Gitchat is an exciting new tool that enables developers to have natural conversations about their code. It aims to make understanding and navigating complex codebases effortless through the power of language models. The core innovation of Gitchat is integrating GitHub repositories with chatbots powered by large language models like LLMs. By linking code hosted on GitHub to conversational interfaces, Gitchat allows developers to simply ask questions and get intelligent answers about their projects. Gitchat works by ingesting and vectorizing code from a connected GitHub account. This allows advanced language models to build contextual understanding of the codebase. Developers can then chat with the language model by asking questions in plain English such as "what is the purpose of this class" or "how is this module interconnected with the rest of the code". The language model leverages the vectorized code, context from docs/commits, and its own reasoning capabilities to provide helpful answers. It can explain code, provide design rationale, suggest improvements, summarize implementations, and more. The conversational nature makes it easy to interact and dig deeper. Gitchat aims to solve many pain points in software development like onboarding, tribal knowledge, finding dependencies, and navigating legacy code. By creating a code companion that developers can have a dialogue with, Gitchat hopes to make developer workflows more intuitive and productive
Category tags:Team member not visible
This profile isn't complete, so fewer people can see it.
Team member not visible
This profile isn't complete, so fewer people can see it.
"Cool use of embeddings, langchain and various models. Would have been good to have a working link to your codebase to get a better sense of how you're doing certain things. Hope you had a good time! Definitely agree that querying your repo can be super helpful."
Alfredo Ramos
Clarifai Humanoid
"Seems like github copilot which is cool! It wasn't hosted as a streamlit UI module. I didn't hear mentions of Clarifai in the video or description, not clear if they even used Clarifai. Broken link to github so can't even verify if Clarifai was used. "
Matthew Zeiler
CEO, Founder, Deep Learning, Machine Learning, and Image Recognition Wizard