Dify framework configuration

# Dify Framework Configuration

Dify is an open-source large language model (LLM) application development platform. It combines the concepts of Backend as a Service and LLMOps, enabling developers to quickly build production-grade generative AI applications.

Its official website is: https://dify.ai/zh

The documentation address is: https://docs.dify.ai/zh-hans/introduction

## Deploying Dify

>i Dify provides a cloud service that can be used without deployment. You can go to: https://cloud.dify.ai/ to log in and use it.

MrDoc recommends that users deploy Dify using Docker Compose. The official documentation address is: https://docs.dify.ai/v/zh-hans/getting-started/install-self-hosted/docker-compose

## Configuring Dify

### Configuring Models

Dify is a large language model framework that supports integration with various model providers (OpenAI's ChatGPT, Baidu's ERNIE Bot, Gitee AI, and self-deployed Ollama, etc.). After deploying Dify, you need to configure the model providers in Dify.

In "Settings" - "Model Providers", configure your models:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_185625_7513860.6042399288846461.png)

### Creating a Blank Knowledge Base

>d The purpose of the Dify knowledge base is to enable Dify applications to retrieve content from the knowledge base during chat, and then answer based on the retrieved content. If you do not need the large language model integrated with Dify to perform targeted knowledge base question answering, you can ignore this step.

After deploying Dify, under the "Knowledge Base" section in Dify, click "Create Knowledge Base":

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_184915_2570750.9796835726438711.png)

You can choose to upload your local knowledge base files, or "Create an empty knowledge base". MrDoc recommends that you create an empty knowledge base specifically for storing data from MrDoc.

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_185029_6975430.16591581362032481.png)

Click "Create an empty knowledge base" and enter the name of the knowledge base to create it:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_185116_8586320.8358555175633485.png)

In this way, a Dify knowledge base is created.

#### Obtaining the Knowledge Base API Key

In the "API" section of the Dify knowledge base page, obtain the knowledge base API key for later use:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_210748_1575710.5861644902323248.png)

### Creating an Application

Dify supports creating different types of applications for different usage scenarios, such as: chat assistant, text generation application, Agent, chatflow, workflow, etc. For detailed differences, see its documentation: https://docs.dify.ai/zh-hans/guides/application-orchestrate/readme#%E5%BA%94%E7%94%A8%E7%B1%BB%E5%9E%8B

The application APIs actually used by MrDoc are the "Chat API" and the "Text Generation API". You can create the corresponding type of application as needed to obtain its application API.

On the Dify Studio page, click "Create Blank App", and select "Chat Assistant" and "Basic Orchestration":

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_190314_3973870.40938912582459375.png)

After creating the application, click to enter the application settings page.

In the upper right corner, select a model for your application:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_190640_5791490.3668058962567673.png)

Write the prompt according to your actual needs, for example:

```
请根据上下文，回答用户的问题；如果上下文中没有相关信息，请回复「我暂时无法回答你的问题」，所有的回答以 Markdown 样式进行输出
```

In the context, add the created knowledge base as the context of the application.

Finally, publish the application.

### Obtaining the Application Embed Website Code

Click the "Publish" button, and select "Embed Website" from the drop-down menu:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_211215_3325860.837526273714381.png)

In the pop-up dialog, select the second type and copy its code for later use.

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_211311_0331120.6417233674702577.png)

### Obtaining the Application API Key

In the application, select "Access API", and click "API Key" to create an application API key for later use:

![](http://file.mrdoc.pro/mrdoc-oss/img/2024/05/31/2024-05-31_211045_3822220.5463592458255666.png)

## Configuring MrDoc

1. In the "Basic Configuration" of the "AI Integration" page in the backend management, select "Dify" for the AI framework.

2. In the "Dify Configuration" of the "AI Integration" page in the backend management, fill in Dify's "API Address", "Chat API Key", "Knowledge Base API Key", and "Text Generation API Key";

3. Click "Read Dify Knowledge Base" and select the Dify knowledge base to be bound.

4. If "Chat Dialog" is enabled in "Basic Configuration", fill in the "Embed Website Code" of the Dify application in the embed code of "Dify Configuration".

### Knowledge Base Synchronization

To upload documents from MrDoc to the Dify framework, you need to select the scope of document upload. Documents in the projects within the selected scope will be uploaded to the selected AI large language model framework.

>d Before knowledge base synchronization, be sure to configure the target knowledge base of the "AI Framework".

> i You need to configure the "Knowledge Base API Key" of the AI framework.

Determine how documents in MrDoc are synchronized to the AI large language model framework.

- Manual synchronization: synchronize the content of specified project documents to the specified knowledge base;
- Automatic synchronization: when documents in the specified project change, synchronize them to the specified knowledge base.

![](/media/202605/3b6daa1d48384b6492cb1fbb43c6ae4d1394.png)