Get started

pnpm add @anvia/core @anvia/chroma chromadb

Embed documents before writing them; the adapter deliberately disables Chroma's collection embedding function.

import { embedDocuments } from '@anvia/core/embeddings';
import { retrieveDocuments } from '@anvia/core/vector-store';
import { ChromaVectorClient } from '@anvia/chroma';
const { documents } = await embedDocuments({
    model: embeddings,
    documents: sourceDocuments,
    id: (document) => document.id,
    content: (document) => document.text
});
const storeClient = new ChromaVectorClient({
    client
});
const store = storeClient.vectorStore({
    collectionName: 'support_docs',
    dimensions: 1536
});
await store.validate();
await store.upsert({
    documents: documents
});
const results = await retrieveDocuments({
    store,
    model: embeddings,
    query: 'How do I reset my password?',
    topK: 5
});

client may be omitted and path may configure the default Chroma client. Use ensure() when this process may create a missing collection; use validate() when infrastructure owns provisioning.

Next

• Collections and indexing
• Search and filters
• Production