Prompt.md

Prompt Engineering with GPTutor

What is GPTutor? How it work?

GPTutor is an AI tool that enhances your programming and code-reading efficiency. Check out this video first:

https://www.youtube.com/watch?v=uTNXQuKrnKI

How to prompt engineering with GPTutor?

You can design prompts in GPTutor by clicking settings → edit prompt (or search GPTutor.prompt in VS Code’s setting), then click Edit in setting.json.

Then, you will see the value of GPTutor.prompts and GPTutor.promptInputBox arrange in the following format:

"GPTutor.prompts": {
    "global": {
      "explain": {
        "display_name": "🧑‍🏫 Explain",
        "type": "chatgpt-prompt",
        "prompt": [
          {
            "role": "system",
            "content": "You are a Senior ${languageId} Developer..."
          },
          {
            "role": "user",
            "content": "...The ${languageId} code I selected is in the following triple quotes \"\"\"${selectedCode}\"\"\". You don't need to repeat the selected code again. Please focus on explain target ${languageId} code. Output in ${outputLanguage}."
          }
        ]
      },
      "comment": {...},
      "Review": {...},
    }
    "specificLanguage": {
      "move": {
        "explain": {
          "display_name": "🧑‍🏫 Explain",
          "type": "chatgpt-prompt",
          "prompt": [
            {
              "role": "system",
              "content": "Move is an open source language for writing safe smart contracts. It's format is similar to Rust. You are a Senior ${languageId} Developer..."
            },
            {
              "role": "user",
              "content": "..."
            }
          ]
        },
        "comment": {...},
        "Review": {...}
      }
    }
  }
"GPTutor.promptsForInputBox": {
    "global": {
        "default": {
        "display_name": "🤖 Code Generate",
        "type": "chatgpt-prompt",
        "prompt": [
            {
            "role": "system",
            "content": "You are a helpful assistant that generate the ${languageId} code according to user's instruction..."
            },
            {
            "role": "user",
            "content": "Instruction: ${user_input}"
            }
        ]
    }
}

(… is the part omitted for a clearer view (otherwise, it is too long).)

The following is the explanation of the GPTutor.prompts and GPTutor.promptsForInputBox:

1. GPTutor.prompts is the prompt configs when you hover the selected code.
2. GPTutor.promptsForInputBox is the prompt configs when you interact with the input box.
3. Global: These prompts will take effect in all languages.
4. Specific Language: These prompts will only affect the key in the specified language. For instance, the prompt labeled "move" will only show up when the document's language is Move.
5. Keys: explain, comment, and Review are the function's keys. If a key in the specific language duplicates a key in the global section, the specific language key will be prioritized.
6. display_name: This is the name that will be displayed in GPTutor's pop-up or dropdown menu.
7. type: This indicates the type of prompt. Currently, only chatgpt-prompt is supported, but we plan to support others such as Google-Bard, Azure, and 3rd party Service Proxy APIs in the future.
8. prompt: This is the specific prompt that will be used for ChatGPT.

Lists of Prompts

Key	Description
languageId	The language identifier of the current document in Visual Studio Code. (For example, when you open a Python file, it will be "python")
entireDocument	The entire content of the current document as a single string.
codeContext	The code surrounding with the current cursor position, including 300 lines before and after.
definitionContext	The code related to the definition of the selected function. It is obtained by calling the getDefinition command in VS Code.
selectedCode	The currently selected code by the user.
codeBefore	All code before the current cursor position.
codeAfter	All code after the current cursor position.
codeContextBefore	The code before the current cursor position, limited to 100 lines.
codeContextAfter	The code after the current cursor position, limited to 100 lines.
outputLanguage	the output language user selected. For example, English.

Examine Prompt Value:

To see the current prompt sent to the OpenAI, you may click the Show Prompt button on top of the output. You can edit it directly and run again. Yet to use it permainately, you should go to VSCode's Setting.json by click Settings → Edit Prompt.