Skip to content

README_EN.md

Latest commit

 

History

History
73 lines (53 loc) · 2.61 KB

README_EN.md

File metadata and controls

73 lines (53 loc) · 2.61 KB

DeepSeek_HTML

Simplified Chinese instructions: README_CN.md

☘️ Project Address: https://github.com/CQUPTLei/DeepSeek_HTML/tree/master

Conversation Screenshot: Insert Image Description Here

1. Project Structure

C:\USERS\14134\DESKTOP\DEEPSEEK
│  .gitignore 
│  DeepSeek.html  # All code in one file (not recommended)
│  index.html     # Webpage launch file
│  README.md      # readme
│
├─chat_history
│      example.json  # Chat history
│
├─css
│      style.css    # Styles
│
└─js
        config.js         	# Configuration
        config_example.js	# Configuration example
        function.js			# Main functionality implementation

2. Supported Features

  • DeepSeek's V3 model and R1 model;
  • Multi-turn dialogue;
  • Streaming output;
  • Save conversations as json files.
  • Markdown rendering;
  • Formula rendering;
  • Copy output as markdown format.

3. Usage Instructions

  1. Clone the project locally:
git clone [email protected]:CQUPTLei/DeepSeek_HTML.git

Downloading the Zip file is also acceptable.

  1. Go to the DeepSeek open platform to recharge and apply for an API key: https://platform.deepseek.com/api_keys
    • The price is quite affordable; I used 60,000 tokens for only 0.3 yuan.
  2. Copy the API key and paste it into the appropriate location in: js/config_example.js. Rename this file to: config.js.
  3. Create a new empty json file in the chat_history directory, or use example.json as well.
  4. Run index.html and follow the prompts to select where to save the chat.

Note: Currently only supports Chrome and Edge browsers.

4. Areas for Improvement

  1. Output rendering optimization;
  2. File storage logic needs optimization;
  3. Output of the reasoning process of the inference model.

5. Parameter Optimization

Refer to the DeepSeek development documentation: https://api-docs.deepseek.com/zh-cn/

  • For compatibility with OpenAI, you can also set base_url to https://api.deepseek.com/v1 to use, but note that v1 here is not related to the model version.

  • The deepseek-chat model has been fully upgraded to DeepSeek-V3, with no changes to the interface. By specifying model='deepseek-chat', you can call DeepSeek-V3.

  • deepseek-reasoner is the latest reasoning model from DeepSeek, DeepSeek-R1. By specifying model='deepseek-reasoner', you can call DeepSeek-R1.