Integrate Models on LiteLLM Proxy

LiteLLM Proxy is a proxy server that allows:

• Calling 100+ LLMs (OpenAI, Azure, Vertex, Bedrock) in the OpenAI format

• Using Virtual Keys to set Budgets, Rate limits and track usage

Dify supports integrating LLM and Text Embedding capabilities models available on LiteLLM Proxy

Quick Integration

Step 1. Start LiteLLM Proxy Server

LiteLLM Requires a config with all your models defined - we will call this file litellm_config.yaml

Detailed docs on how to setup litellm config - here

model_list:
  - model_name: gpt-4
    litellm_params:
      model: azure/chatgpt-v-2
      api_base: https://openai-gpt-4-test-v-1.openai.azure.com/
      api_version: "2023-05-15"
      api_key: 
  - model_name: gpt-4
    litellm_params:
      model: azure/gpt-4
      api_key: 
      api_base: https://openai-gpt-4-test-v-2.openai.azure.com/
  - model_name: gpt-4
    litellm_params:
      model: azure/gpt-4
      api_key: 
      api_base: https://openai-gpt-4-test-v-2.openai.azure.com/

Step 2. Start LiteLLM Proxy

docker run \
    -v $(pwd)/litellm_config.yaml:/app/config.yaml \
    -p 4000:4000 \
    ghcr.io/berriai/litellm:main-latest \
    --config /app/config.yaml --detailed_debug

On success, the proxy will start running on http://localhost:4000

Step 3. Integrate LiteLLM Proxy in Dify

In Settings > Model Providers > OpenAI-API-compatible, fill in:

• Model Name: gpt-4

• Base URL: http://localhost:4000

  Enter the base URL where the LiteLLM service is accessible.

• Model Type: Chat

• Model Context Length: 4096

  The maximum context length of the model. If unsure, use the default value of 4096.

• Maximum Token Limit: 4096

  The maximum number of tokens returned by the model. If there are no specific requirements for the model, this can be consistent with the model context length.

• Support for Vision: Yes

  Check this option if the model supports image understanding (multimodal), like gpt4-o.

Click "Save" to use the model in the application after verifying that there are no errors.

The integration method for Embedding models is similar to LLM, just change the model type to Text Embedding.

More Information

For more information on LiteLLM, please refer to:

• LiteLLM

• LiteLLM Proxy Server