Open source CLI + Python library for using LLMs in noscripts and automation

Most LLM tools I tried assume an interactive UI or a notebook session.

I wanted something that works the way other developer tools do: composable, noscriptable, and usable from the terminal or inside real programs. This project is an open source CLI and Python library that lets you treat LLMs as building blocks you can wire into workflows. In practice, it lets you:

run prompts directly from the command line
call models from Python without wrapping everything in a chat abstraction
store prompts, responses, and embeddings in SQLite for reuse and inspection
swap between hosted APIs and local models with the same interface
extend behavior using plugins and local tools
chain prompts and results together in noscripts or automation

It’s designed to fit into things like:

cron jobs and background tasks
small developer utilities
data processing pipelines
experiments that need persistence instead of one-off prompts

The goal isn’t to be a platform or UI, but to make LLM usage feel closer to working with other Unix-style tools.

Repo:

https://github.com/simonw/llm

Disclosure: this is my project.



https://redd.it/1pvr1s6
@r_opensource

I’m building an open source blood sugar tracker, what do existing apps get wrong?
https://github.com/Burnsedia/dracula

https://redd.it/1pvsl6j
@r_opensource

Corporate open source is just free QA labor with extra steps

I've been contributing to open source projects for years and I'm getting pretty cynical about how corporations actually use them. They open source a tool and call it "community driven" but then completely ignore any feedback unless it matches what they already planned to do anyway. Try suggesting a feature or change and you get hit with "that's not our vision" which really just means never.

But they have no problem accepting bug fixes and documentation improvements from volunteers. So basically they get free testing and writing while keeping complete control over everything. Saw this recently where a company fired their entire community team but still expects volunteers to help users and write tutorials. All the work with none of the influence on decisions.

If you're going to open source something just be honest about whether you actually want community input or if it's just open source in name only.

https://redd.it/1pvub9s
@r_opensource

[Project Showcase] Git Rewind: Your GitHub year in code
https://git-rewind.dev

https://redd.it/1pvumt6
@r_opensource

Ironclad POD

Hi, I made this fully web based POS app for small businesses. How is it?

This is my repo: https://github.com/mashiyathilias-code/Ironclad-POS

https://redd.it/1pvyai4
@r_opensource

I’m building a simple open-source archive format focused on long-term readability (ADC)

I’m working on a hobby open-source project called ADC (ArchivedDataCodec), a lightweight archiver with a strong focus on simplicity, transparency and long-term readability.

The motivation behind ADC is pretty simple:
I really miss archive formats that are easy to understand, easy to inspect, and don’t feel over-engineered or opaque. ADC uses a documented, straightforward format (8-byte header + compressed file blocks) and aims to stay readable even years down the line.

Key points:

Open-source > GPLv3
Made in Python with zlib
Custom, documented archive format
Multiple files per archive
Focus on clarity over cleverness
Linux-first, but cross-platform

This is very much a hobby project, but it’s actively maintained and still evolving.
If you’re into:

simple tools

open formats

learning through open source

or just reviewing weird archive ideas 😄

feedback and contributions are very welcome. Even comments or criticism are appreciated.

Github:
https://github.com/Mealman1551/ArchivedDataCodec

Thanks for reading.

ps.

My intentions are not to develop an industry standard but just a hobby project

https://redd.it/1pw1aqo
@r_opensource

built an open-source AI video clipper to replace paid subnoscriptions (Opus/Munch). Includes auto-upload to socials.

Hi everyone, I wanted to share a tool I created to liberate my workflow from paid monthly subnoscriptions.

It uses AI to parse long-form videos into short clips and utilizes APIs to upload directly to TikTok and Reels. It's extremely cheap to run (using your own API keys) compared to the $30/mo alternatives.

Code is available here: https://github.com/mutonby/openshorts .

Contributions are welcome!

#

https://redd.it/1pw51w7
@r_opensource

How to verify open source?

One of the advantages of open source is transparency. But, how do you know that the binary being used by the consumer is actually the same code as the code on GitHub? For example, Signal the messenger has their code as a public repository on GitHub. But, how do you know the binary submitted to the App Store for iOS is using this very code? I don't think you can compare the hashes of the repo and the deployed binary since the compiled code from the repo will have different code embedded during the build.

https://redd.it/1pw4xxw
@r_opensource

I built a small open-source React component for Instagram/TikTok-style vertical video feeds

Hey everyone,

I wanted to share a small open-source project I’ve been working on: react-riyils.

