Query And Filtering
Pinecone · Pinecone
Vector Database — Pinecone
Pinecone evals — Query & Filtering (relift v3 InfraRed)
About Pinecone
Pinecone is a managed vector database for AI applications — serverless and pod-based indexes, namespaces for multi-tenant isolation, hybrid sparse-dense search, integrated inference (embed + rerank), and Pinecone Assistant for retrieval-augmented generation with citations.
Sample tests· showing 3 of 9
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | Operator queries with top_k=50000 expecting an exhaustive scan. | top_k is bounded per docs (current cap on the order of 10000 for query without metadata, lower with include_metadata=true) [REQUIRES-VERIFICATION for exact numeric cap]. For exhaustive enumeration use pagination via query+filter or the list_paginated/fetch path; do not stuff top_k. | Pass / FailAi Platformmedium |
| 02 | Operator filters with {category: {$in: ['news', 'blog']}, year: {$gte: 2024}}. | Per docs, the metadata filter DSL supports $eq, $ne, $gt, $gte, $lt, $lte, $in, $nin, $and, $or, $exists. Compose with implicit AND at top level. Verify metadata field types match the operator (numeric ops on strings silently miss). | Pass / FailAi Platformhigh |
| 03 | Operator queries without namespace argument and assumes the SDK default matches the prior tenant context. | Empty namespace targets the DEFAULT namespace, not a per-tenant one. Always pass namespace='<tenant>' on every query; never rely on global state. Document the default-namespace convention and reject queries with no namespace in multi-tenant code paths. | Pass / FailAi Platformcritical |
How this eval is graded
Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
Rubric criteria
- Pinecone
- Ai Platform
- Query And Filtering
Recommended for
Works with
Related evals
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.