📦 markovwangrr/iso-topology

iso-topology

You can now generate design-grade, three-dimensional isometric architecture diagrams straight from plain text files, eliminating the need to spend hours manually drawing boxes and aligning grids in Figma. This tool, called iso-topology, translates coordinate-free text layouts into gorgeous, two-and-a-half-D vector graphics that represent complex systems with clean visual depth. It solves the headache of outdated documentation by keeping your system diagrams committed directly alongside your codebase as plain, version-controlled text. The project even includes a built-in feedback loop so AI agents can automatically write, validate, and repair the diagrams before they render. It is the easiest way to keep your technical drawings beautiful and always up to date.

🆔 @hackernewsgithubprojects

📦 arbs-io/github-copilot-llm-gateway

github-copilot-llm-gateway

Your self-hosted AI models can now run inside GitHub Copilot Chat without constantly choking on broken tool calls or context limits. The github-copilot-llm-gateway is a clever VS Code extension that adds a robust resilience layer to smooth out the rough edges of local models like Qwen, Llama, and Mistral. Instead of failing when small or quantized models output sloppy JSON, this companion extension actively repairs malformed tool calls, manages context budgeting, and routes raw reasoning blocks perfectly into Copilot's interface. It gives your self-hosted setup the ultimate stability boost, bringing flawless agentic coding right to your local network.

🆔 @hackernewsgithubprojects

📦 manak-hash/linkbreeze

LinkBreeze

LinkBreeze is the self-hosted link-in-bio platform that lets you bypass monthly subscription fees while keeping full ownership of your data. Think of it as a private, open-source alternative to Linktree that you can launch on your own server with a single Docker command. It packs a fast, built-in admin panel where you can drag and drop unlimited links, schedule when they appear, embed media widgets from Spotify or YouTube, and collect subscriber emails. Best of all, it gives you privacy-first page analytics and customizable themes with zero tracking cookies, loading your personal page in under three hundred milliseconds.

🆔 @hackernewsgithubprojects

📦 loganbooker/sesquilsr

SesquiLSR

Standard latent upscaling destroys image detail, forcing you to use massive denoise settings just to fix the blurry mess. A clever open-source project called sesquilsr solves this by using a tiny three-million parameter model that upscales directly inside the latent space. Instead of relying on crude math, it uses learned downsampling to scale your images to any fractional size up to two-times with incredible perceptual detail. It runs in under six milliseconds, bypasses slow round-trips to the image decoder entirely, and plugs right into ComfyUI. It is the perfect, lightning-fast bridge to prep your AI generations for a clean second pass.

🆔 @hackernewsgithubprojects

📦 lagunaswift/rockyvoice

RockyVoice

Transform your everyday AI assistant into Rocky, the beloved Eridian alien engineer from Andy Weir's sci-fi novel Project Hail Mary. This clever repository lets you install a custom persona skill for Claude Code and Hermes, replacing standard responses with Rocky's signature speech style of small words, no contractions, tripled words for big feelings, and questions that always end with the word question. Even better, it includes a local web application powered by Hume AI voice cloning, letting you hear Rocky speak your code fixes and answers out loud in real time. It is a brilliant way to bring your favorite space friend directly into your terminal.

🆔 @hackernewsgithubprojects

📦 thecodearcher/limen

Limen

Compose your perfect authentication stack in Go without any bloat. Limen is a modular authentication library designed to let you plug in only the security features you actually need, from passwords and OAuth to two-factor auth. Unlike traditional all-in-one libraries, its architecture isolates every single authentication method into its own separate Go package. This means your application stays incredibly lightweight because you never compile unused dependencies into your final binary. It even includes reactive TypeScript client SDKs for popular frontend frameworks. It is a wonderfully clean and modern approach to securing your Go applications.

📰 https://news.ycombinator.com/item?id=48823067

🆔 @hackernewsgithubprojects

📦 seu-paisys/embodied.cpp

embodied.cpp

Run heavy robotic brain models directly on everyday hardware without losing speed. If you want to deploy advanced AI models that let robots see, plan, and physically act in the real world, you usually need massive cloud setups. This project changes that by packaging these complex models into highly efficient files that run natively on standard computer processors and local graphics cards. It handles the translation between camera sensors and robot joints in lightweight code. You can test your models in simulation or load them straight onto a physical robot arm, bringing local, intelligent automation to standard hardware.

