react-agent VS duet-gpt

4️⃣ React Agent 🕵️‍♂️

Well, if you're coding in React, I have an experimental autonomous agent for you https://github.com/eylonmiz/react-agent, still experimental, but nice results so far.

I'm building an OSS React Agent library, in this example, it is reading data and feeding chatGPT from the filesystem, but it can be the same as for any integration, like Google search or scraping results. https://github.com/eylonmiz/react-agent/blob/main/backend/main/react-agent/generateComponents.ts One of the features I'm implementing now is reading library docs from the web, once I finish this feature I could show a full implementation that includes scaping the web, but it should be straightforward using puppeteer integration instead of filesystem integration

Website · Watch Demo · Github Repo · Docs · Discord

I'm using AI assistant tools for coding for the past 3-4 years, for Tabnine to Copilot and ChatGPT. They all go on the auto-completion route, which is nice, but it goes as far as saving me 30% - 70% of the time, making me more productive, let's say I can do 2X instead of 1X. In my humble opinion, it is not enough. I think we are a few steps away from exponential growth in software development speed, but it will be powered by software agents. No linear growth, exponential one. They generate specs, read specs, do PRs, do architecture planning, test code, and search for new libraries and technologies. It won't be linear growth, with no man in the loop (maybe initially), this could lead to the read Extreme programming vision, but autonomous. I've been working on an open-source experiment/project (ReactAgent) doing that for limited spec, using GPT-4, generating React.js (frontend framework) from user stories of a product manager (that BTW, we're also been generated by ChatGPT). I would love to know what you think, or even talk about how you see the future of software development! Repo: https://github.com/eylonmiz/react-agent Demo Video: Watch Demo » Website: https://reactagent.io/

I would love to know what you think, or even talk about how you see the future of software development!

Repo: https://github.com/eylonmiz/react-agent

as a self-taught web developer, that worked at multiple startups and co-founded a few of my own, my only advice is hard work on real projects. you can also work on OSS projects. if you're willing to work on an OSS project, I have one of my own that just started, I could also mentor you and help you get a paying job, this is the project: https://github.com/eylonmiz/react-agent

1) Show HN: Bloop – Answer questions about your code with an LLM agent (github.com/bloopai)

https://news.ycombinator.com/item?id=36260961

2) https://github.com/paul-gauthier/aider

3) Show HN: GPT Repo Loader – load entire code repos into GPT prompts (github.com/mpoon)

https://news.ycombinator.com/item?id=35191303

4) https://github.com/smol-ai/developer

5) codium

6) copilot

7) using gpt in the playground / chatgpt

8) jam.dev/jamgpt

9) magic.dev

10) https://github.com/kristoferlund/duet-gpt

Which ones am I missing?

DUETGPT PROMPT ============== You are an expert developer (AI) on a mission to help me (developer) achieve development related tasks. I will give you tasks, you will reply with a ResponseMessage suggesting bash commands for me run to complete these tasks. I will provide you the outcomes of the commands you suggest, and you will use this information to suggest more commands until the task is complete. This two-way interaction forms the basis of your collaboration. For instance, you may need to understand the contents of a file before suggesting modifications. - AI can only issue one ResponseMessage per interaction. - More complex tasks might require several interactions and several messages. - ResponseMessage consists of the sections defined below, in the order defined below (top to bottom) and nothing else. ResponseMessage: Question | Explanation, -----, More | Done, -----, Command, Question: type: Markdown formatted string description: Ask for clarification and more information about a task rules: - If a Command is issued, ResponseMessage MUST not contain a Question - If a Command is not issued, ResponseMessage MUST contain a Question ... Full prompt at: https://github.com/kristoferlund/duet-gpt

Prompt and code: https://github.com/kristoferlund/duet-gpt