🍬Taffy

AI Code Editor focused on multi file editing and better context

Demo

🏃 Getting Started

Add to VSCode from the Visual Studio Marketplace
Add either a OpenAI, Anthropic or DeepSeek API Key
Select any portion of your code, then press Ctrl+Shift+M to ask a question

💪 Motivation

We found that often with AI code assistants, they just don't seem smart enough to generate the correct code in the context required. Examples include

Not follow codebase best practices
Using outdated versions of libraries
Not understanding types within the codebase
Generation being limited to single file edits

The thing is, all of these are not outside the capabilities of modern LLMs - with the right information in the context window, all these issues can be resolved.

However, with alternative AI code assistants, they fall fall short in one of these ways:

Not having multi-file support
Not having sufficient context to answer the question correctly
Racking up large API bills because of large context windows

With our initial release, we've solved multi-file support, and with subsequent releases we hope to solve the other problems as well.

🎁 Features

Current Features

Multi File Editing
Fast edits (Only edit the section being updated)
LLM Tools (LLM can read other files in your codebase, and make necessary updates)
@mentions files in your codebase to reference them
Full Prompting Transparency (Toggle to see exact prompts sent to LLM)
Bring Your Own key (BYOK)
MIT Licensed

Upcoming Features (as of Oct 2024)

Right now, we have a base of features that we believe make a good foundation for the developer experience that we expect from an AI code assistant.

We have several ideas of how we can improve the developer experience, and we need your feedback and votes in order to better shape the development of this product.

Small Model Context Big Model Thinking: 👍Upvote this feature
Types Context: 👍Upvote this feature
Library Context: 👍Upvote this feature
Diagnostics Context: 👍Upvote this feature
Best Practices Context: 👍Upvote this feature

🏠 Local Development

To run Taffy locally:

npm i          # Install dependencies
npm run serve  # Start the development server

Press F5 to debug or run it in the standard VSCode extension host.

📃 License

MIT License

taffy

FireJet