🆔 @hackernewsgithubprojects

📦 brobert-in-aus/scail-auto-extend

scail-auto-extend

scail-auto-extend is the ComfyUI custom node that finally lets you generate SCAIL-2 videos of any length in a single queue without manual math. Normally, the model limits you to eighty-one frames, forcing you to manually build duplicate node sections, bypass frames, and math out overlap anchors just to extend a clip. This clever node automates the entire loop at runtime by slicing your video into perfect chunks, anchoring the overlaps, matching colors to prevent drift, and stitching the final video together seamlessly. It turns a frustrating multi-step bottleneck into a simple, one-click background process. Give it a spin to effortlessly scale up your video generations.

🆔 @hackernewsgithubprojects

📦 haitaowutju/brainjanus

BrainJanus

BrainJanus is the neural translation model that finally links brain activity, vision, and language together in a single system. By mapping complex brain signals like fMRI scans into a shared digital vocabulary, this project allows a computer to translate what you are looking at directly into images or text, and even run the process completely in reverse. Instead of needing different custom software for every individual task, this setup handles visual reconstruction, brain captioning, and brain signal synthesis all at once. It makes translating complex neural activity into everyday human communication and pictures incredibly simple and unified.

🆔 @hackernewsgithubprojects

📦 almanaccode/codealmanac

codealmanac

Build a living, breathing reference manual for your AI coding agents to stop them from making the same silly mistakes over and over. This project helps you create a dedicated local wiki that captures critical context your raw code leaves out, like architectural decisions, hidden gotchas, and complex business flows. By feeding this structural map directly to your AI assistants, you prevent them from hallucinating or breaking delicate parts of your system. It is a brilliant way to keep your autonomous developers fully aligned with your real-world constraints. Give your AI agents the missing instructions they need to write smarter code.

🆔 @hackernewsgithubprojects

📦 carloluisito/omnidesk

OmniDesk AI Terminal

OmniDesk is an elegant desktop terminal that wraps artificial intelligence coding tools like Claude Code directly into a high-performance workspace. Instead of juggling dozens of messy terminal windows, this app organizes everything around your actual coding repositories, giving you a dedicated activity bar and a session rail to navigate. You can instantly switch between a single focused terminal or a live grid view that displays all your active sessions side by side. It even manages your git worktrees behind the scenes so you can safely test changes. It is the ultimate command center for developers who want to run multiple local AI sessions without the chaos.

🆔 @hackernewsgithubprojects

📦 jeakwon/ai-engram

ai-engram

ai-engram is the Python library that finally makes LLM unlearning as simple as subtracting a matrix. Instead of running expensive training loops or backpropagation to make a model forget specific facts, this tool uses a clever, training-free approach based on covariance. It isolates the exact weights responsible for the target data and subtracts them in a single, closed-form step. You just collect forward-pass statistics on what to forget and what to keep, compute the difference, and the model instantly unlearns the target info while preserving everything else. It is incredibly fast, native to Hugging Face, and completely bypasses gradient descent.

🆔 @hackernewsgithubprojects

📦 omarh-creator/autonomous-forge

Autonomous Forge

Run safe, repository-native autonomous software-improvement loops using an intelligent Python command-line tool that lets you plan and review code changes without any risk of runaway AI edits. Autonomous Forge acts as a highly cautious local-first assistant that parses your roadmap, checks repository policies, and generates detailed change proposals, all within a strictly read-only safety boundary. The single coolest thing it does is compile a complete review artifact that bundles the selected task, policy compliance checks, planned patch areas, and validation steps into a deterministic JSON handoff before a single line of code is actually written. This gives human developers absolute control and visibility, ensuring that autonomous maintenance remains entirely transparent...

🆔 @hackernewsgithubprojects

📦 tsinghua-fib-lab/omniscientist

OmniScientist

Artificial intelligence is transitioning from a basic writing assistant into a fully automated researcher capable of proposing its own scientific hypotheses. This is the core vision behind omniscientist, an open research ecosystem developed by Tsinghua University and the Zhongguancun Academy. The project connects large language models to scientific concept networks, allowing them to spot hidden connections between theories and propose genuinely novel research ideas. Beyond brainstorming, it automates experiment design by searching for codebases and datasets to build real testing plans. It is a fascinating glimpse into how human and machine researchers might soon collaborate to solve complex global challenges.

