Integration Guide

Dify + AIPower

Use AIPower as an OpenAI-compatible provider in Dify, then run GPT, Claude, Gemini, DeepSeek, Qwen, GLM, Kimi, and Doubao through one account.

Get API key Try demo

Customer support app

Route common questions to DeepSeek/Qwen, escalate complex cases to Claude/GPT/Gemini.

Internal knowledge base

Use one Dify workspace with shared model access for China and global teams.

E-commerce pre-sales bot

Answer product, shipping, refund, and policy questions with low-cost starter models.

Team prototype

Give product and ops teams one billing layer instead of separate provider accounts.

Setup in Dify

Open your Dify workspace and go to Settings → Model Providers.
Select OpenAI or an OpenAI-compatible provider configuration.
Paste your AIPower key and set the API base URL:
API Base URL: https://api.aipower.me/v1
API Key: your-aipower-key
Start with deepseek/deepseek-chat, qwen/qwen-plus, or zhipu/glm-4-flash for trial-friendly workflows.
After your first top-up, add GPT, Claude, Gemini, Kimi, MiniMax, and premium routing models to the same Dify workspace.

Dify's model provider docs describe provider setup and OpenAI custom base URL configuration. AIPower supplies the OpenAI-compatible endpoint and billing layer.

Recommended model routing for Dify apps

Dify job	AIPower model	Why
Support / FAQ bot	deepseek/deepseek-chat	Good default for customer support, routing, and bilingual answers.
Chinese customer service	qwen/qwen-plus	Strong Chinese answers and practical cost control.
Workflow classification	zhipu/glm-4-flash	Very low-cost intent routing, tags, and short structured outputs.
Complex troubleshooting	anthropic/claude-sonnet	Use after top-up for long replies, coding, and deeper analysis.
Long documents	google/gemini-2.5-flash	Use for large docs, policies, and long-context workflows.

Why this works well for Dify teams

One account for many models. Keep Dify on one OpenAI-compatible endpoint while still testing China and global model families.
Payment flexibility. Use card, WeChat Pay, or Alipay instead of opening separate provider billing accounts.
Better support-bot economics. Use low-cost models for intent routing and FAQs, then reserve premium models for harder cases.
Team/BYOK path. When your Dify app moves from prototype to team usage, add shared budgets, invoices, logs, and provider-key routing.

Power your next Dify app with AIPower

10 trial calls. One OpenAI-compatible endpoint. China + global models.

Get API key