This documentation describes the integration of MindsDB with LiteLLM, a framework that simplifies access to models of various providers.

​
Prerequisites

Before proceeding, ensure the following prerequisites are met:
  1. Install MindsDB locally via Docker or Docker Desktop.
  2. To use LiteLLM within MindsDB, install the required dependencies following this instruction.
  3. Obtain the API key of the model provider, if required.

​
Setup

Create an AI engine from the LiteLLM handler. 
CREATE ML_ENGINE litellm
FROM litellm;
Create a model using litellm as an engine. 
CREATE MODEL litellm_model
PREDICT target_column
USING
    engine = "litellm_engine",
    model = "gpt-4",
    base_url = "https://api.openai.com/v1",
    api_key = "sk-xxx",
    prompt_template = "answer questions in three bullet points: {question}";
The parameters include:
  • engine is the LiteLLM engine created based on the LiteLLM handler with the CREATE ML_ENGINE statement.
  • model is the one of the models supported by LiteLLM. See the complete list of the supported providers and models here.
    Note that you may need to provide the model name following this format: provider/model_name. For example, ollama/llama2 or openai/gpt-4o.
  • base_url is an optional parameter that stores the base URL for accessing models.
  • api_key stores the API key of the provider whose model is used.
  • prompt_template stores the instructions to the model.

​
Usage

Here is how to create and use models through LiteLLM in MindsDB. 
CREATE ML_ENGINE litellm
FROM litellm;

CREATE MODEL chat_model
PREDICT answer
USING
    engine = "litellm",
    model = "ollama/llama2:latest",
    base_url = "http://localhost:11434";

SELECT *
FROM chat_model
WHERE question = "what is ai?";