Show HN: Solidis – Tiny TS Redis client, no deps, for serverless https://ift.tt/byCF431

Show HN: Solidis – Tiny TS Redis client, no deps, for serverless Hey everyone! Over the past two years I threw myself back into full-time engineering with a simple goal: write code that gives back to the community. After a lot of late-night FOMO (“AI will do it all for us, right?”) and some painful production incidents, I finally turned my weekend project into an open-source library. [ What is Solidis ? ] - Super-light (< 30 KB) RESP2/RESP3 client with zero runtime deps and first-class ESM/CJS support. - Fully tree-shakable – import only the commands you need. - Written with SOLID principles & full TypeScript typings for every command. - Designed for cold-start sensitive serverless platforms (small bundle + tiny memory footprint). [ Why I built it ] 1. node-redis & ioredis pain - ESM is still an after-thought. - Hidden deadlocks on RST, vague error surfaces. - Everything gets bundled, even commands you’ll never call. 2. I refuse to add a dependency I don’t fully understand – I literally read candidates 10× before `npm i`. 3. Serverless bills love to remind me that every KB and millisecond matters. [ Key features ] - Protocols: RESP2 and RESP3 (auto-negotiation) - Bundle size: `<30 KB` (core) / `<105 KB` (full) - Dependencies: 0 - Extensibility: Drop-in command plugins, custom transactions - Reliability: Auto-reconnect, per-command timeouts, type-checked replies [ Roadmap / Help wanted ] - Benchmarks against `node-redis` & `ioredis` (PRs welcome!) - More first-class Valkey love - Fuzz-testing the parser - Docs site – the README came first; I’d love help polishing full docs This might be my last big OSS push for a while, so stars, issues, and PRs mean the world . If Solidis saves you some cold-start time or just scratches a TypeScript itch, let me know! Repo: https://github.com/vcms-io/solidis License: MIT Thanks for reading, and happy hacking! (Feel free to AMA in the comments – I’m around.) https://github.com/vcms-io/solidis May 17, 2025 at 02:50AM

Show HN: KVSplit – Run 2-3× longer contexts on Apple Silicon https://ift.tt/UopV93T

Show HN: KVSplit – Run 2-3× longer contexts on Apple Silicon I discovered that in LLM inference, keys and values in the KV cache have very different quantization sensitivities. Keys need higher precision than values to maintain quality. I patched llama.cpp to enable different bit-widths for keys vs. values on Apple Silicon. The results are surprising: - K8V4 (8-bit keys, 4-bit values): 59% memory reduction with only 0.86% perplexity loss - K4V8 (4-bit keys, 8-bit values): 59% memory reduction but 6.06% perplexity loss - The configurations use the same number of bits, but K8V4 is 7× better for quality This means you can run LLMs with 2-3× longer context on the same Mac. Memory usage scales with sequence length, so savings compound as context grows. Implementation was straightforward: 1. Added --kvq-key and --kvq-val flags to llama.cpp 2. Applied existing quantization logic separately to K and V tensors 3. Validated with perplexity metrics across context lengths 4. Used Metal for acceleration (with -mlong-calls flag to avoid vectorization issues) Benchmarked on an M4 MacBook Pro running TinyLlama with 8K context windows. Compatible with Metal/MPS and optimized for Apple Silicon. GitHub: https://ift.tt/CAYbgkl https://ift.tt/CAYbgkl May 17, 2025 at 01:34AM

Show HN: Inconveniently operating my computer with voice and hand gestures https://ift.tt/Bmzwqnf

Show HN: Inconveniently operating my computer with voice and hand gestures Introducing Iron OS: it's like a regular computer, but much more inconvenient Created with threejs, rosebud AI, web speech API, and mediapipe computer vision Any feedback would be appreciated! I've been having fun experimenting with computer vision and voice control lately. https://twitter.com/measure_plan/status/1923452731248795856 May 17, 2025 at 12:46AM

Show HN: I vibe coded an open-source Go app to back up DBs using Docker labels https://ift.tt/BaYL1us

Show HN: I vibe coded an open-source Go app to back up DBs using Docker labels Hey HN, I'm excited to share Label Backup, an open-source project I developed. This tool, written in Go, automates backups for databases running in your Docker containers using simple container labels. I built this with vibe coding. How it Works: - You add a few descriptive labels to your containers that run databases you want to back up (e.g., backup.enabled=true, backup.type=postgres, backup.cron="0 2 * * *", backup.conn=..., or backup.type=volume, backup.path=/my/data). - Label Backup, running as a separate container, automatically discovers these, then schedules and performs backups for: - PostgreSQL databases - MySQL databases - MongoDB databases - Redis data - Backups can be streamed to local storage or any S3-compatible service (such as AWS S3, MinIO, or Cloudflare R2). - Key features include on-the-fly Gzip compression, configurable retention policies, and webhook notifications for backup status. My main goal was to build something lightweight, easy to configure, and reliable, aiming for a tool that (I hope!) feels smooth and intuitive to use. You can find the source code, a detailed README, and a docker-compose.yml for a complete test environment on GitHub: https://ift.tt/mkRtHd4 The Docker Hub image is available at: resulgg/label-backup I'd love to hear your thoughts and feedback. https://ift.tt/mkRtHd4 May 16, 2025 at 07:53PM

