joanfabregat/jina-embed repository overview

⁠Multilingual Embedding API

A FastAPI service that provides text embedding capabilities using the jinaai/jina-embeddings-v3⁠ model.

⁠Overview

This API allows you to generate vector embeddings for text documents using the jinaai/jina-embeddings-v3 model. The service supports different embedding tasks (query and indexing) and is designed to process batches of texts efficiently.

⁠Features

• Generate embeddings for multiple texts in a single request
• Support for both query and index embedding tasks
• Efficient batch processing
• Token counting endpoint
• Service information endpoint

⁠Requirements

• Python 3.13+
• FastAPI
• fastembed
• Pydantic
• uvicorn

⁠Environment Variables

The service can be configured using the following environment variables:

• PORT: The port to run the server on (default: 8000)
• VERSION: The version of the service (default: "unknown")
• BUILD_ID: The build identifier (default: "unknown")
• COMMIT_SHA: The commit SHA (default: "unknown")

⁠Usage

⁠Starting the Server

The recommended way to run this service is using Docker.

docker run -p 8000:8000 joanfabregat/jina-embed:latest

Documentation for the API can be found at /docs or /redoc when running the server.

⁠API Endpoints

⁠GET /info

Returns information about the service.

{
  "model_name": "jinaai/jina-embeddings-v3",
  "version": "1.0.0",
  "build_id": "12345",
  "commit_sha": "abc123"
}

⁠POST /embed

Generates embeddings for a list of texts.

Request body:

{
  "texts": [
    "This is a sample text",
    "Another sample text"
  ],
  "task": "query",
  "batch_size": 4
}

Response:

[
  [
    0.123,
    0.456,
    ...
  ],
  [
    0.789,
    0.012,
    ...
  ]
]

Parameters:

• texts: List of texts to embed (required)
• task: Embedding task, either "query" or "index" (default: "query")
• batch_size: Batch size for processing texts (default: 4)

⁠License

This project is licensed under the MIT License - see the license notice in the code for details.

⁠Credits

Developed by Joan Fabrégat, [email protected]⁠