search
Find relevant content across namespaces using natural language queries or vector similarity. Filter results by metadata or keywords for precise discovery.
Instructions
Search for data in a namespace using semantic search or vector similarity. This tool provides powerful search capabilities across your namespaces, supporting both text-based semantic search and vector-based similarity search. For text search, you can use natural language queries to find relevant documents based on meaning rather than just keywords. For vector search, you can find similar content by comparing vector embeddings. The tool supports advanced features like result filtering, similarity thresholds, metadata filters, keyword filters, and kiosk mode for production environments. This is ideal for building intelligent search interfaces, recommendation systems, or content discovery features.
Filtering Capabilities:
Metadata Filters: Use #key:value format (e.g., #category:tech, #priority:high)
Keyword Filters: Use #keyword format (e.g., #important, #urgent)
Filters only apply to text search and metadata must be manually uploaded with documents
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| namespaces | Yes | Namespaces to search in. Provide an array of namespace names where you want to search for content. You can search across multiple namespaces simultaneously. All namespaces must be accessible with your API key. | |
| query | Yes | Search query. For text search: provide a natural language query string (e.g., 'tell me about the company?', 'how to configure authentication?'). For vector search: provide an array of numbers representing a vector embedding (e.g., [0.1, 0.2, 0.3, ..., 0.768]). The query type will be automatically detected based on the input format. DO NOT USE QUOTES IN THE QUERY FOR VECTOR SEARCH. Filtering: For text search, you can include filters in your query: - Metadata filters: #category:tech #priority:high - Keyword filters: #important #urgent - Combine both: 'serverless benefits #category:tech #important' | |
| query_type | No | Type of query to perform. 'text' for semantic search using natural language queries. 'vector' for similarity search using vector embeddings. If not specified, the type will be automatically detected based on the query format (string for text, array for vector). | |
| top_k | No | Number of top results to return. Controls how many search results are returned, with higher values providing more comprehensive results. Default is 10. Use lower values (3-5) for focused results, higher values (10-20) for broader exploration. | |
| threshold | No | Similarity threshold for results. A value between 0 and 1 that filters results based on similarity score. Higher values (0.7-0.9) return only highly similar results, lower values (0.3-0.5) return more comprehensive results. Required when kiosk_mode is true. | |
| kiosk_mode | No | Kiosk mode for restricted search. When true, search is restricted to specific namespaces with threshold filtering, providing more controlled results suitable for production environments. When false, search across all specified namespaces without strict filtering. |