Pinecone Developer MCP

Official

cascading-search

Search across multiple indexes for similar records, deduplicate results, and rerank them for relevance using advanced models. Ideal for precise, multi-source data retrieval.

Instructions

Search across multiple indexes for records that are similar to the query text, deduplicate and rerank the results.

Input Schema

NameRequiredDescriptionDefault
indexesYesThe indexes to search across. Records in each index should share a common schema.
queryYesA query to search for records.
rerankYesSpecifies how the results should be reranked. Use a "query" with a "topK" that returns more results than you need; then use "rerank" to select the most relevant "topN" results.

Input Schema (JSON Schema)

{ "$schema": "http://json-schema.org/draft-07/schema#", "additionalProperties": false, "properties": { "indexes": { "description": "The indexes to search across. Records in each index should share a common schema.", "items": { "additionalProperties": false, "properties": { "name": { "description": "An index to search.", "type": "string" }, "namespace": { "description": "A namespace to search.", "type": "string" } }, "required": [ "name", "namespace" ], "type": "object" }, "type": "array" }, "query": { "additionalProperties": false, "description": "A query to search for records.", "properties": { "filter": { "description": "A filter can be used to narrow down results. Use the syntax of\n MongoDB's query and projection operators: $eq, $ne, $gt, $gte, $lt,\n $lte, $in, $nin, $exists, $and, $or. Make sure the records in the index\n contain the fields that you are filtering on." }, "inputs": { "additionalProperties": false, "properties": { "text": { "description": "The text to search for.", "type": "string" } }, "required": [ "text" ], "type": "object" }, "topK": { "description": "The number of results to return.", "type": "number" } }, "required": [ "topK", "inputs" ], "type": "object" }, "rerank": { "additionalProperties": false, "description": "Specifies how the results should be reranked. Use a \"query\" with a \"topK\"\n that returns more results than you need; then use \"rerank\" to select the\n most relevant \"topN\" results.", "properties": { "model": { "description": "Choose a reranking model:\n- \"cohere-rerank-3.5\" is Cohere's leading reranking model, balancing performance\nand latency for a wide range of enterprise search applications.\n- \"bge-reranker-v2-m3\" is a high-performance, multilingual reranking model that\nworks well on messy data and short queries expected to return medium-length\npassages of text (1-2 paragraphs).\n- \"pinecone-rerank-v0\" is a state of the art reranking model that out-performs\ncompetitors on widely accepted benchmarks. It can handle chunks up to 512 tokens\n(1-2 paragraphs).", "enum": [ "cohere-rerank-3.5", "bge-reranker-v2-m3", "pinecone-rerank-v0" ], "type": "string" }, "query": { "description": "An optional query to rerank documents against. If not specified, the\n same query will be used for both the initial search and the reranking.", "type": "string" }, "rankFields": { "description": "The fields to rerank on. This should include the field name specified\n in the index's \"fieldMap\". The \"bge-reranker-v2-m3\" and\n \"pinecone-rerank-v0\" models support only a single rerank field.\n \"cohere-rerank-3.5\" supports multiple rerank fields.", "items": { "type": "string" }, "type": "array" }, "topN": { "description": "The number of results to return after reranking. Must be less than or\n equal to the value of \"query.topK\".", "type": "number" } }, "required": [ "model", "rankFields" ], "type": "object" } }, "required": [ "indexes", "query", "rerank" ], "type": "object" }

You must be authenticated.

Other Tools from Pinecone Developer MCP

Related Tools

ID: 7lxm56a88w