📦 diegoscarabelli/garmin-health-data

Garmin Health Data: Your Watch's Secret Database

One command pulls every metric from your Garmin watch into a single, local database file you can actually query. Garmin Health Data solves the problem of scattered health logs by downloading sleep, stress, and activity data from Garmin Connect and storing it in a local SQLite database for deep analysis. This tool is useful because it turns your wearable’s hidden data into an accessible resource without needing heavy cloud infrastructure. It also lets you manage multiple family accounts in one place. The result is a personal health archive that is fully under your control and ready for custom insights.

📰 https://news.ycombinator.com/item?id=47970512

🆔 @hackernewsgithubprojects

📦 loverain1997/h3-prompt-journal

The 3:1 Dance Trick

Turn vague dance instructions into precise, count-locked choreography with the H3 Prompt Journal. This collection documents real experiments where creators turned fuzzy adjectives like 'fast' and 'slow' into hard numbers the model can actually track. The standout technique uses a strict action-count ratio, forcing one character to move three times for every single move the other makes. This prevents the AI from defaulting to boring, synchronized motion, creating a deliberate, lagging dynamic that feels like real performance. It’s a brilliant lesson in how specificity beats description when guiding AI video tools.

🆔 @hackernewsgithubprojects

📦 tinysuitehq/tinysearch

TinySearch Cuts AI Web Research Costs

TinySearch acts as a local filter that shrinks web pages before your AI agent sees them, cutting research costs by up to sixty percent. Instead of feeding your model entire websites full of ads and navigation links, this tool crawls the internet locally and extracts only the specific evidence chunks your agent actually needs. It uses local machine learning to rank these passages without sending your data to third-party servers, keeping your setup private and free. The result is a compact, source-grounded summary that lets your agent answer questions accurately while spending significantly fewer tokens.

🆔 @hackernewsgithubprojects

📦 bookorbit/bookorbit

BookOrbit: Read Anywhere, Sync Everywhere

Start reading on your phone and finish on your e-reader without losing your place, thanks to BookOrbit. This self-hosted platform acts as a central hub for your ebooks, PDFs, and audiobooks, keeping your progress and highlights synced across the web, Kobo devices, and KOReader. Instead of juggling separate apps for each format, you get one unified library that automatically fetches metadata from fourteen different sources. It turns your scattered collection into a cohesive reading space where your entire history lives in one spot.

🆔 @hackernewsgithubprojects

📦 wecomteam/wecom-cli

wecom-cli: The Terminal Bridge for AI and Humans

wecom-cli is a Rust-powered command line tool that turns your terminal into a direct control panel for WeCom, the enterprise communication platform. The surprising part? It’s designed for both you and your AI agents. Instead of just reading data, it actively pushes messages, creates documents, manages meetings, and handles emails right from the terminal. Think of it as a universal remote for your workday that doesn’t care if a human or a script is holding it. It handles the messy details like authentication and API calls so you can focus on the task.

🆔 @hackernewsgithubprojects

📦 apache/maka

Maka: Your AI Agent's Black Box

Maka is the local-first AI workspace that treats your agent's actions like a flight recorder. Instead of just chatting, it records every message, tool call, and permission decision into an append-only log that never gets lost. This means if your AI agent crashes or you need to audit exactly what it did to your files, you have a perfect, recoverable history. It runs locally, so your data stays on your machine, and it lets you branch sessions to explore different solutions. If you want an AI agent that actually answers for its work, Maka is your new standard for accountability.

🆔 @hackernewsgithubprojects

📦 mahlernim/google-timeline-visualizer

Turn Your GPS Data Into a Travel Movie

Your phone’s background location history is quietly building a map of your entire year, and this open source app turns that data into a cinematic travel film. Google Timeline Visualizer takes the raw location file you export from your phone settings and animates your movement across a map, showing where you went and how far you traveled. The most surprising part is that it does all of this entirely on your device without sending your personal data to any cloud servers. You can select specific months or days, choose a camera style, and export a high quality video to share with friends.

