Setup Guide for TextEmbed

This guide provides detailed instructions for setting up the TextEmbed server using both PyPI and Docker. Follow the steps below to get started!

📝 Prerequisites

Python Version: Ensure you have Python 3.10 or higher installed.
Dependencies: Install all required dependencies for the TextEmbed server.

⚙️ Installation via PyPI

Install the required dependencies:
```
pip install -U textembed
```

Start the TextEmbed server with your desired models:

python3 -m textembed.server --models <Model1>,<Model2> --port <Port>

For more information and additional options, run:
```
python3 -m textembed.server --help
```

🔧 Available Server Options:

--models: Comma-separated list of Huggingface models to be used, e.g., <Model1>,<Model2>.
--served_model_names: Comma-separated list of names under which the models will be served.
--host: The host address where the application will run.
--port: The port number where the application will run.
--workers: Number of worker processes for batch processing.
--batch_size: The batch size for processing requests.
--embedding_dtype: The data type for the embeddings (binary, float16, or float32).
--api_key: Your API key for authentication (Keep it secure and do not share it with others).

🐳 Running with Docker (Recommended)

Run TextEmbed using Docker for a more streamlined deployment. The Docker image is available on Docker Hub.

Pull the Docker image:

docker pull kevaldekivadiya/textembed:latest

Run the Docker container:

docker run -p 8000:8000 kevaldekivadiya/textembed:latest --models <Model1>,<Model2> --port <Port>

For more information and additional options, run:

docker run kevaldekivadiya/textembed:latest --help

This command displays the help message for the TextEmbed server, detailing the available options and usage instructions.

🌐 Accessing the API

Once the server is running, you can access the API documentation via Swagger UI by navigating to http://localhost:8000/docs in your web browser.

🖼️ Image Embedding Example

TextEmbed now supports generating embeddings for images, such as using the SentenceTransformer CLIP model (sentence-transformers/clip-ViT-B-32).

📷 Steps to Generate Image Embeddings

Convert Image to Base64 String:

import base64

def image_to_base64(image_path: str) -> str:
    with open(image_path, "rb") as image_file:
        encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
    return encoded_string

image_path = '<IMAGE_PATH>'
base64_string = image_to_base64(image_path)

Make a POST Request to the TextEmbed Server:

import requests

resp = requests.post(url="http://0.0.0.0:8000/v1/image_embedding", json={
  "input": [
    base64_string
  ],
  "model": "sentence-transformers/clip-ViT-B-32",
  "user": "string"
})

print(resp.json())

📩 Example Request and Response

Request:

{
  "input": [
    "<Base64EncodedImageString>"
  ],
  "model": "sentence-transformers/clip-ViT-B-32",
  "user": "string"
}

Response:

{
  "embeddings": [
    [0.1, 0.2, ..., 0.3]
  ],
  "model": "sentence-transformers/clip-ViT-B-32",
  "user": "string"
}