Prompt Runner VSCode Extension

Overview

Prompt Runner is a Visual Studio Code extension that allows users to run prompt files against various LLMs directly from the editor. It turns VSCode into a powerful prompt IDE. You no longer have to tolerate the tiny chat interface or copy pasting between apps for complex prompts.

The prompt file format is designed to help users define structured prompts for interacting with large language models (LLMs). It supports a variety of tools such as input variable, web browsing, image, promtp chaining, code execution, control loop, etc. to help users develop and evaluate prompts more efficiently. In fact, it has the potential to be a fully featured agent framework.

Check out the this doc for a detailed documentation of what this extension supports.

demo

Installation

Install on VSCode Marketplace

Features

Run Prompt Files: Run the content of the active editor as a prompt against the selected LLM provider and model.
Extensive Tools: The prompt file format supports tools such as input variable, web browsing, image, promtp chaining, code execution, control loop, etc.. In fact you can build agents entirely with prompt files.
Multiple Providers (Ongoing): Support for various providers including OpenAI, Ollama. I am working on Anthropic and VSCode Copilot as model providers.
Observability (TODO): Integrate with prompt management platforms such as Langfuse for observability, feedback and prompt management.

Usage

Commands

Press Ctrl+Shift+P or F1 to open the command palette, and type the following commands:

Compile Prompt File: Compile the current prompt file.
- Command: Prompt Runner: Compile Prompt File
Run Prompt File: Run the current document as a prompt using the default provider and model.
- Command: Prompt Runner: Run Prompt File
Run Prompt File with Selected Provider: Run the current document as a prompt after selecting a provider and model.
- Command: Prompt Runner: Run Prompt File with Selected Provider
Select Default Model: Set the default provider and model to use for future prompts.
- Command: Prompt Runner: Select Default Model
Select Output Location: Set the output location of the compiled prompts or running outputs. Choose webview panel for better readability, or output channel for easier copy.
- Command: Prompt Runner: Select Output Location

Configuration

The extension allows users to configure providers and models in the VSCode settings. If no configuration is provided, the extension will initialize with default settings.

Default Providers

The default providers include:

OpenAI: Supports models like gpt-3.5-turbo, gpt-4, etc.
Ollama: A locally hosted LLM with support for custom models.
Anthropic (TODO): Includes models such as claude-3.5-sonnet, claude-3-opus, etc.
VSCode Copilot (TODO): Supports several versions of the GPT models provided by VSCode.

Provider Configuration Schema

The configuration for providers can be defined in the VSCode settings file. Currently supported provider types are openai and ollama. anthropic is coming next. Example config:

{
    "prompt-runner.providers": [
        {
            "name": "openai",
            "type": "openai",
            "base_url": "https://api.openai.com/v1", // So that you can use OpenAI-compatible servers
            "api_key": "<your-api-key>",
            "models": [
                "gpt-3.5-turbo",
                "gpt-4",
                "gpt-4o",
                "gpt-4o-mini"
            ],
            "organization_id": "",
            "project_id": ""
        }
    ]
}

Manual Compilation

Clone this repository or download the source code.
Install the necessary dependencies using npm or yarn:
```
npm install
```
or
```
yarn install
```
Open the project in Visual Studio Code.
Press F5 to launch the extension in a new VSCode window for testing and development.

Debug

The logs of this extension are at the output channel. Select Prompt Runner in vscode output channel for debug logs.

Development

TODO

Contribution

Contributions are welcome! Please fork the repository, make your changes, and submit a pull request.

License

This project is licensed under the MIT License.

Prompt Runner

Jin Shang