Skip to main content
POST

Introduction

Convert text to vector embeddings using Gemini native interface. The model is specified via URL path (e.g., gemini-embedding-001), suitable for scenarios requiring Google embedding models or alignment with Gemini API. Complements the OpenAI format in Text Embedding: this document covers the Gemini native path; the same capability is also available via POST /v1/embeddings.

Authentication

Bearer Token, e.g., Bearer sk-xxxxxxxxxx

Path Parameters

string
required
Embedding model name, e.g., gemini-embedding-001.

Request Parameters

object
required
Content to embed. Must contain a parts array, each item as { "text": "text content" }.
integer
Output vector dimensions (only supported by some models, e.g., gemini-embedding-001, text-embedding-004).
string
Task type, e.g., RETRIEVAL_DOCUMENT, RETRIEVAL_QUERY (optional).

Code Examples

Response Example

Batch Interface (batchEmbedContents)

For batch embedding, use: POST /v1/models/{model}:batchEmbedContents. The request body is a requests array, each item with the same structure as single requests (including content.parts), and do not include the model field in each item.

Supported Models

Notes

  • content.parts is required, at least one text must be non-empty
  • Model is specified via URL path, do not include model field in request body
  • Usage information is in metadata.usage in the response (prompt_tokens, total_tokens)
If your application already uses the OpenAI SDK, consider using the /v1/embeddings compatible interface to minimize code changes.