Qdrant Search Tool PhoenixInstall

Summary

Perform a semantic search on a specified Qdrant vector collection using an input embedding and return relevant document contents with optional metadata, while tracing the search process with Arize Phoenix.

Inputs

• Tool Name (required) — Name of the tool to identify and invoke it within an agent.
• Tool Description (required) — Description of what the tool does to help the agent decide when to use it.
• Collection Name (required) — The name of the Qdrant collection to search for relevant documents.
• Options — Configuration options including content payload key, whether to include metadata in results, metadata payload key, minimum similarity score threshold, Qdrant search filter expressed as JSON, and the number of top results to return (Top K).
• Phoenix Options — Tracing options for Arize Phoenix including session ID to group traces and user ID for attribution.
• Input Embedding (required) — Embedding vector from an upstream embedding node used as the semantic search query.

Output shape

a single string containing concatenated, numbered search results with their similarity scores, content, and optionally metadata, or an error message string.

Results are filtered by the provided score threshold and optional Qdrant filter JSON. If no documents are found, the output is a message indicating no relevant documents. The operation employs Arize Phoenix tracing to monitor the embedding creation and search phases. The output is delivered synchronously as a single text response suitable for subsequent AI tool consumption.