wealthsimple / llm-gateway

Gateway for secure & reliable communications with OpenAI and other LLM providers

LLM Gateway

🤔 What is this?

llm-gateway is a gateway for third party LLM providers such as OpenAI, Cohere, etc. It tracks data sent and received from these providers in a postgres database and runs PII scrubbing heuristics prior to sending.

Per OpenAI's non-API consumer products data usage policy, they "may use content such as prompts, responses, uploaded images, and generated images to improve our services" to improve products like ChatGPT and DALL-E.

Use llm-gateway to interact with OpenAI in a safe manner. The gateway also recreates the ChatGPT frontend using OpenAI's /ChatCompletion endpoint to keep all communication within the API.

📦 Supported Models

Provider	Model
OpenAI	GPT 3.5 Turbo
OpenAI	GPT 3.5 Turbo 16k
OpenAI	GPT 4
AI21 Labs	Jurassic-2 Ultra
AI21 Labs	Jurassic-2 Mid
Amazon	Titan Text Lite
Amazon	Titan Text Express
Amazon	Titan Text Embeddings
Anthropic	Claude 2.1
Anthropic	Claude 2.0
Anthropic	Claude 1.3
Anthropic	Claude Instant
Cohere	Command
Cohere	Command Light
Cohere	Embed - English
Cohere	Embed - Multilingual
Meta	Llama-2-13b-chat
Meta	Llama-2-70b-chat

⚒️ Usage

The provider's API key needs to be saved as an environment variable (see setup further down). If you are communicating with OpenAI, set OPENAI_API_KEY.

For step-by-step setup instructions with Cohere, OpenAI, and AWS Bedrock, click here.

API Usage

[OpenAI] Example cURL to /completion endpoint:

curl -X 'POST' \
  'http://<host>/api/openai/completion' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "temperature": 0,
  "prompt": "Tell me what is the meaning of life",
  "max_tokens": 50,
  "model": "text-davinci-003"
}'

[OpenAI] When using the /chat_completion endpoint, formulate as conversation between user and assistant.

curl -X 'POST' \
  'http://<host>/api/openai/chat_completion' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "messages": [
    {"role": "assistant", "content": "You are an intelligent assistant."},
    {"role": "user", "content": "create a healthy recipe"}
  ],
  "model": "gpt-3.5-turbo",
  "temperature": 0
}'

Python Usage

from llm_gateway.providers.openai import OpenAIWrapper

wrapper = OpenAIWrapper()
wrapper.send_openai_request(
    "Completion",
    "create",
    max_tokens=100,
    prompt="What is the meaning of life?",
    temperature=0,
    model="text-davinci-003",
)

🚀 Quick Start for Developers

This project uses Poetry, Pyenv for dependency and environment management. Check out the official installation documentation for Poetry and Pyenv to get started. For front-end portion, this project use npm and yarn for dependency management. The most up-to-date node version required for this project is declared in .node-version.

Backend Dependencies

If using Docker, steps 1-3 are optional. We recommend installing pre-commit hooks to speed up the development cycle.

1. Install Poetry and Pyenv
2. Install pyenv install 3.11.3
3. Install project requirements

brew install gitleaks
poetry install
poetry run pre-commit install

4. Run cp .envrc.example .envrc and update with API secrets

🐳 Docker Development Loop (backend & frontend)

To run in Docker:

# spin up docker-compose
make up

# open frontend in browser
make browse

# open FastAPI Swagger API
make browse-api

# delete docker-compose setup
make down