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

# Voyage AI

> Create text and contextualized chunk embeddings and rerank documents with Voyage AI models for retrieval and RAG pipelines

## At a glance

* **Availability**: Experimental ([how to enable](/cloud/integrations)).
* **Auth**: API key.
* **Connection**: The key from `VOYAGE_API_KEY` is sent as the `Bearer` prefixed `Authorization` header.
* **Docs**: [https://docs.voyageai.com/docs/api-key-and-installation](https://docs.voyageai.com/docs/api-key-and-installation)

## Credentials

Set these per environment. See [Connect an integration](/cloud/integrations).

| Variable         | Required | Description                                                                                                                    |
| ---------------- | -------- | ------------------------------------------------------------------------------------------------------------------------------ |
| `VOYAGE_API_KEY` | Yes      | Voyage AI API key from the Voyage dashboard (starts with pa-) [Docs](https://docs.voyageai.com/docs/api-key-and-installation). |

## Setup

1. **Create a Voyage AI account**: Sign up at [https://dashboard.voyageai.com](https://dashboard.voyageai.com). New accounts include free trial tokens for every model, which is enough for evaluation without billing.
2. **Create an API key**: In the dashboard, open the API Keys section and create a new key. Copy it immediately.
3. **Store the key**: Add VOYAGE\_API\_KEY=.... to your .env file. Requests authenticate with an Authorization: Bearer header.
4. **Verify access**: Run the Create Embeddings tool with model voyage-3.5 and a single short input string. A 401 means the key is wrong or revoked.

* Use the same embedding model (and input\_type convention) for indexing and querying - vectors from different models are not comparable
* Contextualized chunk embeddings (voyage-context-3) expect each inner list to be one document's chunks in order; regular embeddings treat every array item independently
* Usage beyond the free trial tokens is billed per million tokens and requires adding a payment method in the dashboard

Provider API reference: [https://docs.voyageai.com/reference/embeddings-api](https://docs.voyageai.com/reference/embeddings-api)

## Tools

| Tool                             | Access | Description                                                                                                                          |
| -------------------------------- | ------ | ------------------------------------------------------------------------------------------------------------------------------------ |
| Create Embeddings                | Read   | Create embedding vectors for one or more texts - use input\_type to optimize for retrieval queries vs documents                      |
| Rerank Documents                 | Read   | Rank candidate documents by relevance to a query with a Voyage reranker - use as a precision pass after vector search                |
| Create Contextualized Embeddings | Read   | Embed document chunks with document-level context (voyage-context-3) so each chunk vector also captures the surrounding document     |
| Create Multimodal Embeddings     | Read   | Embed interleaved text and images into one vector space (voyage-multimodal models) for visual RAG over screenshots, slides, and PDFs |

## Example prompts

* Create Voyage AI embeddings for the texts I provide and report the model, dimensions, and token usage.
* Use Voyage AI rerank-2.5 to order a list of documents I provide by relevance to my query and show the relevance scores.