Skip to main content

Using Dify

Dify is an open-source platform for creating AI-powered applications with a visual interface. This guide shows how to integrate Azerion Intelligence with Dify.

What is Dify?​

Dify provides a no-code/low-code environment for building AI applications with visual workflow builders and multi-modal support.

Prerequisites​

Integration Steps​

1. Add Azerion Intelligence as a Model Provider​

  1. Access Model Settings:

    • Log into your Dify workspace
    • Navigate to Settings → Model Provider
    • Click Add Model Provider

  2. Configure OpenAI-Compatible Provider:

    • Select OpenAI-API-compatible as the provider type
    • Set the following configuration:
    Provider Name: Azerion Intelligence
    Base URL: https://api.azerion.ai/v1
    API Key: YOUR_AZERION_API_KEY

  3. Test Connection:

    • Click Test to verify the connection
    • If successful, click Save to add the provider

2. Configure Available Models​

  1. Add Model Configuration:

    • In the Model Provider settings, click Add Model
    • Configure the model:
    Model Name: meta.llama3-3-70b-instruct-v1:0
    Model Type: Text Generation
    Context Length: 32768
    Max Tokens: 4096

3. Create an Application​

  1. Create New App:

    • Go to Studio and click Create App
    • Choose your application type (Chatbot, Agent, or Workflow)

  2. Select Model:

    • Choose Azerion Intelligence as your provider
    • Select meta.llama3-3-70b-instruct-v1:0 as the model
    • Configure temperature and max tokens as needed

Basic Example​

Simple chatbot configuration:

{
  "model": {
    "provider": "azerion-intelligence",
    "name": "meta.llama3-3-70b-instruct-v1:0",
    "mode": "chat",
    "completion_params": {
      "temperature": 0.7,
      "max_tokens": 1024
    }
  }
}

Troubleshooting​

Connection Issues​

Problem: "Failed to connect to model provider"

Solutions:

  • Verify your API key is correct and active
  • Check that the base URL is set to https://api.azerion.ai/v1
  • Ensure your network allows outbound HTTPS requests

Model Not Available​

Problem: "Model not found" error

Solutions:

  • Verify the model name is correct: meta.llama3-3-70b-instruct-v1:0
  • Check if the model is available in your Azerion Intelligence plan

Rate Limiting​

Problem: "Rate limit exceeded" errors

Solutions:

  • Implement request queuing in your Dify workflows
  • Consider upgrading your Azerion Intelligence plan
  • Add retry logic with exponential backoff