This guide is for users who want to run Keep with locally hosted LLM models.
If you encounter any issues, please talk to us at our (Slack
community)[https://slack.keephq.dev].
Overview
This guide will help you set up Keep with LiteLLM, a versatile tool that supports over 100 LLM providers. LiteLLM acts as a proxy that adheres to OpenAI standards, allowing seamless integration with Keep. By following this guide, you can easily configure Keep to work with various LLM providers using LiteLLM.Motivation
Incorporating LiteLLM with Keep allows organizations to run local models in on-premises and air-gapped environments. This setup is particularly beneficial for leveraging AIOps capabilities while ensuring that sensitive data does not leave the premises. By using LiteLLM as a proxy, you can seamlessly integrate with Keep and access a wide range of LLM providers without compromising data security. This approach is ideal for organizations that prioritize data privacy and need to comply with strict regulatory requirements.Prerequisites
Running LiteLLM locally
- Ensure you have Python and pip installed on your system.
- Install LiteLLM by running the following command:
- Start LiteLLM with your desired model. For example, to use the HuggingFace model:
http://0.0.0.0:4000
.
Running LiteLLM with Docker
To run LiteLLM using Docker, you can use the following command:Configuration
Env var | Purpose | Required | Default Value | Valid options |
---|---|---|---|---|
OPEN_AI_ORGANIZATION_ID | Organization ID for OpenAI/LiteLLM services | Yes | None | Valid organization ID string |
OPEN_AI_API_KEY | API key for OpenAI/LiteLLM services | Yes | None | Valid API key string |
OPENAI_BASE_URL | Base URL for the LiteLLM proxy | Yes | None | Valid URL (e.g., “http://localhost:4000”) |
These environment variables should be set on both Keep frontend and
backend.