Dify

Dify: configure a model connection in an automation workflow.

Published instructions · App version: unverified

Chat, ping, invoke and content-generation examples may send real billable requests. Check the model, balance, rate and minimum before running them. GET /v1/models checks authentication and the catalogue only. Set up environment variables safely

Base URL to paste

OpenAI SDKs generally use a URL ending in /v1; Anthropic SDKs use the root URL. Avoid adding /v1 twice. Protocol compatibility does not guarantee every provider feature.

https://thapi.cloud/v1

Prerequisites

Dify cloud or self-hosted
Model Provider admin permissions

Recommended model: claude-fable-5

Quick way — command / config

Settings → Model Provider → OpenAI-API-compatible → Add
  Model Name        : claude-fable-5
  API Key           : sk-thapi-...
  API endpoint URL  : https://thapi.cloud/v1
  Model type        : LLM
  Context size      : 1000000
  Max tokens        : 128000
  ☑ Vision support (if needed)

Where to place file / fill Settings

Interfaceany

Settings → Model Provider

Use OpenAI-API-compatible (rather than OpenAI official) to change the endpoint.

Verify connection

Select a ThAPI model in Dify Studio and try Run

Troubleshooting

Symptom	Reason	How to handle
Invalid URL	The endpoint has an extra /chat/completions suffix.	Chỉ dán https://thapi.cloud/v1 — Dify tự nối path.

n8n · Flowise