🆔 @hackernewsgithubprojects

📦 mfranzon/dgx-models

dgx-models: 3D Models for DGX Cases

dgx-models is the open 3D model library that gives creators printable enclosures for high-end Nvidia hardware. The project provides twelve distinct designs, including the DGX Spark and ASUS GX10, ready for 3D printing. Each model includes detailed fan placements and airflow paths so you can actually build a case that works. It solves the problem of finding functional, aesthetically pleasing shells for expensive computing gear. The browser viewer lets you inspect every angle without downloading heavy software. If you are building custom hardware, this is your go-to resource for tested, realistic designs.

🆔 @hackernewsgithubprojects

📦 omricn/stfu

This App Punishes You For Yelling While Gaming

This Windows tray app takes a few seconds to measure your normal speaking volume, then sets a threshold sixty percent of the way up to your shout, ensuring it catches only sudden yells. When you spike, it immediately minimizes your game, plays a random sound effect, and launches a fullscreen overlay that actively resists your attempts to close it. The app is built for shared households where loud noises wake others, using a clear escalation ladder that gets more disruptive with every repeat offense. It logs every trigger for later review and processes audio locally without ever storing or transmitting recordings.

🆔 @hackernewsgithubprojects

📦 zaneham/voyager-fds-emulator

Emulating The Computer Leaving The Solar System

The same tiny computer that processed images of Jupiter and is currently drifting through interstellar space is now emulated on your laptop. The Voyager Flight Data Subsystem has been running for forty-eight years, processing data from a machine that runs slower than a dial-up modem. This project recreates that exact 1970s hardware, including its unique memory banking and lack of a multiply instruction. It proves how robust the early spacecraft computers were. Now you can run that legacy software locally and see exactly how the farthest human-made objects stay alive.

📰 https://news.ycombinator.com/item?id=49373701

🆔 @hackernewsgithubprojects

📦 adambenhassen/minimax-music-ui

Minimax Music UI: Self-Hosted Suno

Write a style prompt and some lyrics, then watch your computer turn them into a full song right on your local machine. Minimax Music UI gives you a simple web page where you can generate music locally, just like using an online service, but without sending your data to a cloud. The coolest part is that you can press play while the song is still being created. You hear the audio stream in as the computer renders it, so the waveform grows in real-time before the file is even finished. It keeps everything private and organized in a local library for you to revisit.

🆔 @hackernewsgithubprojects

📦 adobe-research/tpips

TPIPS: The AI That Knows Why Photos Look Different

TPIPS is the image comparison tool that finally lets you ask what specific part of a photo makes it look different. Most tools just guess how similar two images are overall, but TPIPS lets you type in a clue like lighting, pose, or even the number of ducks. It then measures similarity strictly for that one thing. This is huge because it turns a vague feeling of difference into a precise, testable number. It helps researchers and creators understand exactly where a visual change matters, turning messy human perception into clean data you can actually use in your projects.

🆔 @hackernewsgithubprojects

📦 stanlukuvka/comfyui-minimax-h3-speed

Make MiniMax H3 Video Generation Faster with SPEED Sampling

ComfyUI MiniMax H3 SPEED is the ComfyUI sampler that cuts video generation time by rendering low resolution first. Instead of churning through twenty full resolution steps, it builds the basic structure at a fraction of the size, then steps up to full quality. This saves massive amounts of compute and memory without changing how you use ComfyUI. The audio stays crisp while the video speeds up. It is a work in progress, but the core idea is proven and the setup is simple. Give your GPU a break and generate faster.

🆔 @hackernewsgithubprojects

📦 albertobeiz/kindergrimm

Kindergrimm: Procedural Doodle Characters

Kindergrimm procedurally generates unique doodle characters that look like hand-drawn sketches on cream paper, complete with a built-in game where they come alive. The system treats every face as a simple recipe, meaning you can save, share, or rebuild any specific character at any time without losing the original look. This is fascinating because it turns complex art into reproducible data, allowing for endless variety while keeping every drawing consistent and animatable. The result is a playful blend of procedural art and interactive storytelling that feels surprisingly personal. If you enjoy creative coding, this project offers a beautiful look at how randomization can mimic human whimsy in digital art.

