#OpenSourceDiscovery 81: Open Interpreter

Let LLMs run code locally in your terminal

Today I discovered…

Open Interpreter

A Python-based command-line tool that converts natural language instructions to code (Python, Javascript, bash, etc.) using LLMs and executes it locally. Start chatting by simply running `interpreter` command in the terminal.

While it has many use cases, I am particularly excited about its use in server management and building developer tools.

💖 What I like:

• Ridiculously easy to setup and use without any learning curve

• Auto corrects itself without additional input. Also considers the user feedback when provided

• A well-written summary of the steps combined with the prompt for approval makes everything feel under control

• Support for CodeLLaMA (in addition to GPT4 Code Interpreter)

👎 What I dislike:

• I wished it worked completely offline but when I tried it using LLaMA, it produced only “hello world” after a long wait. So that dream is still far away

• It was unable to fix its problem in an ffmpeg command. I wished it had browsing capabilities to consider new information such as a Stackoverflow question or GitHub issue on that topic

• Overall, I found it highly useful for general tasks and felt productive in areas where I am a beginner, I am impressed with the accuracy and the developer experience. But for specific tasks and the areas where I already am an expert, it felt like it is slowing me down as I find my way to achieve the same goals via interpreter. Having said that, it might change few more releases down the line or when I develop intuition about when to use it and when not to

⭐ Ratings and metrics

Based on my experience, I would rate this project as following

• Production readiness: 8/10

• Docs rating: 6/10

• Time to POC(proof of concept): 1 min

Author: Killian @hellokillian

Demo | Source

Tech Stack: Python, GPT4 or LLaMA

---

Around the web

• Discussion on HackerNews

• Discussion on ProductHunt

• You can also discuss in response to this post on Substack

---

If you discovered an interesting Open-Source project and want me to feature it in the newsletter, get in touch via the form above. To support this newsletter and Open-Source authors, follow #OpenSourceDiscovery on LinkedIn and Twitter