Ooor: 1.5MB Open-Source Windows App Turns llama.cpp Into a Local AI Workstation

·5 views

Ooor is a MIT-licensed, open-source Windows desktop application that packages the llama.cpp ecosystem into a lightweight, click-to-run local AI workstation. Developed by a contributor known as oshine and available on GitHub, the app compiles to just 1.5MB using WinForms and .NET Framework 4.8, idling in single-digit megabytes of RAM. It bundles an engine manager, a GGUF model library, a Hugging Face model marketplace, a resumable downloader, a streaming chat console, and an agent tool-calling layer in a single executable. Unlike alternatives such as LM Studio, which exceeds 500MB and relies on Electron, Ooor requires no cloud services, no user accounts, and collects no telemetry. The app exposes an OpenAI-compatible HTTP endpoint at 127.0.0.1:6080, allowing direct integration with tools like Cursor, Continue, and Cherry Studio.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer Launches Technical Blog to Share Full-Stack Engineering Insights

A software engineer has launched a technical blog aimed at sharing real-world engineering challenges and architectural patterns with the developer community. The author works across a stack that includes React, TypeScript, React Native, and backend infrastructure involving database design and API integrations. The blog plans to cover topics such as full-stack architecture breakdowns, advanced TypeScript patterns, mobile app deployment with Expo, and freelance development practices. Content is intended for fellow engineers, engineering managers, and recruiters interested in practical, in-depth technical knowledge. The first technical article is expected to be published soon.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Pin Object Identity in Tests Before Extracting Mutating Functions

Software developers are advised to freeze object identity checks in tests before extracting any single mutating function from a shared module. Value-based assertions alone can miss alias bugs, since a copied list may pass item checks while silently altering the original object's order or structure. The guidance focuses specifically on in-place container mutations, recommending that developers capture object IDs of mutable arguments before and after each function call to detect unexpected identity changes. A reference table categorizes common leaf-function behaviors — such as in-place sorting or nested dict aliasing — and indicates which are safe to extract only when identity contracts remain stable. The proposal includes a draft test harness to track container IDs and mapping key changes, intended to be adapted locally before any code restructuring begins.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Termai: A Lightweight AI Terminal Agent Built with Bash and jq

A developer has built Termai, a terminal-based AI assistant using only Bash and jq, without relying on Python or Node.js. The tool connects to a large language model and supports capabilities such as web search, weather checks, URL fetching, and command execution. Tool-calling functionality was implemented using a simple case statement, where the model selects a tool, Bash executes it, and the result is returned to the model. The project demonstrates that a functional AI agent can be constructed with minimal dependencies using standard shell scripting tools.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

SignalWatch Delivers Text Diffs via Email and Webhooks Without Screenshots

A developer built SignalWatch, a hosted web-monitoring tool that detects changes on public URLs and delivers unified text diffs rather than screenshots. The service polls pages every 15 minutes on the free plan, scopes changes using CSS selectors, and filters noise through ignore-regex and minimum change-ratio settings. Alerts can be sent via email, Slack, Discord, or HMAC-signed JSON webhooks compatible with automation tools like n8n. Unlike self-hosted alternatives such as changedetection.io, SignalWatch is a fully managed solution requiring no container setup, though it does not support JavaScript rendering or login-gated pages. The free tier allows up to three monitors with no credit card required, and a live public demo tracks the Hacker News front page.

0 comments Read more at DEV Community