Show HN: Easel – Code multiplayer games like singleplayer https://ift.tt/c3zLYgD

Show HN: Easel – Code multiplayer games like singleplayer For the past 3 years, I've been creating a new 2D game programming language where the multiplayer is completely automatic. The idea is that someone who doesn't even know what a "remote procedure call" is can make a multiplayer game by just setting `maxHumanPlayers=5` and it "just works". The trick is the whole game simulation, including all the concurrent threads, can be executed deterministically and snapshotted for rollback netcode. Normally when coding multiplayer you have to worry about following "the rules of multiplayer" like avoiding non-determinism, or not modifying entities your client has no authority over, but all that is just way too hard for someone who just wants to get straight into making games. So my idea was that if we put multiplayer into the fabric of the programming language, below all of your code, we can make the entire language multiplayer-safe. In Easel the entire world is hermetically sealed - there is nothing you can do to break multiplayer, which means it suits someone who just wants to make games and not learn all about networking. I've had people make multiplayer games on their first day of coding with Easel because you basically cannot go wrong. There were so many other interesting things that went into this project. It's written in Rust and compiled to WebAssembly because I think that the zero-download nature of the web is a better way of getting many people together into multiplayer games. The networking is done by relaying peer-to-peer connections through Cloudflare Calls, which means Cloudflare collates the messages and reduces the bandwidth requirements for the clients so games can have more players. I also took inspiration from my experience React when creating this language, here's how you would make a ship change color from green to red as it loses health: `with Health { ImageSprite(@ship.svg, color=(Health / MaxHealth).BlendHue(#ff6600, #66ff00)) }` There is a lot of hidden magic that makes the code snippet above work - it creates a async coroutine that loops each time Health sends a signal, and the ImageSprite has an implicit ID assigned by the compiler so it knows which one to update each time around the loop. All of this lets you work at a higher level of abstraction and, in my opinion, make code that is easier to understand. Speaking of async coroutines, my belief is that they don't get used enough in other game engines because their lifetimes are not tied to anything - you have this danger where they can outlive their entities and crash your game. In Easel each async task lives and dies with its entity, which is why we call them behaviors. Clear lifetime semantics makes it safe to use async tasks everywhere in Easel, which is why Easel games often consist of thousands of concurrently-executing behaviors. In my opinion, this untangles your code and makes it easier to understand. That's just the beginning, there is even more to talk about, it has been a long journey these past 3 years, but I will stop there for now! I hope that, even for those people who don't care about the multiplayer capabilities of Easel, they just find it an interesting proposal of how a next-generation game programming language could work. The Editor runs in your web browser and is free to play around with, so I would love to see more people try out making some games! Click the "Try it out" button to open the Sample Project and see if you can change the code to achieve the suggested tasks listed in the README. https://ift.tt/OU4rpTg May 14, 2025 at 04:01PM

Show HN: Convert JSON Schema to SQL DDL https://ift.tt/eECrLUc

Show HN: Convert JSON Schema to SQL DDL While doing research for an architectural change at work, I couldn’t find a nice npm library that let’s you create SQL tables from a JSON Schema. That’s how I decided to create one myself. https://ift.tt/e3X2LdU May 16, 2025 at 02:49AM

Show HN: AsianMOM – WebGPU Vision-LLM app that roasts you like ur mom in-browser https://ift.tt/t0ZqEgk

Show HN: AsianMOM – WebGPU Vision-LLM app that roasts you like ur mom in-browser Randomly got inspired yesterday seeing SmolVLM working on WebGPU and had the silly idea for this project. it's not perfect and super limited because of the current limitations of WebML (and admittedly, because I suck at prompting, but that's why it's Open Source haha) but it is 1.5B WORTH OF AI (SmolVLM 500M and LLama 3.2 1B) working RIGHT IN YOUR BROWSER with you not having to install anything! In fact, the whole thing is actually just an index.html that you can install and even use directly! It might be a little bit slow on first try (takes about 3 mins) when it installs models, but it caches it so it's way faster the second time (also, it's available offline after it's cached haha) Works on any modern web browser It may be a funny little project, but it's genuinely taught me so much about WebML and Vision models, and the technologies we're getting with WebML will 100% democratize AI access and make it way simpler and easier to be used everywhere :p GH Repo in case you're interested: https://ift.tt/LbtlVM8 https://ift.tt/hUnMYAB May 16, 2025 at 12:50AM