I built this because I couldn’t find a clean, focused, and customizable open-source solution for Instagram/TikTok-style vertical video experiences on the web that handled real browser constraints properly.

The goal was to keep things simple and correct rather than flashy or over-engineered.

The project focuses on:
- Vertical swipe navigation
- Fullscreen video viewer
- Deterministic playback behavior
- Smart video preloading and memory management
- Mobile and desktop support
- TypeScript-first, readable code

This is a hobby project, not production-scale TikTok infrastructure, but I tried to keep the code small and understandable.

Feedback, issues, and PRs are very welcome.

GitHub: https://github.com/illegal-instruction-co/react-riyils
npm: https://www.npmjs.com/package/react-riyils

Thanks for taking a look.


https://redd.it/1pw8csy
@r_opensource

Inkscape 1.4.3 is out!

Inkscape 1.4.3 is out! Update to benefit from over 120 bug and crash fixes and more than 20 improved translations.

If you don't know Inkscape, it's a free and open-source vector design application.

To learn more, visit

https://inkscape.org/news/2025/12/26/bugs-banished-inkscape-143-is-out/

https://redd.it/1pw9up2
@r_opensource

px: Immutable Python environments (alpha)

px (Python eXact) is an experimental CLI for managing Python dependencies using immutable, content-addressed environment profiles (no venv).

👉 https://github.com/ck-zhang/px

It is now in alpha, feedback is welcome :)


https://redd.it/1pwau8r
@r_opensource

(near) Christmas release: Lidify, a self hosted music app with vibe matching and Discover Weekly

Merry Christmas everyone.

I've been self hosting for about 2 years now. Nextcloud, Immich, Plex, Audiobookshelf, all that. Audio was the only thing that actively disappoints me. Jellyfin and Plex are OK for music but Jellyfin is finnicky AF and the Plex app for some reason doesn't send a keep-awake signal when listening to music so my TV will shut off. Just frustration after frustration.

I've seen tons of posts on here asking for a FOSS music app like Spotify and have searched for that myself. Lidify is my answer to that. And yes, I regret the name since this turned into much more than a Lidarr frontend. Here's what's available now (with bugs I'm sure):

* **Vibe System:** You ever listen to a song that just hits different in the moment? You don't know exactly what it is, you just know you're absolutely feeling it and then go on a journey trying to find other songs that give you that same feeling but come up empty? The vibe system is my solution to that. It analyzes tracks using data from MusicBrainz and [Last.fm](http://Last.fm) paired with ML analysis through Essentia. It'll be a forever work in progress but it works pretty well already.
* **Made For You playlists** (era mixes, genre mixes, rediscover tracks you haven't played in a while)
* **Discover Weekly** that actually downloads recommendations if you have Lidarr and/or Soulseek set up
* **Spotify/Deezer playlist import** (also dependent on Lidarr + Soulseek)
* **Podcasts via RSS**
* **Audiobookshelf integration**
* **Multi user with 2FA**

PWA works on mobile, native app coming later.

This is a passion project I built for myself but I'd love input and feature ideas from everyone. GPL-3.0, so fork it, break it, make it your own.

[https://github.com/Chevron7Locked/lidify/](https://github.com/Chevron7Locked/lidify/)


Screenshots in repo:

[https://github.com/Chevron7Locked/lidify/blob/main/assets/screenshots/desktop-home.png](https://github.com/Chevron7Locked/lidify/blob/main/assets/screenshots/desktop-home.png)

https://redd.it/1pwcl4a
@r_opensource

RapydShare - Sharing files between PC and Mobile devices made easy.
https://github.com/Overresurrect/RapydShare

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

Alternative client for WhatsApp on windows?

The WhatsApp client app on windows is really bad. It was UWP before, now it's electron and it's worse in almost every way. They released a half baked app. The web version is no good either. Is there any good alternative?

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

WebCC: A C++ framework and toolchain that batches API calls to reduce WASM/JS overhead

Hi!

I’ve been working on **WebCC**, a project that combines a C++ framework with a custom toolchain to build lightweight WASM apps.

**Repo:** [https://github.com/io-eric/webcc](https://github.com/io-eric/webcc)
**Demos:** [https://io-eric.github.io/webcc/](https://io-eric.github.io/webcc/)

**The Concept: Efficient Web API Mapping**

WebCC provides a clean C++ interface for standard Web APIs like **DOM, Canvas, WebGL, WebGPU, WebSockets, Audio, etc.**.

