When using an index with integrated inference, embedding and reranking operations are tied to index operations and do not require extra steps. This allows working with an index that accepts source text and converts it to vectors automatically using an embedding model hosted by Pinecone.
For full-text search over document fields without generating embeddings, see Working with Documents.
For more information, see Upsert and search with integrated inference.
Integrated inference requires a serverless index configured for a specific embedding model. Use the pc.indexes.createForModel method to create an index that will automatically handle embeddings:
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
await pc.indexes.createForModel({
name: 'integrated-index',
cloud: 'aws',
region: 'us-east-1',
embed: {
model: 'multilingual-e5-large',
fieldMap: { text: 'chunk_text' },
},
waitUntilReady: true,
});
The fieldMap specifies which field in your records contains the text to be embedded. In this example, records must have a chunk_text field.
Once you have an index configured for a specific embedding model, use the upsertRecords operation to convert your source data to embeddings and upsert them into a namespace.
Each record must contain:
id (or _id) fieldfieldMap specified when creating the index (e.g., chunk_text)import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
const index = pc.index({ name: 'integrated-index', namespace: 'namespace1' });
const records = [
{
id: 'rec1',
chunk_text:
"Apple's first product, the Apple I, was released in 1976 and was hand-built by co-founder Steve Wozniak.",
category: 'product',
},
{
id: 'rec2',
chunk_text:
'Apples are a great source of dietary fiber, which supports digestion and helps maintain a healthy gut.',
category: 'nutrition',
},
{
id: 'rec3',
chunk_text:
'Apples originated in Central Asia and have been cultivated for thousands of years, with over 7,500 varieties available today.',
category: 'cultivation',
},
{
id: 'rec4',
chunk_text:
'In 2001, Apple released the iPod, which transformed the music industry by making portable music widely accessible.',
category: 'product',
},
];
await index.upsertRecords({ records });
Use the searchRecords method to convert a query to a vector embedding and then search your namespace for the most semantically similar records, along with their similarity scores.
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
const index = pc.index({ name: 'integrated-index', namespace: 'namespace1' });
// Search for 4 records most semantically relevant to the query
const response = await index.searchRecords({
query: {
topK: 4,
inputs: { text: 'Disease prevention' },
},
});
console.log(response);
// {
// matches: [
// {
// id: 'rec2',
// score: 0.847,
// record: {
// id: 'rec2',
// chunk_text: 'Apples are a great source of dietary fiber...',
// category: 'nutrition'
// }
// },
// // ... more matches
// ],
// usage: { readUnits: 5 }
// }
To rerank initial search results based on relevance to the query, add the rerank parameter, including the reranking model you want to use, the number of reranked results to return, and the fields to use for reranking.
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
const index = pc.index({ name: 'integrated-index', namespace: 'namespace1' });
const response = await index.searchRecords({
query: {
topK: 10,
inputs: { text: 'Disease prevention' },
},
rerank: {
model: 'bge-reranker-v2-m3',
topN: 3,
rankFields: ['chunk_text'],
},
fields: ['category', 'chunk_text'],
});
console.log(response);
// {
// matches: [
// {
// id: 'rec2',
// score: 0.952, // Reranked score
// record: {
// id: 'rec2',
// chunk_text: 'Apples are a great source of dietary fiber...',
// category: 'nutrition'
// }
// },
// // ... top 3 reranked results
// ],
// usage: {
// readUnits: 5,
// rerankUnits: 1
// }
// }
You can combine text search with metadata filtering:
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
const index = pc.index({ name: 'integrated-index', namespace: 'namespace1' });
const response = await index.searchRecords({
query: {
topK: 4,
inputs: { text: 'Apple products' },
filter: { category: { $eq: 'product' } },
},
});
By default, all fields are returned in search results. You can specify which fields to return:
import { Pinecone } from '@pinecone-database/pinecone';
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
const index = pc.index({ name: 'integrated-index', namespace: 'namespace1' });
const response = await index.searchRecords({
query: {
topK: 4,
inputs: { text: 'Apple products' },
},
fields: ['category', 'chunk_text'], // Only return these fields
});
For a list of available embedding models, see the model gallery. You can also use pc.inference.listModels() to see all available models programmatically.
Here's a complete workflow using integrated inference:
import { Pinecone } from '@pinecone-database/pinecone';
async function integratedInferenceExample() {
const pc = new Pinecone({ apiKey: 'YOUR_API_KEY' });
// 1. Create an index for a model
await pc.indexes.createForModel({
name: 'my-integrated-index',
cloud: 'aws',
region: 'us-east-1',
embed: {
model: 'multilingual-e5-large',
fieldMap: { text: 'content' },
},
waitUntilReady: true,
});
const index = pc.index({ name: 'my-integrated-index' });
// 2. Upsert records with text
await index.upsertRecords({
records: [
{
id: 'article1',
content: 'Machine learning is transforming healthcare...',
category: 'healthcare',
},
{
id: 'article2',
content: 'Artificial intelligence in finance...',
category: 'finance',
},
],
});
// 3. Search with text and reranking
const results = await index.searchRecords({
query: {
topK: 5,
inputs: { text: 'AI in medical applications' },
},
rerank: {
model: 'bge-reranker-v2-m3',
topN: 2,
rankFields: ['content'],
},
});
console.log(results);
}
integratedInferenceExample();
For standalone inference without index integration, see Inference API.