irthomasthomas / undecidability

13 stars 2 forks source link

Plandex - open source, terminal-based AI coding engine #864

Open ShellLM opened 3 months ago

ShellLM commented 3 months ago

Intro | Plandex Docs

Snippet

"Introduction

Plandex is an open source, terminal-based AI coding engine that helps you work on complex, real-world development tasks with LLMs.

It combines multiple agents to complete tasks that span many files and model responses. When you give Plandex a task, it continues working automatically until the task is determined to be complete.

Use cases

What makes Plandex different?

Version Control

Plandex gives the LLM its own version-controlled staging area/sandbox (separate from your project's git repo) where all of its proposed changes are accumulated. This allows you to:

Context Management

Apart from version control, Plandex also helps you manage what's in the LLM's context:

Focus

For now, Plandex is focused on:

It doesn't yet do automatic execution of code or automatic selection of context—both are left to the developer.

In other words, Plandex isn't (yet) shooting for full autonomy. While we do plan to move in this direction over time, we think the current level of model capabilities make Plandex's focus a sweet spot for achieving real productivity gains.

Though full autonomy is certainly an enticing prospect, in practice it often means wasting a lot of time and tokens letting the LLM spin its wheels on problems that are trivial for a developer to identify and fix. You can definitely use Plandex to help debug its own code, but for now we think it's best to make this opt-in rather than the default behavior.

Models

Plandex uses OpenAI models by default, as so far we've found them to offer the best balance between intelligence and reliability.

That said, you can also use it with a wide range of other models, including Anthropic Claude, Google Gemini, Mixtral, Llama and many more via OpenRouter.ai, Together.ai, or any other OpenAI-compatible provider.

Languages and Platforms

You can use Plandex to work with any language or framework that the underlying LLM has been trained on. For the largest models, this includes just about any language or framework you can think of, though output quality will tend to be best for those that are more popular and therefore better represented in the training data.

Plandex is cross-platform and easy to install. It supports Mac, Linux, FreeBSD, and Windows. It runs from a single binary with no dependencies.

Hosting

Plandex runs on a client-server architecture. The Plandex server is open source and can be self-hosted. A cloud-hosted option is also offered for getting started as quickly as possible with minimal setup."

Suggested labels

{'label-name': 'version-control', 'label-description': 'Facilitates management of code changes in a version-controlled environment for AI development.', 'gh-repo': 'Plandex Docs', 'confidence': 56.22}

ShellLM commented 3 months ago

Related content

681 similarity score: 0.87

734 similarity score: 0.87

750 similarity score: 0.86

743 similarity score: 0.85

508 similarity score: 0.85

706 similarity score: 0.85