pinecone

Manage vector indexes, upserts, and hybrid queries with the Pinecone API.

14|5|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/MLT-OSS/hermes-agent-go --skill pinecone-mlt-oss
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pinecone
Source: https://github.com/MLT-OSS/hermes-agent-go/tree/main/optional-skills/mlops/pinecone
Command: npx skills add https://github.com/MLT-OSS/hermes-agent-go --skill pinecone-mlt-oss

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pinecone-client, and includes references (resource) components.

What problem does it solve? Building production RAG, semantic search, or recommendation systems requires a scalable vector database, and this Skill provides the operational knowledge to create indexes, upsert embeddings, and query Pinecone without managing infrastructure. ## Core Features & Use Cases - Index Management: Create serverless or pod-based indexes, inspect stats, and delete indexes through the Pinecone Python client. - Vector Operations: Batch upsert embeddings with metadata, run similarity queries with metadata filters, and partition data using namespaces for multi-tenancy. - Hybrid Search: Combine dense and sparse vectors with an alpha weighting parameter for semantic plus keyword retrieval. - Framework Integration: Connect Pinecone to LangChain and LlamaIndex as a vector store or retriever. - Use Case: You are building a production RAG chatbot over company documentation. Use this Skill to create a serverless index, upsert document embeddings with category metadata, and query with filters to retrieve the top-k relevant chunks per user namespace. ## Quick Start Ask the agent to create a Pinecone serverless index and upsert your document embeddings, then run a filtered similarity query against it.

Frequently Asked Questions about pinecone

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a Pinecone index in Python?

Install pinecone-client, initialize the Pinecone client with your API key, and call create_index with a name, dimension matching your embedding model, metric such as cosine, and a ServerlessSpec specifying cloud and region.

How do I query Pinecone with metadata filters?

Pass a filter dictionary to the query method alongside your vector and top_k. Filters support equality, comparison operators like $gte and $lte, logical $and/$or, and $in for matching metadata fields such as category or price.

Pinecone vs Chroma vs FAISS for vector search?

Pinecone is a fully managed serverless service suited to production workloads needing auto-scaling and low latency. Chroma is self-hosted and open-source, while FAISS is an offline library for pure similarity search without a server.

Does Pinecone support hybrid search with sparse vectors?

Yes. Upsert vectors with both dense values and sparse_values containing token indices and weights, then query with a sparse_vector and an alpha parameter where 0 means sparse only, 1 means dense only, and 0.5 balances both.

How do I isolate data for multiple users in Pinecone?

Use namespaces to partition vectors by user or tenant. Upsert and query with a namespace parameter so each tenant's data stays logically separated within a single index, and inspect namespaces via describe_index_stats.

When should I not use Pinecone?

Avoid Pinecone when you need fully self-hosted infrastructure, offline operation, or want to avoid SaaS costs. In those cases use Chroma or Weaviate for self-hosting, or FAISS for local offline similarity search.