OpenLM

Drop-in OpenAI-compatible library that can call LLMs from other providers (e.g., HuggingFace, Cohere, and more).

1c1
< import openai
---
> import openlm as openai

completion = openai.Completion.create(
    model=["bloom-560m", "cohere.ai/command"], 
    prompt=["Hello world!", "A second prompt!"]
)
print(completion)

Features

• Takes in the same parameters as OpenAI's Completion API and returns a similarly structured response.
• Call models from HuggingFace's inference endpoint API, Cohere.ai, OpenAI, or your custom implementation.
• Complete multiple prompts on multiple models in the same request.
• Very small footprint: OpenLM calls the inference APIs directly rather than using multiple SDKs.

Installation

pip install openlm

Examples

• Import as OpenAI
• Set up API keys via environment variables or pass a dict
• Add a custom model or provider
• Complete multiple prompts on multiple models

OpenLM currently supports the Completion endpoint, but over time will support more standardized endpoints that make sense.

Example with Response

import sys
from pathlib import Path

sys.path.append(str(Path(__file__).resolve().parent.parent))

import openlm 
import json

completion = openlm.Completion.create(
    model=["ada", "huggingface.co/gpt2", "cohere.ai/command"],
    prompt=["The quick brown fox", "Who jumped over the lazy dog?"],
    max_tokens=15
)
print(json.dumps(completion, indent=4))

{
    "id": "504cc502-dc27-43e7-bcc3-b62e178c247e",
    "object": "text_completion",
    "created": 1683583267,
    "choices": [
        {
            "id": "c0487ba2-935d-4dec-b191-f7eff962f117",
            "model_idx": 0,
            "model_name": "openai.com/ada",
            "index": 0,
            "created": 1683583233,
            "text": " jumps into the much bigger brown bush.\" \"Alright, people like you can",
            "usage": {
                "prompt_tokens": 4,
                "completion_tokens": 15,
                "total_tokens": 19
            },
            "extra": {
                "id": "cmpl-7E3CCSpJHXfx5yB0TaJU9ON7rNYPT"
            }
        },
        {
            "id": "bab92d11-5ba6-4da2-acca-1f3398a78c3e",
            "model_idx": 0,
            "model_name": "openai.com/ada",
            "index": 1,
            "created": 1683583233,
            "text": "\n\nIt turns out that saying one's name \"Joe\" is the",
            "usage": {
                "prompt_tokens": 7,
                "completion_tokens": 15,
                "total_tokens": 22
            },
            "extra": {
                "id": "cmpl-7E3CDBbqFy92I2ZbSGoDT5ickAiPD"
            }
        },
        {
            "id": "be870636-9d9e-4f74-b8bd-d04766072a7b",
            "model_idx": 1,
            "model_name": "huggingface.co/gpt2",
            "index": 0,
            "created": 1683583234,
            "text": "The quick brown foxes, and the short, snuggly fox-scented, soft foxes we have in our household\u2026 all come in two distinct flavours: yellow and orange; and red and white. This mixture is often confused with"
        },
        {
            "id": "c1abf535-54a9-4b72-8681-d3b4a601da88",
            "model_idx": 1,
            "model_name": "huggingface.co/gpt2",
            "index": 1,
            "created": 1683583266,
            "text": "Who jumped over the lazy dog? He probably got it, but there's only so much you do when you lose one.\n\nBut I will say for a moment that there's no way this guy might have picked a fight with Donald Trump."
        },
        {
            "id": "08e8c351-236a-4497-98f3-488cdc0b6b6a",
            "model_idx": 2,
            "model_name": "cohere.ai/command",
            "index": 0,
            "created": 1683583267,
            "text": "\njumps over the lazy dog.",
            "extra": {
                "request_id": "0bbb28c0-eb3d-4614-b4d9-1eca88c361ca",
                "generation_id": "5288dd6f-3ecf-475b-b909-0b226be6a193"
            }
        },
        {
            "id": "49ce51e6-9a18-4093-957f-54a1557c8829",
            "model_idx": 2,
            "model_name": "cohere.ai/command",
            "index": 1,
            "created": 1683583267,
            "text": "\nThe quick brown fox.",
            "extra": {
                "request_id": "ab5d5e03-22a1-42cd-85b2-9b9704c79304",
                "generation_id": "60493966-abf6-483c-9c47-2ea5c5eeb855"
            }
        }
    ],
    "usage": {
        "prompt_tokens": 11,
        "completion_tokens": 30,
        "total_tokens": 41
    }
}

Other Languages

r2d4/llm.ts is a TypeScript library that has a similar API that sits on top of multiple language models.

Roadmap

• Streaming API
• Embeddings API

Contributing

Contributions are welcome! Please open an issue or submit a PR.

License

MIT