Custom Providers

If the AI service you use isn't in MoryFlow's preset list, you can add it yourself.

When to Use This

• Using third-party services compatible with OpenAI protocol
• Connecting to your company's internal AI service
• Using other aggregation platforms
• Trying new AI services

OpenAI Compatible Protocol

Most AI services are compatible with OpenAI's API protocol, meaning you just need:

1. API Endpoint (Base URL)
2. API Key
3. Model Name

And you can use it in MoryFlow.

Adding a Provider

How to Add

1. Open Settings → Model Management
2. Scroll to "Custom Providers"
3. Click "Add Provider"

What to Fill In

Common Provider Configurations

Azure OpenAI

Base URL: https://{your-resource}.openai.azure.com/openai/deployments/{deployment-name}
API Key: Your Azure API Key

Groq

Base URL: https://api.groq.com/openai/v1
API Key: Your Groq API Key
Models: llama-3.1-70b-versatile, mixtral-8x7b-32768

Together AI

Base URL: https://api.together.xyz/v1
API Key: Your Together API Key
Models: meta-llama/Llama-3-70b-chat-hf

Perplexity

Base URL: https://api.perplexity.ai
API Key: Your Perplexity API Key
Models: llama-3.1-sonar-large-128k-online

Local LM Studio

Base URL: http://localhost:1234/v1
API Key: Leave empty or any value
Models: Name of model loaded in LM Studio

Local vLLM

Base URL: http://localhost:8000/v1
API Key: Leave empty
Models: Your deployed model name

Adding Custom Models

After configuring a provider, add available models:

1. Click "Add Model" in provider settings
2. Fill in model info:
   • Model ID: Name used in API calls
   • Display Name: Name shown in interface
   • Context Window: Maximum tokens supported
   • Capability Tags: Mark what features the model supports

Test It

After configuration, test it:

1. Select the new model in chat panel
2. Send a simple message
3. Check if it responds normally

Troubleshooting

Connection Failed

1. Check Base URL is correct (usually ends with /v1)
2. Confirm API Key is valid
3. Check if your network can reach the service

Model Not Working

1. Confirm model ID spelling is correct
2. Confirm the model is available in your account
3. Check if provider has regional restrictions

Response Format Wrong

Some providers may not be fully OpenAI compatible. You might need to:

• Check the provider's documentation
• Adjust request parameters
• Contact the provider

Security Notes

• API Keys are stored locally, not uploaded
• When using self-hosted services, ensure encrypted communication (HTTPS)
• Rotate API Keys periodically