We’ve open-sourced our internal AI coding IDE

We built this IDE internally to help us with coding
and to experiment with custom workflows using AI.
We also used it to build and improve the IDE itself.
It’s built around a flexible
extension system,
making it easy to develop, test, and tweak new ideas fast.
Each extension is a Python package that
runs locally.

**GitHub Repo:** [https://github.com/notbadai/ide/tree/main](https://github.com/notbadai/ide/tree/main)
**Extensions Collection:** [https://github.com/notbadai/extensions](https://github.com/notbadai/extensions)
**Discord:** [https://discord.gg/PaDEsZ6wYk](https://discord.gg/PaDEsZ6wYk)

## Installation (macOS Only)

To install or update the app:

```bash
curl -sSL https://raw.githubusercontent.com/notbadai/ide/main/install.sh | bash
```

***We have a set default extensions installed with the above installation command, ready to use with the IDE.***

## Extensions

Extensions have access to the file system, terminal content, cursor position, currently opened tabs, user selection, chat history etc.
So a developer can have own system prompts, call multiple models, and orchestrate complex agent workflows.

**Chat and apply** is the workflow I use the most.
You can quickly switch between different chat extensions for different types tasks from the dropdown menu.
To apply code suggestions we use [Morph](https://x.com/morphllm).

**For complex code sometimes code completions** are better.
We have a extensions that suggests code completions and the editor shows them inline in grey.
These can be single or multi-line.
It's easy to switch the models and prompts for this to fit the project and workflow.

**Extensions can also have simple UIs**.
For instance, we have an extension that suggest commit messages (according to a preferred format) based on the changes.
It shows the the suggestion in a simple UI and user can edit the message and commit.

More features and extensions are listed in our [documentation](https://github.com/notbadai/ide/tree/main).

## Example Extension Ideas We’ve Tried

* Determine the file context using another call to a LLM based on the request

In our initial experiments, the user had to decide the context by manually selecting which files to add. We later tried
asking an LLM to choose the files instead, by providing it with the list of files and the user’s request, and it turned
out to be quite effective at picking the right ones to fulfill the request. Newer models can now use tools like read
file to handle this process automatically.

* Tool use

Adding tools like get last edits by user and git diff proved helpful, as models could call them when they needed more
context. Tools can also be used to make edits. For some models, found this approach cleaner than presenting changes
directly in the editor, where suggestions and explanations often got mixed up.

* Web search

To provide more up-to-date information, it’s useful to have a web search extension. This can be implemented easily
using free search APIs such as [DuckDuckGo](https://duckduckgo.com/) and
open-source [web crawlers](https://github.com/unclecode/crawl4ai).

* Separate planning and building

When using the IDE, even advanced models weren’t great at handling complex tasks directly. What usually worked best was
breaking things down to the function level and asking the model to handle each piece separately. This process can be
automated by introducing multiple stages and model calls for example, a dedicated planning stage that breaks down
complex tasks into smaller subtasks or function stubs, followed by separate model calls to complete each of them.

* Shortcut based use-cases like refactoring, documenting, reformatting


https://redd.it/1ofqa8c
@r_opensource

I just open-sourced an offline "mini-Google" semantic search engine you can install and forget until you need it, for emergencies, off-grid use, or personal notes

In case you want to have a look, the link is: [https://github.com/Ohrest88/offlinesearchengine](https://github.com/Ohrest88/offlinesearchengine)

It was an experiment where I wanted to see if something like an offline "mini-Google" could run completely on-device (on my Android phone), with semantic search (searching by meaning, like popular search engines, not just keywords).

