> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pinecone.io/llms.txt
> Use this file to discover all available pages before exploring further.

# voyage-multimodal-3

> Voyage AI's voyage-multimodal-3 embeds interleaved text and images, like PDFs and slides, in 1024 dimensions. Generate vectors with Voyage AI, then upsert them.

[Back to all models](https://docs.pinecone.io/models/overview.md)

![Voyage AI](https://cdn.sanity.io/images/vr8gru94/production/f2bce9e78e611fb34bf845928f6ede4cc377a8a5-100x100.png)

# voyage-multimodal-3

Voyage AI

Bring your own vectors

Specifications

<dl />

## When to use it

Rich multimodal embedding model that can vectorize interleaved text and content-rich images, such as screenshots of PDFs, slides, tables, figures, and more. See [blog post](https://blog.voyageai.com/2024/11/12/voyage-multimodal-3/) for details. Visit the [Voyage documentation](https://docs.voyageai.com/docs/multimodal-embeddings) for an overview of all Voyage embedding models and rerankers.

## Quickstart

### Installation

```python theme={null}
!pip install -qU voyageai pinecone
```

### Create index

```python theme={null}
from pinecone import Pinecone, ServerlessSpec

pc = Pinecone(api_key="API_KEY")

# Create Index
index_name = "voyage-multimodal-3"

if not pc.has_index(index_name):
    pc.create_index(
        name=index_name,
        dimension=1024,
        metric="cosine",
        spec=ServerlessSpec(
            cloud="aws",
            region="us-east-1"
        )
    )

index = pc.Index(index_name)
```

### Embed & upsert

```python theme={null}
from typing import Union

# Embed data
data = [
    {"id": "vec1", "data": {"content": [{"type": "image_url", "image_url": "https://example.com/image.jpg"}, {"type": "text", "text": "Frontier intelligence at 2x the speed"}]}},
    {"id": "vec2", "data": {"content": [{"type": "image_url", "image_url": "https://example.com/page1.jpg"}, {"type": "image_url", "image_url": "https://example.com/page2.jpg"}]}},
    {"id": "vec3", "data": {"content": [{"type": "image_base64", "image_base64": "data:image/jpeg;base64,..."}]}}
]

# You can also use lists of texts and PIL Images, e.g.:
# "data": ["This is a banana", PIL.Image.open("banana.jpg")] ]

import voyageai

vo = voyageai.Client(api_key=VOYAGE_API_KEY)

model_id = "voyage-multimodal-3"

def embed(docs: Union[list[dict[str, str]], list[list[Union[str, Image]]]], input_type: str) -> list[list[float]]:
    embeddings = vo.multimodal_embed(
		    docs,
		    model=model_id,
		    input_type=input_type
		).embeddings
    return embeddings

# Use "document" input type for documents
embeddings = embed([d["data"] for d in data], input_type="document")

vectors = []
for d, e in zip(data, embeddings):
    vectors.append({
        "id": d['id'],
        "values": e,
        "metadata": {'inputs': d['inputs']}
    })

index.upsert(
    vectors=vectors,
    namespace="ns1"
)
```

### Query

```python theme={null}
query = ["Strong LLMs in 2024"]

# Use "query" input type for queries
x = embed([query], input_type="query")

results = index.query(
    namespace="ns1",
    vector=x[0],
    top_k=3,
    include_values=False,
    include_metadata=True
)

print(results)
```

## Good to know

To generate vectors, use Voyage AI's Python client with a Voyage AI API key. [Register](https://dash.voyageai.com/) to get one.