📰 https://news.ycombinator.com/item?id=49346080

🆔 @hackernewsgithubprojects

📦 tealios/errata

Errata: Write Stories with AI

Errata turns your scattered notes into a living story engine. You can chat directly with your characters, letting them voice their thoughts in real time, while the app keeps track of every detail you have written. It builds a memory for your narrative so the AI never forgets a plot point or a character trait. You can even see exactly what prompt is being sent to the model before it writes, giving you total control over the output. Because it runs locally on your computer, your work stays private and safe. It is a powerful tool for anyone who wants to create long-form fiction without losing their thread.

🆔 @hackernewsgithubprojects

📦 s1dashu/ip-as-logo-skill

The AI Skill for Instant Cute Logos

ip-as-logo-skill is the compact agent tool that finally makes cute brand mascots easy. It guides AI agents to generate extremely simple, rounded character images that pop instantly. The magic is its strict complexity budget, forcing designs into just a few basic shapes and three colors. This ensures the result looks great even as a tiny app icon. It skips the usual messy branding process by handing you six clean, high-quality options at once. Perfect for developers who need a friendly face for their product fast. Just tell your agent to use the skill, approve the direction, and grab your new mascot.

🆔 @hackernewsgithubprojects

📦 mothparkzo6249/tg-polymarket-bot

Tg Polymarket Bot: One Tap Whale Copying

Tg Polymarket Bot is the Telegram tool that lets you copy whale trades in a single tap. It plugs directly into Polymarket's live feed, spotting big bets before the price moves, so you aren't stuck hunting for signals. The killer feature is how it handles entry: the wallet lives inside the bot, so when an alert pops up, you tap one button and your bet signs instantly. No waiting for clunky popups or fighting laggy interfaces that often drop trades. It even lets you mirror specific traders automatically, keeping your keys locked behind a personal PIN.

🆔 @hackernewsgithubprojects

📦 lbh-123-ai/comfyui_minimax_h3_latent_upscaler

ComfyUI Minimax H3 Latent Upscaler: Skip the Slow VAE

Minimax H3 video generation usually stalls on a massive five-billion parameter decode step that eats your time, but this tool cuts straight through that bottleneck by working entirely in latent space. Instead of the slow, heavy cycle of decoding low-res pixels, stretching them, and re-encoding them, this neural node upscales the compressed data directly. It skips the expensive math and the annoying ghosting artifacts that simple stretching creates. You just generate a rough draft, let this node sharpen the latent data, and refine it to high resolution. It is a pure speed win for high-quality video creation, saving you hours without changing how much memory you use.

🆔 @hackernewsgithubprojects

📦 hybroai/a2a-adapter

A2A Adapter: The Universal Translator for AI Agents

A2A Adapter is the bridge that finally lets different AI agents talk to each other without custom code. If you are running Claude, LangGraph, or a custom Python function, this tool wraps them into a standard protocol server in just a few lines. It handles the complex networking and task management for you, so you only need to worry about your logic. The coolest part is that it turns any agent into a compatible service instantly, allowing your various tools to collaborate as if they were one system. It is the missing link for anyone building multi-agent workflows who wants simple, open standards instead of messy integrations.

🆔 @hackernewsgithubprojects

📦 jithin-sabu/purge-app

Purge App Keeps Your Mac Safe

Purge is the macOS cleaner that refuses to delete anything permanently. It scans for the hidden cache and junk your computer collects, labeling each item so you know exactly what is safe to remove. The biggest surprise is that it treats cleanup as reversible; everything moves to your Trash, not into the void. This open source tool also recognizes developer folders, so it can free up space from code projects without breaking your work. Because the deletion logic is public and the app runs locally, you have full control over your data.

🆔 @hackernewsgithubprojects