for requirements: Step 2 Tool: Search_Vector_Store

{
  "query": "Summarize cyber incident response requirements and deadlines.",
  "govtribe_vector_store_id": "<VECTOR_STORE_ID>",
  "max_num_results": 10,
  "rewrite_query": true
}

Reuse an existing store with tighter result count: Tool: Search_Vector_Store

{
  "query": "List mandatory deliverables and submission format instructions.",
  "govtribe_vector_store_id": "<VECTOR_STORE_ID>",
  "max_num_results": 5,
  "rewrite_query": false
}