remoteinference 0.0.8

Remote inference for language models

Remoteinference

Simple package to perform remote inference on language models of different providers.

Getting Started

Install the package

pip install remoteinference

To access an OpenAI model simply import the OpenAILLM and use the chat_completion endpoint to send your contents to the server endpoint. As a response you will receive a valid JSON containing the typicall OpenAI API conform response in a dictionary:

import os

from remoteinference.models import OpenAILLM
from remoteinference.util import user_prompt

model_type = 'gpt-4o-mini'
model = OpenAILLM(
    api_key=os.environ.get('OPEANI_API_KEY'),
    model=model_type
    )

response = model.chat_completion(
    prompt=[user_prompt('Who are you?')],
    temperature=0.5,
    max_tokens=50
)

print(response['choices'][0]['message']['content'])

If you have a LLM running on a remote server using llama.cpp you can initalize the model by running:

from remoteinference.models import LlamaCPPLLM
from remoteinference.util import user_prompt

# initalize the model
model = LlamaCPPLLM(
    server_address='localhost',
    server_port=8080
    )

# run simple completion
response = model.chat_completion(
    prompt=[user_prompt('Who are you?')],
    temperature=0.5,
    max_tokens=50
)

print(response['choices'][0]['message']['content'])

Supported Models

OpenAI

Initialize an OpenAI model by calling:

from remoteinference.models import OpenAILLM

model = OpenAILLM(
    api_key='your_key',
    model='gpt-4o-mini'
)

To view a full list of available models for the OpenAI endpoint see OpenAI docs

TogetherAI

Initialize an OpenAI model by calling:

from remoteinference.models import TogetherAILLM

model = TogetherAILLM(
    api_key='your_key',
    model='meta-llama/Llama-3-8b-hf'
)

To view a full list of available models for the OpenAI endpoint see TogetherAI docs

LlamaCPP

This package also provides functionality to query a self-hosted language model via llama.cpp

To initalize a model which is hosted locally just do:

from remoteinference.models import LlamaCPPLLM

model = LlamaCPPLLM(
    server_address='localhost',
    server_port=8080
)

To see the full specifications of the llama.cpp webserver see server docs.