Bring your own vectors

Create an index with a dense-vector field, upsert embeddings you already have, and run semantic search.

If you already generate embeddings, store them directly. Create an index with a dense-vector field, upsert your vectors, and rank by similarity. Pinecone does no embedding on your behalf here.

Prerequisites

A Pinecone account and API key (get one).
Python 3.10+.
The Pinecone Python SDK: pip install --upgrade pinecone.

Upsert your vectors and search

Set your API key
Set your API key as an environment variable so the SDK can authenticate:
Bash
```
export PINECONE_API_KEY="YOUR_API_KEY"
```

Create an index with a dense-vector field

Set dimension to match your embedding model's output, and pick a distance metric (cosine, dotproduct, or euclidean). Only the vector field goes in the schema; other fields like text are stored on the documents (non-schema fields are stored as metadata, capped at 40 KB per document).

Python

import os, time
from pinecone import Pinecone, SchemaBuilder

pc = Pinecone(api_key=os.environ["PINECONE_API_KEY"])

schema = (
    SchemaBuilder()
      .add_dense_vector_field(name="embedding", dimension=1024, metric="cosine")
      .build()
)
if not pc.indexes.exists(name="byo-vectors"):
    pc.indexes.create(name="byo-vectors", schema=schema)

while not pc.indexes.describe(name="byo-vectors").status.ready:
    time.sleep(2)

index = pc.Index(name="byo-vectors")

Upsert your vectors

Each document carries its embedding (a list of floats matching the schema's dimension) plus any fields you want to store. Replace the truncated vectors below with your real embeddings.

Python

index.documents.upsert(
    namespace="__default__",
    documents=[
        # each `embedding` is a full list of 1024 floats (shown truncated)
        {"_id": "1", "embedding": [0.12, 0.04, ...], "text": "Refund requests must be submitted within 30 days."},
        {"_id": "2", "embedding": [0.08, 0.21, ...], "text": "Enterprise support responds within 4 hours."},
    ],
)

time.sleep(5)  # documents are indexed asynchronously, so wait a moment

Search by vector similarity

Embed your query with the same model you used for the documents, then rank by the dense-vector field (embedding).

Python

query_embedding = [0.10, 0.05, ...]  # embed your query text with the same model

resp = index.documents.search(
    namespace="__default__",
    top_k=3,
    score_by=[{"type": "dense_vector", "fields": ["embedding"], "values": query_embedding}],
    include_fields=["*"],
)

for m in resp.matches:
    print(m._id, m._score, getattr(m, "text", ""))

Next steps

Match keywords

Add a full_text_search field to your schema for keyword search, or combine it with your vectors for hybrid search.

Data modeling

How to design a schema for your workload