Connect Dify
Add a text model through Dify's OpenAI-API-compatible provider and select it in an app or workflow.
https://aelder.dev/v1Use an existing Dify workspace where you can manage model providers. The service key you add can be used across that workspace. This guide configures text generation through OpenAI-compatible Chat Completions.
1. Open the compatible provider
In current Dify, open Integrations → Model Provider. Older versions place Model Provider under workspace Settings. Install the OpenAI-API-compatible provider by LangGenius if it is absent, then select Add Model on its card.
Use this generic compatible provider for the connection, regardless of the model’s vendor name. Keep existing provider credentials and workspace defaults.
2. Fill in the model settings
| Field | Value |
|---|---|
| Model Type | LLM |
| Model Name | An exact ID from the authenticated models endpoint |
| API Key | Your service key, entered privately without Bearer |
| API Base URL | The service Base URL above, including /v1 |
| Completion mode | Chat |
| API Type, if shown | Chat Completions API (/chat/completions) |
| Compatibility mode, if shown | OpenAI compatible |
| Model context size | The selected model’s verified context limit |
Do not append /chat/completions to the Base URL. Leave model name for API endpoint empty when Model Name already contains the exact API ID. Do not assume that a prefilled context size describes your selected model. If its context limit is not published in the service catalog, confirm it with support before saving.
Leave unverified vision, audio, web search and tool capabilities off. Save the model. Dify validates credentials with model requests, which can spend balance.
3. Select it in your app
Open the app or workflow you want to configure. In its LLM node or model selector, select the new compatible provider and model, then save the app’s draft. Test with one short text prompt in preview before publishing; preview requests also spend balance.
The provider’s saved credentials persist in the workspace. If a workflow still uses another model, inspect each LLM node rather than changing all workspace defaults. For validation failures, check the exact model ID, /v1 suffix and supported parameters. Knowledge retrieval and agent tools require separate setup.
See Dify’s model provider guide and the official plugin’s configuration schema.
Need the key? Open your AELDER dashboard. For direct API calls, see the API reference.