orign 0.1.6

A Python client for Orign

orign-py

A Python client for Orign

Installation

pip install orign

Install the Orign CLI

curl -fsSL -H "Cache-Control: no-cache" https://storage.googleapis.com/orign/releases/install.sh | bash

Login to Orign

$ orign login

Usage

Get a list of available models

$ orign get models

Chat

Define which model we would like to use

from orign import ChatModel

model = ChatModel(model="allenai/Molmo-7B-D-0924", provider="vllm")

Open a socket connection to the model

await model.connect()

Chat with the model

async for response in model.chat(msg="What's in this image?", image="https://tinyurl.com/2fz6ms35"):
    print(response)

Stream tokens from the model

async for response in model.chat(msg="What is the capital of France?", stream_tokens=True):
    print(response)

Send a thread of messages to the model

async for response in model.chat(prompt=[
    {"role": "user", "content": "What is the capital of France?"},
    {"role": "assistant", "content": "Paris"},
    {"role": "user", "content": "When was it built?"}
]):
    print(response)

Send a batch of threads to the model

async for response in model.chat(batch=[
    [{"role": "user", "content": "What is the capital of France?"}, {"role": "assistant", "content": "Paris"}, {"role": "user", "content": "When was it built?"}],
    [{"role": "user", "content": "What is the capital of Spain?"}, {"role": "assistant", "content": "Madrid"}, {"role": "user", "content": "When was it built?"}]
]):
    print(response)

Embeddings

Define which model we would like to use

from orign import EmbeddingModel

model = EmbeddingModel(provider="sentence-tf", model="clip-ViT-B-32")

Embed a text

await model.embed(text="What is the capital of France?")

Embed an image

await model.embed(image="https://example.com/image.jpg")

Embed text and image

await model.embed(text="What is the capital of France?", image="https://example.com/image.jpg")

OCR

Define which model we would like to use

from orign import OCRModel

model = OCRModel(provider="easyocr")

Detect text in an image

await model.detect(image="https://example.com/image.jpg")