🆔 @hackernewsgithubprojects

📦 openbmb/minicpm-desk-pet

MiniCPM Desk Pet

MiniCPM-Desk-Pet is the desktop companion that brings a local AI assistant right to your screen as an animated pet. After a simple guided setup, the app downloads a lightweight language model that runs entirely on your own computer, so your conversations stay private and offline. Instead of just sitting there, this little pet actively monitors your coding workspace. It automatically notices when tools like Cursor or Claude Code are busy, displays custom animations to match their progress, and even rings a bell when an agent is waiting for your feedback. It is a brilliant way to make local AI feel alive.

📰 https://news.ycombinator.com/item?id=48824849

🆔 @hackernewsgithubprojects

📦 microsoft/healthagentbench

HealthAgentBench

Drop your AI agents into a terminal environment and see if they can actually survive real-world healthcare challenges. With healthagentbench, you can put frontier AI models to the test across fifty-four intense clinical and biomedical scenarios. We are talking about critical tasks like scanning chest X-rays, flagging errors in messy electronic health records, and matching patients to life-saving clinical trials. This is not a simple multiple-choice exam; the agents must inspect raw medical data, use actual terminal tools, and solve complex problems under realistic constraints. It turns out medicine is incredibly tough, as even the smartest models only solve a minority of these tasks. Grab the code and see how...

🆔 @hackernewsgithubprojects

📦 weiyixue999/signal_structure_aware_gaussian

Signal Structure-Aware Gaussian Splatting

Reconstructing massive 3D environments from photos usually crushes consumer hardware under the weight of massive image files and runaway memory usage. A new framework called signal_structure_aware_gaussian changes that by dynamically adjusting image resolution during training to match how stable the 3D scene is at any moment. It starts with lower-detail images to build a coarse layout, partitions the scene into blocks, and only dials up the resolution when the local geometry stabilizes. This smart scheduling prevents the system from creating redundant 3D points too early. The result is a highly efficient pipeline that lets developers reconstruct large-scale scenes on standard hardware in just over an hour.

🆔 @hackernewsgithubprojects

📦 xcltql666/omintryon

OmniTryOn

OmniTryOn is the video-generating tool that finally lets you swap outfits on moving people in a single step. Instead of spending hours meticulously editing frames or trying to warp flat images of clothes onto a moving body, this clever tool lets you virtually dress anyone in a video with absolutely any garment. It works by treating your clothing items as prompts, injecting them directly into the video generation process to keep everything looking natural, even when the person moves around. It is an incredibly smart way to see how clothes actually move and fit on a real, walking body before making a purchase.

🆔 @hackernewsgithubprojects

📦 sunrainyg/randopt

RandOpt

Pretrained language models are already surrounded by a dense thicket of highly specialized task experts just waiting to be uncovered through random noise. This is the core discovery behind randopt, an open-source codebase that introduces a surprisingly simple way to post-train AI models. Instead of relying on complex, mathematically heavy optimization steps, this approach samples random perturbations around pretrained weights to find highly diverse, top-performing task experts. It turns out that a single random step actually has a high chance of outperforming a traditional gradient update on strong models. Once these diverse experts are discovered, they can be distilled back into a single powerful model. It is a fascinating rethink...

🆔 @hackernewsgithubprojects

📦 26worldcup/26worldcup.github.io

26worldcup.github.io

The 26worldcup.github.io companion is the ultimate fan-built web app that keeps you ahead of every kick-off without any of the typical internet clutter. It serves up a beautifully clean schedule, group standings, real-time weather forecasts, and broadcasting guides for thirty-two countries. But the best part is the built-in tournament simulator. Driven by an Elo mathematical model built on forty-nine thousand historic international matches, you can instantly simulate the entire tournament up to ten thousand times to see exactly how your favorite team is predicted to finish. It runs fully offline, contains zero ads, and is translated into twenty-three languages. Check out this incredibly smart tool before the tournament begins.

🆔 @hackernewsgithubprojects