That opened a whole can of worms:

* running a small in-built model for generating embeddings
* storing the vector embeddings in a local database
* doing vector search for semantic similarity
* keeping everything offline
* make it work on android

The second part of the experiment was making it ideally multiplatform, so it's in flutter and currently there are pre-built executables for Android (play store) and Linux (AppImage)

Of course, happy with any feedback :)

https://redd.it/1ofs1wv
@r_opensource

Tired of tailoring your resume to multiple Job denoscriptions? I made this!
https://github.com/ricky-aufvaa/Dynamic-Resume-Builder

https://redd.it/1ofoypf
@r_opensource

(Gathering Feedback and Activity) MCServerNap – Automatically start & stop your Minecraft server on player join/leave (Rust application)
https://github.com/uwuhazelnut/MCServerNap

https://redd.it/1ofv9nr
@r_opensource

Any open source photoshop alternative?

Any open source photoshop alternatives that can actually rival with adobe?

https://redd.it/1ofvv11
@r_opensource

Is there a forum standard that combines the best from traditional forums and IMs?

Often a project has both conventional forums (Discussion / BBPress / MyBB) and Instant Messaging chatrooms (Matrix / SimpleX / IRC), often causing community fragmentation on small projects. Is there a standard that unites their advantages in one place?

https://redd.it/1ofwv1c
@r_opensource

looking for an open source mobile OS that runs on a huawei y5p

I recently changed phone because my old Huawei y5p was blocked everywhere by google supported programms and couldn't made any security update even from huawei itself. So now I am looking for some degooglelised and huawei-free open-source os (basicaly kind of linux for mobile) that could run on this phone to give it a second life instead of forget it in a drawer.

https://redd.it/1ofzp3f
@r_opensource

FloatView - A video browser that finds and fills unused screen space automatically
https://github.com/mrconter1/FloatView

https://redd.it/1og028g
@r_opensource

Markdrop - A powerful visual markdown editor and builder

Hey everyone! I just launched Markdrop, a minimalist, feature-rich markdown editor designed for speed and simplicity!

GitHub Repo: https://github.com/rakheOmar/Markdrop

If you’re into webdev, open-source, or just looking to make your first contribution, we’d love your feedback, ideas, and help!

How you can help:

Open a PR if you see something you want to fix or build! We review and merge good PRs quickly!
Starring the repo! This is the #1 way to help—it massively boosts our visibility and helps others find the project!
Check out our "good first issue" and "beginner friendly" labels. We've set up several issues specifically for new contributors.
Suggest new features you'd like to see.

Every contribution, (even a small doc fix or a star!) means a lot to us. Let's build something cool together!

https://redd.it/1og26d0
@r_opensource

GitHub - whitlocktech/Otexum-Pulse: A windows application to auto launch an application based on user idle time.
https://github.com/whitlocktech/Otexum-Pulse

https://redd.it/1og3ox7
@r_opensource

Looking for someone to fork/update a project

https://github.com/szszss/TpacTool

ever since the latest update, this tool is no longer able to open tpac files, it crashes when you click on any of the assets. unfortunately i'm not a developer just a concerned modder whod like to be able to extract 3d models and textures again.

https://redd.it/1og2wxy
@r_opensource

Best realistic FOSS driving simulator?

Realistic not in graphics, but to prepare for driving school. Like FlightGear from the world of cars. It should support wheels, pedals and stuff.

https://redd.it/1og5tnd
@r_opensource

MouseUtils - An autoclicker for Windows and Linux

MouseUtils is an autoclicker, currently supporting Windows and Linux (X11 only), written in C++ with a Qt6 GUI.

This is my first C++ project I've released, and to be honest, made (excluding the classic "Hello, world!"). There will most likely be (and I believe there is) bad practices and general mistakes in the source code. I'm a bit of a newb :P

Feel free to play around, and if you're interested in contributing, please do!

GitHub: https://github.com/AngusAU293/MouseUtils

Thank you, and have a great day!

https://redd.it/1og7bu1
@r_opensource

What are your favorite open source services?

Hey everyone! I have been DeGoogling and I’m prioritizing replacing Google services with ones that are also open source. I thought it would be interesting to share favorite opensource tools!

Mine are:
Browser: Librewolf
Email: Tuta Mail
Calendar: Tuta Calendar
Photos: Ente Password manager: Bitwarden & KeepassXC
Google docs: Cryptpad



https://redd.it/1ofz88q
@r_opensource

What are you using for mailing lists?

