Response API - File Search

The file search functionality using the OpenAI Responses API in the Compass platform enables users to search the files and content within their authorized projects.

A File Search tool that lets AI models search through your uploaded documents to answer questions with contextual information from uploaded and processed files.

Core File Search Parameters

Examples

1. Basic File Search Request

# Environment variables
export BASE_URL="https://<compass-url>/v1"
export API_KEY="api-key"

# Basic file search within project scope
curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key: $API_KEY" \
  -d '{
    "model": "gpt-4o-mini",
    "input": "What are the key findings in our Q4 financial reports?",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj123_financial", "vs_proj123_reports"]
      }
    ],
    "metadata": {
      "user_id": "user_456",
      "project_id": "proj_123",
      "request_type": "file_search"
    }
  }'

2. Advanced File Search with Ranking Options

curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key: $API_KEY" \
  -d '{
    "model": "gpt-4o",
    "input": "Find all mentions of machine learning algorithms in our research papers",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj123_research"],
        "max_num_results": 10,
        "ranking_options": {
          "ranker": "default_2024_08_21",
          "score_threshold": 0.7
        }
      }
    ],
    "temperature": 0.3,
    "metadata": {
      "user_id": "user_789",
      "project_id": "proj_123",
      "search_category": "research"
    }
  }'

3. Multi-Turn Conversation with File Search

# First request
curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key: $API_KEY" \
  -d '{
    "model": "gpt-4o-mini",
    "input": "Summarize our product roadmap documents",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj456_roadmap"]
      }
    ],
    "metadata": {
      "user_id": "user_101",
      "project_id": "proj_456"
    }
  }'

# Continue conversation using previous_response_id
curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key $API_KEY" \
  -d '{
    "model": "gpt-4o-mini",
    "input": "What are the timeline dependencies mentioned?",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj456_roadmap"]
      }
    ],
    "previous_response_id": "resp_abc123def456",
    "metadata": {
      "user_id": "user_101",
      "project_id": "proj_456"
    }
  }'

4. Multiple File Search Tools in Single Request

You can include multiple file_search tools with different vector stores and configurations in the same request for comprehensive cross-document analysis.

# Multiple file search tools with different configurations
curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key $API_KEY" \
  -d '{
    "model": "gpt-4o",
    "input": "How do our financial projections align with our product roadmap and marketing strategy?",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj123_financial"],
        "max_num_results": 10,
        "ranking_options": {
          "score_threshold": 0.8
        }
      },
      {
        "type": "file_search", 
        "vector_store_ids": ["vs_proj123_roadmap"],
        "max_num_results": 15,
        "ranking_options": {
          "score_threshold": 0.7
        }
      },
      {
        "type": "file_search",
        "vector_store_ids": ["vs_proj123_marketing"],
        "max_num_results": 12,
        "ranking_options": {
          "score_threshold": 0.75
        }
      }
    ],
    "metadata": {
      "user_id": "user_456",
      "project_id": "proj_123",
      "request_type": "multi_source_search",
      "search_categories": "financial,roadmap,marketing"
    }
  }'

# Alternative: Single tool with multiple vector stores
curl -X POST $BASE_URL/responses \
  -H "Content-Type: application/json" \
  -H "api-key $API_KEY" \
  -d '{
    "model": "gpt-4o",
    "input": "Analyze our complete business strategy across all departments",
    "tools": [
      {
        "type": "file_search",
        "vector_store_ids": [
          "vs_proj123_financial",
          "vs_proj123_roadmap", 
          "vs_proj123_marketing",
          "vs_proj123_hr_docs"
        ],
        "max_num_results": 25
      }
    ],
    "metadata": {
      "user_id": "user_789",
      "project_id": "proj_123", 
      "request_type": "comprehensive_search"
    }
  }'