1.6 Configure AI Models

AI Copilot is one of the core features of Pop. It provides intelligence for modules such as document Q&A, knowledge base retrieval, workflow nodes, and application outputs. This chapter will help you add and configure AI models.

---

🤖 1. Which Models Does Pop Support?

Pop supports all models compatible with the OpenAI API, including:

Provider	Example Models	Features
OpenAI	gpt-4o, gpt-4o-mini	Stable, versatile, industry standard
DeepSeek	deepseek-chat, deepseek-reasoner	Strong Chinese ability, cost-effective
Moonshot (Kimi)	moonshot-v1	Excellent long-text Chinese performance
OneAPI / Custom Gateways	Any custom model	Private deployment
Local Model Services	LM Studio / Ollama	Offline + strong privacy

💡 In enterprise environments, you can use OneAPI to centrally manage model permissions and billing.

---

🔧 2. How to Add a Model?

Navigate to:
System Settings → AI Models → Add Model

You will see the following fields:

Field	Description
Name	Any custom name, e.g., “DeepSeek Main Model”
Base URL	API base address, e.g., https://api.deepseek.com/v1
API Key	Your model API key
Model Name	For example: deepseek-chat
Max Tokens	Controls output length
Timeout	Request timeout (optional)
Set as Default	Determines the global default model for Copilot

---

📌 3. Configuration Examples (Common Models)

OpenAI Example

Base URL: https://api.openai.com/v1
Model Name: gpt-4o
API Key: sk-xxxxx

DeepSeek Example

Base URL: https://api.deepseek.com
Model Name: deepseek-chat
API Key: sk-xxxxx

OneAPI Example

Base URL: https://api.xxx.com/v1
Model Name: Any model name

---

🧪 4. Test Model Connection

After filling in the fields, click “Test Connection”.

• If successful → you will see a success message or model list
• If failed → check the following:

Possible Cause	What to Check
Incorrect BaseURL	Does it contain /v1? Is it HTTPS?
Network Issues	Proxy, VPN, corporate network limits
Invalid API Key	Expired or incorrectly entered?
Unsupported Model	Use “Test Connection” to check supported list

⚠️ If you access external APIs via a proxy, configure HTTP/HTTPS proxies in system settings.

---

🎯 5. Choosing Different Models for Different Modules

Pop allows you to assign different models to different modules:

Module	Recommended Model	Reason
Document Q&A	gpt-4o / deepseek-chat	Strong comprehension
Knowledge Base	deepseek / moonshot	Concise and accurate summary generation
Workflow AI Node	gpt-4o-mini / custom gateway	Cost-effective and stable
App Output	Choose based on scenario	Balance quality & cost

---

🛠 6. Advanced Settings (Optional)

① Custom Headers

Useful for enterprise gateways or authentication systems:

{
  "X-User-ID": "123",
  "X-Dept": "AI"
}

② Limit Maximum Concurrent Requests

Prevents your model service from being overloaded.

③ Override System Prompt

Ideal for building specialized AI assistants:

• Legal advisor
• Medical assistant
• Game design assistant
• Customer service reply generator

---

🔚 Summary

Once your model configuration is complete, you can immediately use:

• AI-powered document interpretation
• Knowledge Base Q&A
• Copilot conversations
• Workflow AI nodes
• Application AI outputs

Next, we will learn how to import your first document.