vector-databases skill
✓ VerifiedIntegrates vector databases for embeddings, similarity search, index management, and metadata filtering across multiple providers.
- GitHub stars
- 47
- Files
- 1
- Last commit
- Nov 9, 2026
- License
- MIT
- Allowed tools
- 5
What does the vector-databases skill do?
vector-databases integrates vector databases for embeddings, similarity search, index management, and metadata filtering across multiple providers.
Summary generated from the skill's SKILL.md.
About the vector-databases skill
Managing embeddings and similarity searches across different database providers is what this skill enables. Developers can use it when working with specific vector technologies like Pinecone, Weaviate, Qdrant, ChromaDB, or the pgvector extension for PostgreSQL. It provides the necessary logic for index management, metadata filtering, and hybrid search operations.
An agent should load this skill when a user mentions vector databases, HNSW, or ANN. It produces code implementations for tasks such as upserting vectors, querying with metadata filters, and configuring distance metrics like cosine similarity. The skill relies on tools like Read, Write, Edit, Grep, and Glob to assist in the development process.
This integration focuses on the technical implementation of vector storage rather than high-level RAG architectures or LangChain workflows. It helps prevent common mistakes like embedding model mismatches or failing to use batching during upsert operations. Users can implement production-ready setups by configuring index types and managing metadata schemas correctly.
What an agent can do with it
- 1Implement Pinecone similarity searchesWrite code to query a Pinecone namespace using a vector and a metadata filter.
- 2Configure pgvector in PostgreSQLCreate a table with a vector column and an HNSW index for similarity searches.
- 3Generate OpenAI embeddingsUse the OpenAI API to convert text arrays into numerical embeddings.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the vector-databases skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the vector-databases skill in Claude Code
~/.claude/skills/vector-databases/SKILL.md
$ npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases ~/.claude/skills/vector-databasesInstall the vector-databases skill in Claude Code (project)
.claude/skills/vector-databases/SKILL.md
$ npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases .claude/skills/vector-databasesInstall the vector-databases skill in Codex
~/.agents/skills/vector-databases/SKILL.md
$ npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases ~/.agents/skills/vector-databasesInstall the vector-databases skill in Gemini CLI
~/.gemini/skills/vector-databases/SKILL.md
$ npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases ~/.gemini/skills/vector-databasesInstall the vector-databases skill in Claude.ai
npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases vector-databases && zip -r vector-databases.zip vector-databases # claude.ai → Customize → Skills → + → Create skill → Upload a skill → vector-databases.zip
Automated checks
Ran Nov 14, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work
- Source repository linked and reachableclaude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases is public and reachablegithub.com/claude-dev-suite/claude-dev-suite/tree/HEAD/skill ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit 2026-11-09github.com/claude-dev-suite/claude-dev-suite/commits ↗Pass
- No download-and-run or credential theft in scriptsNo scripts in the skill folderPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/claude-dev-suite/claude-dev-suite/blob/HEAD/skill ↗Pass
- Pre-approved tools are scopedPre-approved: Read, Grep, Glob, Write, Editgithub.com/claude-dev-suite/claude-dev-suite/blob/HEAD/skill ↗Pass
- SKILL.md follows the Agent Skills formatValid name and description; says when to use it; 160 linesgithub.com/claude-dev-suite/claude-dev-suite/blob/HEAD/skill ↗Pass
SKILL.md
View on GitHubVector Databases
Pinecone
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: process.env.PINECONE_API_KEY! });
const index = pc.index('my-index');
// Upsert
await index.namespace('docs').upsert([
{ id: 'doc-1', values: embedding, metadata: { source: 'manual', topic: 'auth' } },
]);
// Query with metadata filter
const results = await index.namespace('docs').query({
vector: queryEmbedding,
topK: 5,
filter: { topic: { $eq: 'auth' } },
includeMetadata: true,
});
ChromaDB (local/self-hosted)
import chromadb
client = chromadb.PersistentClient(path="./chroma_db")
collection = client.get_or_create_collection(
name="documents",
metadata={"hnsw:space": "cosine"},
)
# Add documents (auto-embeds with default model)
collection.add(
ids=["doc1", "doc2"],
documents=["Auth guide content", "API reference content"],
metadatas=[{"source": "manual"}, {"source": "api"}],
)
# Query
results = collection.query(query_texts=["how does login work?"], n_results=5)
pgvector (PostgreSQL extension)
CREATE EXTENSION IF NOT EXISTS vector;
CREATE TABLE documents (
id SERIAL PRIMARY KEY,
content TEXT NOT NULL,
embedding vector(1536),
metadata JSONB DEFAULT '{}'
);
CREATE INDEX ON documents USING hnsw (embedding vector_cosine_ops);
-- Similarity search
SELECT id, content, 1 - (embedding <=> $1::vector) AS similarity
FROM documents
WHERE metadata->>'source' = 'manual'
ORDER BY embedding <=> $1::vector
LIMIT 5;
Node.js with pgvector
import pgvector from 'pgvector';
await pgvector.registerTypes(pool);
await pool.query(
'INSERT INTO documents (content, embedding) VALUES ($1, $2)',
[text, pgvector.toSql(embedding)]
);
const { rows } = await pool.query(
'SELECT *, 1 - (embedding <=> $1) AS similarity FROM documents ORDER BY embedding <=> $1 LIMIT $2',
[pgvector.toSql(queryEmbedding), 5]
);
Qdrant
import { QdrantClient } from '@qdrant/js-client-rest';
const client = new QdrantClient({ url: 'http://localhost:6333' });
// Create collection
await client.createCollection('documents', {
vectors: { size: 1536, distance: 'Cosine' },
});
// Upsert
await client.upsert('documents', {
points: [{ id: 1, vector: embedding, payload: { source: 'manual' } }],
});
// Search with filter
const results = await client.search('documents', {
vector: queryEmbedding,
limit: 5,
filter: { must: [{ key: 'source', match: { value: 'manual' } }] },
});
Embedding Generation
import OpenAI from 'openai';
const openai = new OpenAI();
async function embed(texts: string[]): Promise<number[][]> {
const response = await openai.embeddings.create({
model: 'text-embedding-3-small', // 1536 dims, cheapest
input: texts,
});
return response.data.map((d) => d.embedding);
}
Anti-Patterns
| Anti-Pattern | Fix |
|---|---|
| No metadata filtering | Always store filterable metadata with vectors |
| Wrong distance metric | Match metric to embedding model (cosine for OpenAI) |
| Embedding model mismatch | Same model for indexing and querying |
| No batching on upsert | Batch upserts (100-1000 vectors per call) |
| Storing raw text in vector DB | Store text in primary DB, only IDs + vectors in vector DB |
Production Checklist
- Embedding model locked (changing requires full re-index)
- Batch upserts with error handling
- Metadata schema documented
- Index type configured (HNSW for most cases)
- Backup strategy for vector data
- Monitoring: query latency, index size, recall metrics
Files in this skill
1vector-databases skill questions
How do I install the vector-databases skill in Claude Code?
Run npx degit claude-dev-suite/claude-dev-suite/skills/ai-integration/vector-databases ~/.claude/skills/vector-databases. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the vector-databases skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
Which vector databases are supported?
The skill supports Pinecone, Weaviate, Qdrant, ChromaDB, and pgvector.
What should I avoid when using vector databases?
Avoid using this skill for LangChain integration, RAG architecture, or traditional database tasks.