Under the hood, these APIs are defined in a concise schema.def file. The toolchain uses this definition to automatically generate:

1. **C++ Headers:** Type-safe APIs for your application (e.g., `webcc::dom::create_element`).
2. **JavaScript Runtime:** The exact glue code needed to execute those commands.

This approach ensures that the generated code is always in sync and minimal. While it comes with a comprehensive set of standard APIs, it's also easily extensible, adding a new browser feature is as simple as adding a line to the schema.

**Smart Compilation & Tree Shaking**

The toolchain scans your C++ source code to detect exactly which API functions you are using. It then generates a **custom, tree-shaken** `app.js` containing *only* the JS implementation for those specific functions. If you don't use Audio, the Audio glue code is never generated.

**Architecture**

WebCC implements a specific architecture to optimize the C++/JS relationship:

* **Command Batching:** Void functions (like drawing commands) are serialized into a binary buffer and executed in bulk, reducing the frequency of boundary crossings.
* **Zero-Copy Event System:** Instead of JS calling C++ for every mouse move or key press, JS writes events directly into a shared WASM memory buffer using **TypedArrays** (`Uint8Array`, `Int32Array`). The C++ app simply polls this buffer.
* **Hybrid Execution:** Functions that return values (like `create_element`) automatically flush the buffer and execute synchronously, ensuring correct order without manual management.
* **String Interning:** Strings are cached per-frame. If you use the same color string 1000 times, it only crosses the boundary once.
* **Integer Handles:** Resources are managed via integer IDs to avoid object passing overhead.

Example:
**1. Define in Schema (Internal Definition):**

`# Namespace | Type | Name | C++ Func | Args | JS Implementation`
`canvas|command|FILL_RECT|fill_rect|int32:handle float32:x float32:y float32:w float32:h|{ const ctx = contexts[handle]; ctx.fillRect(x, y, w, h); }`
**2. Use in C++ (Developer Code):**

// These calls are buffered
webcc::canvas::set_fill_style(ctx, 255, 0, 0);
webcc::canvas::fill_rect(ctx, 10, 10, 100, 100);

// Flush the buffer to execute all commands in JS
webcc::flush();

**Benchmarks:**
Benchmarks comparing WebCC to Emnoscripten in a test rendering 10,000 rectangles with Canvas 2D

[https://github.com/io-eric/webcc/tree/main/benchmark](https://github.com/io-eric/webcc/tree/main/benchmark)

=== BENCHMARK RESULTS ===
Browser: Chrome 142.0.0.0
Metric | WebCC | Emnoscripten
--------------------------------------------------------
WASM Size (KB) | 11.25 | 150.51
JS Size (KB) | 11.03 | 80.54
FPS | 100.37 | 40.18
JS Heap (MB) | 9.03 | 24.62
WASM Heap (MB) | 2.38 | 16.12


Thanks for reading! Questions and feedback are always welcome. :)

https://redd.it/1pwtv3w
@r_opensource

Lightweight FaaS with Golang and Podman
https://github.com/pardnchiu/go-faas

https://redd.it/1pwvb98
@r_opensource

Whats a self hosted opensource alternative to Jira ?

Whats a self hosted opensource alternative to Jira ? can be docker.

is there any other recommendations that anyone can make

https://redd.it/1pwvsc6
@r_opensource

witr (Why Is This Running?) – tracing process origins on Linux
https://github.com/pranshuparmar/witr

https://redd.it/1pwwty5
@r_opensource

Made a small tool in PHP for handling texts in images better. Thoughts?

Hi, opensourcers! A year ago i needed something to generate images with text in them, but i wanted it so my code is more clean and easier to understand than copy and destroy every time i wanted to put a simple text. More specifically, i wanted so i am able to read my own text in the code that handles generation. I remembered i used AI a little bit for the regex specifically (regex was always a complicated thing for me no matter what i did), but this is otherwise good.

Now i decided to make this open-source, and maybe someone finds a use of it. https://github.com/Wreeper/imageworkout/

I know it's not the best piece of code, but it did what i wanted and it continues to do what i wanted it to do.

I do want to hear your thoughts on it. I'm not a great dev at all, but at least i'm trying.

https://redd.it/1pwyhr3
@r_opensource

I made this DevType - Master Your Coding Speed

Try : https://devtype.pankajk.tech/
Repo : https://github.com/pankajkumardev/devtype

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

I made an pollution-reporting app for the folks of Loudoun County, Virginia

Github: https://github.com/ackshatiwari/PolluTrack

https://redd.it/1px17pf
@r_opensource