Pinecone
Claude Plugin - Personal knowledge base for Claude Code — patterns and integrations across projects
npx -y skills add phucbm/skills --skill pineconeAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Set up Pinecone as the vector DB in a RAG pipeline — index config, batch upsert, semantic query, metadata filters. Use when the user is adding Pinecone or debugging vector search.
SKILL.md
2.4 KB, as published. Nobody here has run it
Pinecone
Managed vector database. Used as the vector store in a RAG pipeline — see the rag skill for the full pipeline pattern.
Install
pnpm add @pinecone-database/pinecone
Env
PINECONE_API_KEY=
PINECONE_INDEX_NAME=chatbot-knowledge
Client
import { Pinecone } from "@pinecone-database/pinecone";
const pinecone = new Pinecone({ apiKey: process.env.PINECONE_API_KEY! });
const index = pinecone.index(process.env.PINECONE_INDEX_NAME ?? "chatbot-knowledge");
Batch upsert
const BATCH_SIZE = 100;
export async function upsertVectors(vectors: VectorRecord[], namespace: string) {
const ns = index.namespace(namespace);
for (let i = 0; i < vectors.length; i += BATCH_SIZE) {
await ns.upsert(vectors.slice(i, i + BATCH_SIZE));
}
}
Query
export async function queryVectors(
embedding: number[],
topK: number,
filter?: MetadataFilter,
) {
const result = await index.namespace("default").query({
vector: embedding,
topK,
includeMetadata: true,
filter,
});
return result.matches;
}
Metadata filter pattern
Store token arrays on each vector for flexible keyword matching:
// On ingest
metadata: {
tokens: ["com", "tam", "com tam"],
district_tokens: ["quan 1", "quan binh thanh"],
}
// At query time
const filter = { tokens: { $in: tokenize(userQuery) } };
If no tokens match, return {} to fall through to pure semantic search (no filter applied).
Supported operators: $in, $eq, $and, $or.
Increase topK when filters are active — filtering reduces recall before ranking.
e.g. RAG_TOP_K=5 unfiltered, RAG_TOP_K_FILTERED=10-30 filtered.
References
- food-tour-sg
lib/rag/pinecone.ts: https://github.com/phucbm/food-tour-sg/blob/main/lib/rag/pinecone.ts - food-tour-sg
lib/rag/filters.ts: https://github.com/phucbm/food-tour-sg/blob/main/lib/rag/filters.ts - food-tour-sg
lib/rag/normalize.ts: https://github.com/phucbm/food-tour-sg/blob/main/lib/rag/normalize.ts - Pinecone Node.js client docs: https://docs.pinecone.io/reference/node-sdk