So I have an open source project that has a mailing list where people can sign up to hear about new versions. It has a few hundred subscribers and I send 0-2 e-mails a month, most months nothing. Everyone on it has explicitly signed up for it.

Up until now I've been running a self-hosted phpList instance but that means I'm dealing with issues with my web host's IP address reputation etc. I'd like to move to something hosted. So question 1 is: What are people using?

MailChimp is an option. I tried phpList.com but something's wrong with my account configuration and I'm not getting a response from their support. Searching around here I found someone recommend SendFox, which looks really nice, so I thought I'd try that.

But that brings us to question 2: A lot of these services require a physical address be attached to each e-mail to comply with the CAN-SPAM act, including SendFox, and I'd really rather not blast my personal address out like that. What are people doing for that? Paying for a P.O. box or one of these services that give you an address and scan your mail for you? Or should I stay with self-hosted and try to fight out the deliverability issues myself? I know I'm not the first person to run into this.

https://redd.it/1ogc7sn
@r_opensource

ROM /e/OS

Is anyone else testing this ROM?
I had 3 problems with it.
1° I can't install applications from outside the store, it says "unable to install".
2° It always fails when I put a password on my cell phone, not even one type of password remains.
3° On the map I cannot locate where I am with echolocation.
I haven't found how to solve it yet, not one of the 3. .-.
If anyone managed to do it, please give a tip to your friends.

https://redd.it/1ogcrbd
@r_opensource

Flathub announces toolchain fixes to address longstanding license and copyright compliance issues
https://docs.flathub.org/blog/enhanced-license-compliance-tools

https://redd.it/1ogdab6
@r_opensource

My First Open Source Project: GitRead

Hey everyone! 👋

I'm excited to share my first open-source project with the community — GitRead, an AI-powered README generator that helps developers create professional project documentation in seconds.

🔧 GitRead analyzes your GitHub repository, generates a high-quality README, and allows you to customize it with a live Markdown editor and preview. Whether you're launching a new project or improving an existing one, GitRead can save you time and make your repo shine!

This project means a lot to me — it’s my first open-source contribution and I'm really looking forward to feedback from other developers. I'm super happy (and a little nervous 😅).

# 💻 GitHub Repository

👉 https://github.com/PoRiFiRo123/gitread

## 🌐 Live Demo
👉 https://git-read.vercel.app

https://redd.it/1ogegx5
@r_opensource

🚀 Built a tool to make open source contributions easier — looking for feedback!

Hey everyone! 👋

I’ve been working on something called Open Source Contribution Captain — a free tool that helps newcomers find beginner-friendly GitHub issues matched to their tech stack.

It also uses AI-generated summaries to explain what each issue needs, what’s been tried, and any blockers — so you can skip hours of manual digging and get started faster.

🌐 Try it here: https://opencontributioncaptain.com/

I’d really appreciate your feedback —

Does it actually help you find issues more easily?
What can be improved or added?

Thanks for checking it out! ⚓️

https://redd.it/1ogf0rq
@r_opensource

miniLLM: MIT Licensed pretrain framework for language models

It's been a long time I haven't published anything open source (and it was really a shame for me) then I remembered how much I loved idea of nanoGPT by Andrej Karpathy. Recently, most of my pipelines and AI-backed projects however were on Qwen models so I thought to myself, what happens if I do the same thing with Qwen?

And here is MiniLLM which is working more like a "framework" for pretraining and not a standalone model itself. Although I have made a 360 million parameters model using the code which works fine (it understands English, although hallucinates a lot).

So here is the code:

https://github.com/prp-e/minillm

And I'd love to see your comments, contributions and opinions on the project.

https://redd.it/1oghy0q
@r_opensource

Fully open source peer-to-peer 4chan alternative built on IPFS
https://github.com/plebbit/5chan

https://redd.it/1ogbwjx
@r_opensource