The DMS AI plug-in connects private model services and Alibaba Cloud Model Studio services to your Dify workspace through the Model Provider page. Supported service types include large language model (LLM), Embedding, and Rerank.
Prerequisites
Before you begin, make sure you have:
-
A deployed Dify instance. See Deploy Dify.
-
(Required for Alibaba Cloud Model Studio LLM only) Internet access enabled for the virtual private cloud (VPC) where your Dify instance is deployed. See Use the SNAT feature of an Internet NAT gateway to access the Internet.
-
(Required for Alibaba Cloud Model Studio LLM only) An API key from Alibaba Cloud Model Studio. See Get an API key.
Integrate LLM services
Integrate private LLM services
Supported models:DeepSeek-R1-Distill-Qwen-1.5B,DeepSeek-R1-Distill-Qwen-7B,DeepSeek-R1-Distill-Qwen-14B,DeepSeek-R1-Distill-Qwen-32B, andDeepSeek-R1-Distill-Llama-70B.
Required parameters
| Parameter | Description | Format |
|---|---|---|
| Model Name | The name of the private model | e.g., DeepSeek-R1-Distill-Qwen-7B |
| API Key | The API key for your private model service | — |
| Endpoint URL | The endpoint of your private model service | http://172.17.XXX.XXX:XX/v1 |
Steps
-
Go to your Dify workspace. On the Studio tab, click the account name in the upper-right corner, and then click Settings.
-
On the Model Provider page, configure Model Name, API Key, and Endpoint URL, and then click Save.
Integrate Alibaba Cloud Model Studio LLM services
The following example uses DeepSeek-R1.
Required parameters
| Parameter | Description | Format / source |
|---|---|---|
| Model Type | The type of model to integrate | Select LLM |
| Model Name | The name of the model deployed in Alibaba Cloud Model Studio | e.g., DeepSeek-R1 |
| API Key | Your Alibaba Cloud Model Studio API key | Get an API key |
| Endpoint URL | The endpoint for Alibaba Cloud Model Studio services | https://dashscope.aliyuncs.com/compatible-mode/v1 |
Steps
-
Go to your Dify workspace. In the left-side navigation pane, click the
icon, and then click the Studio tab. -
Click the account name in the upper-right corner, and then click Settings.
-
On the Model Provider page, set Model Type to LLM, configure Model Name, API Key, and Endpoint URL, and then click Save.
Integrate Embedding or Rerank services
Connecting an Embedding or Rerank service to Dify requires two steps:
-
Register the model on the Model Provider page in DMS AI.
-
Select the model on the knowledge base configuration page in Dify.
The supported private Embedding and Rerank models are:
| Model type | Model name | Maximum context size |
|---|---|---|
| Embedding | bge-m3 | 8192 |
| Embedding | bge-large-zh-v1.5 | 512 |
| Rerank | bge-reranker-v2-m3 | 8192 |
Use private Embedding or Rerank services
Step 1: Register the model in DMS AI
Required parameters
| Parameter | Description |
|---|---|
| Model Type | Select Text Embedding or Rerank |
| Model Name | The name of the model |
| Service Provider | Your service provider |
| API Key | The API key for your private model service |
| Endpoint URL | The endpoint of your private model service |
Steps
-
Go to your Dify workspace. On the Studio tab, click the account name in the upper-right corner, and then click Settings.
-
On the Model Provider page, set Model Type to Text Embedding or Rerank, configure Model Name, Service Provider, API Key, and Endpoint URL, and then click Save.
Step 2: Select the model on the knowledge base configuration page
-
Go to your Dify workspace.
-
On the Knowledge tab, click Create Knowledge.
-
On the knowledge base configuration page, select the model you registered in Step 1.
Use Embedding or Rerank services provided by AnalyticDB for PostgreSQL
Step 1: Register the model in DMS AI
The maximum context size for AnalyticDB for PostgreSQL services is 2048.
Required parameters
| Parameter | Description | Source |
|---|---|---|
| Model Type | Select Text Embedding or Rerank | — |
| Model Name | The name of the model | — |
| Service Provider | Your service provider | — |
| Access Key | Your AccessKey ID | View AccessKey pairs of a RAM user |
| Secret Key | Your AccessKey Secret | View AccessKey pairs of a RAM user |
| Instance ID | The ID of your AnalyticDB for PostgreSQL instance | — |
| Region ID | The region where your instance is deployed | — |
| Context Size | Maximum input size in tokens | Maximum: 2048 |
For parameter details, see Rerank.
Steps
-
Go to your Dify workspace. On the Studio tab, click the account name in the upper-right corner, and then click Settings.
-
On the Model Provider page, set Model Type to Text Embedding or Rerank, configure all required parameters listed above, and then click Save.
Step 2: Select the model on the knowledge base configuration page
-
Go to your Dify workspace.
-
On the Knowledge tab, click Create Knowledge.
-
On the knowledge base configuration page, select the model you registered in Step 1.
Troubleshooting
Model cannot connect — endpoint URL rejected
Verify the endpoint URL matches the expected format exactly. For private LLM services, the format is http://172.17.XXX.XXX:XX/v1. For Alibaba Cloud Model Studio, use https://dashscope.aliyuncs.com/compatible-mode/v1.
Authentication failed — invalid API key
Check that the API key is copied correctly with no leading or trailing spaces. For Alibaba Cloud Model Studio, generate a new API key from the Model Studio console if the existing one has expired or been revoked.
Alibaba Cloud Model Studio model cannot be reached
Internet access must be enabled for the VPC where your Dify instance is deployed. If you skipped this step, configure an Internet NAT gateway with SNAT rules. See Use the SNAT feature of an Internet NAT gateway to access the Internet.
Model name not recognized after saving
Confirm the model name matches exactly what is deployed in your service — including capitalization and hyphens (e.g., DeepSeek-R1-Distill-Qwen-7B, not deepseek-r1-distill-qwen-7b).