Welcome to PromptManager

PromptManager allows you to define your prompts in a simple UI directly in VSCode. These prompts are then saved in a generated_prompts folder as protobuf files which can easily be read by your preferred programming language.

Features

• Write and edit prompts
• Include variables with {{<variable_name>}}
• Save and remove prompts easily
• See how many tokens are included in your prompt

How to use

Open up your command pallette in VSCode (⌘⇧P on Mac) and type Show Prompt Manager UI. This will immediately create a generated_prompts directory which will house the prompt.proto file along with any prompts you create using the extension. You can configure the generated_prompts path in the extension settings (⌘, on Mac). When you Save a prompt, it is stored as a .pb file which can be read in your code using the protobuf library. Below is the Prompt schema along with example code to read the prompts for Python. Make sure you have protobuf installed on your machine.

Protobuf Schema

syntax = "proto3";

message Prompt {
  string id = 1;
  string title = 2;
  string content = 3;
  repeated string variables = 4;
  int32 tokenCount = 5;
}

Python

1. Install the protobuf library:

pip install protobuf

2. Compile the prompt.proto file to generate the Python module. This takes place after you have saved a prompt:

protoc --python_out=. <path_to_prompt.proto>

3. Now, you should have a prompt_pb2.py file in the same directory as your prompt.proto. You can use this module to read the .pb files. Here's a Python code snippet as an example:

import prompt_pb2
import os

# This function reads a protobuf file and parses it into a Prompt object.
def read_prompt_file(file_path):
    prompt = prompt_pb2.Prompt()
    
    with open(file_path, "rb") as f:
        prompt.ParseFromString(f.read())
    
    return prompt

# This function replaces variables in the content of a Prompt object.
# It uses the Python built-in string format method to replace variables.
# The SafeDict class is a subclass of dict that returns the original key
# wrapped in curly braces if the key is not found in the dictionary. This is
# useful because it allows us to safely use the format method on the content
# string, without having to worry about missing keys.
def replace_content_variables(prompt, new_values):
    class SafeDict(dict):
        def __missing__(self, key):
            return '{' + key + '}'

    content = prompt.content

    # Use string formatting to replace variables in the content string
    return content.format_map(SafeDict(new_values))

# Define the actual path to the generated_prompts and example.pb file
generated_prompts_path = "path/to/generated_prompts"
pb_file = os.path.join(generated_prompts_path, "example.pb")

# Read the protobuf file and parse it into a Prompt object
prompt_data = read_prompt_file(pb_file)

# Define the new values to replace in the content
new_values = {"user_name": "Gabe", "customer_name": "Akash"}

# Replace the variables in the content
updated_content = replace_content_variables(prompt_data, new_values)

# Print the updated content
print(updated_content)

Javascript (Node)

1. Install the protobufjs library:

npm install protobufjs

2. Compile the prompt.proto file to generate the Javascript module. This takes place after you have saved a prompt:

npx pbjs -t static-module -w commonjs -o prompt_pb.js prompt.proto

3. Create a new JavaScript file, e.g., read_pb.js. You can use this module to read the .pb files. Here's a JavaScript code snippet as an example::

const fs = require('fs');
const protobuf = require('protobufjs');

protobuf.load("prompt_pb.js", function(err, root) {
    if (err) {
        throw err;
    }

    // Obtain a message type
    const Prompt = root.lookupType("Prompt");

    function readPromptFile(filePath) {
        const buffer = fs.readFileSync(filePath);
        const decodedPrompt = Prompt.decode(buffer);
        return decodedPrompt;
    }

    function replaceContentVariables(prompt, newValues) {
        let content = prompt.content;
        
        // Use regular expression to replace variables in the content string
        for (let variable in newValues) {
            content = content.replace(new RegExp(`{{${variable}}}`, 'g'), newValues[variable]);
        }

        return content;
    }

    // Replace with the actual path to the generated_prompts and example.pb file
    const generatedPromptsPath = "path/to/generated_prompts";
    const pbFile = `${generatedPromptsPath}/example.pb`;

    const promptData = readPromptFile(pbFile);

    // Define the new values to replace in the content
    const newValues = {"user_name": "Gabe", "customer_name": "Akash"};

    // Replace the variables in the content
    const updatedContent = replaceContentVariables(promptData, newValues);

    console.log(updatedContent);
});

Release Notes

Courtesy of Zenfetch

Upcoming features

1. Group prompts into collections

1.0.0

Initial release of PromptManager