TokenReply

Loading…

Reload

Dify with TokenReply

Use a Dify provider that exposes a custom API base URL, enter TokenReply /v1 plus a TokenReply key, and add the exact model ID as a custom model.

Last checked October 23, 2026 · Sources: Official client documentation and TokenReply route review

Check the installed Dify provider plugin version and confirm it exposes a custom base URL.
Set the base URL to TokenReply /v1 and store the API key in Dify credentials.
Add the exact catalog ID as a custom LLM model when it is not auto-discovered.
Run Dify credential validation and a minimal chat application before adding tools or knowledge retrieval.

Example

Provider type: OpenAI compatible
API base URL: https://api.tokenreply.com/v1
API key: sk-your-token-here
Model: YOUR_MODEL_ID

Dify accepts the provider credentials without falling back to the official OpenAI host.
The custom model produces a response in a minimal chat application.
Embedding, rerank, speech, and LLM model types are configured separately when used.

Dify provider fields and plugin availability vary by version and deployment.
An LLM configuration does not automatically configure embedding, rerank, speech, or image operations.
If the installed provider cannot set a custom base URL, use a maintained custom provider plugin rather than editing requests in transit.

Symptom	What it usually means	What to do
Credential validation still reaches OpenAI	The selected provider plugin ignores or does not expose a custom base URL.	Use a compatible provider/plugin version with an explicit endpoint field.
Chat works but knowledge retrieval fails	The workflow also needs an embedding or rerank model type.	Configure each required model type with a catalog route that advertises that endpoint.