#typescript #dsh_plugin

billion-context helps AI coding agents manage long conversations by compressing older messages into layered summaries before sending them to a model. The model can search summaries or restore original details when needed. It works with many coding clients and supports Anthropic and OpenAI-style APIs. This can reduce token use and help sessions continue without exceeding the model’s context limit. It also provides a web dashboard for viewing sessions, usage, logs, and settings.

https://github.com/ranxianglei/billion-context

#kotlin #android #ikev2 #java #kotlin #openvpn #protonvpn #strongswan #vpn

ProtonVPN for Android is an open-source VPN app for Android devices. It provides a way to use ProtonVPN on a phone or tablet. Its source code is available under the GPLv3 license, which is useful for people who want to inspect or contribute to the app.

https://github.com/ProtonVPN/android-app

#css #dark_mode #dark_theme #light_theme #obsidian #obsidian_md #obsidian_theme #theme

Minimal is a theme for Obsidian on desktop, mobile, and tablet. It lets users change the look of their workspace with color schemes, fonts, background styles, and layout options. Focus mode hides parts of the interface, while image grids, card layouts, and table styles help organize visual content. Companion plugins provide more customization, including options to hide interface elements.

https://github.com/kepano/obsidian-minimal

#python #ai #claude #hacs #home_assistant #home_automation #llm #mcp #model_context_protocol

HA-MCP connects AI assistants such as Claude and ChatGPT to Home Assistant, letting users manage their smart home through everyday language. It can control devices, create and edit automations, scripts and dashboards, check device history, and help diagnose problems. It also supports backups and safety controls such as read-only access and approval policies. This makes it easier to configure and maintain a smart home without doing every task by hand.

https://github.com/homeassistant-ai/ha-mcp

#typescript

GBrain gives AI agents a shared memory that users can control. It stores facts with their sources, supports corrections, and can search notes and other connected information. Optional features can produce answers with citations, identify gaps in knowledge, and maintain information in the background. This helps agents keep useful context across tasks and makes it easier to find and use stored knowledge. GBrain is free to self-host, with paid hosting available.

https://github.com/garrytan/gbrain

#ruby

Sure is a personal finance app that helps people view their cash flow, net worth, and spending in one place. It can also use an AI assistant to answer questions about savings. Users can host Sure themselves and access it through a browser, desktop or mobile app, API clients, and AI agents. It is a community-maintained fork of Maybe Finance.

https://github.com/we-promise/sure

#typescript

Modern Web Guidance gives AI coding agents expert guidance on modern web features, browser support, and best practices. It searches and retrieves guides on topics such as layouts, accessibility, performance, and security, and can scan code for outdated patterns. It helps agents use built-in browser features instead of older, more complex solutions, while choosing sensible alternatives for browsers that lack support. Experimental features require explicit permission before their guides can be retrieved.

https://github.com/GoogleChrome/modern-web-guidance-src

#other

improve is an agent skill that reviews codebases and writes detailed plans for fixing problems or adding features. It checks areas such as security, performance, tests, and documentation, then ranks findings so users can choose what to address. Its plans include context, steps, checks, and limits, making them easier for another agent or a person to follow. It can also send a plan to a cheaper agent, review the result, and keep the plan backlog up to date. This helps users benefit from a capable model’s judgment while using cheaper models for implementation.

https://github.com/shadcn/improve

#typescript #ai_security #application_security #cli #code_scanning #codex #codex_security #cybersecurity #devsecops #nodejs #npm #openai #sdk #security #typescript #vulnerability_scanning

Codex Security helps find, investigate, and fix security weaknesses in code. It can scan entire repositories, selected files, or Git changes, check findings, suggest patches, and verify fixes. It also helps teams assess severity, identify duplicate findings, draft security policies, and save threat models. Teams can use it in Codex or automate scans through a command-line tool, TypeScript SDK, or GitHub Action. It can also export results and track findings in tools such as Jira, Linear, and GitHub.

https://github.com/openai/codex-security

#python #bert #chinese #dataset #ernie #gpt #gpt_2 #large_language_models #llm #multimodel #nezha #nlp #nlu_nlg #pangu #pretrained_models #roberta #simbert #xlnet

Awesome Pretrained Chinese NLP Models is a collection of publicly available Chinese language and multimodal models. It organizes models by purpose, including text understanding, text generation, conversation, reasoning, and specialist fields such as medicine and coding. It also lists evaluation benchmarks, datasets, and model platforms. The collection helps researchers and developers find and compare resources for working with Chinese language AI.

https://github.com/lonePatient/awesome-pretrained-chinese-nlp-models

#objective_c

HermesEnabler 2 is an Apple Watch app for watchOS 27.0 through 27.3 beta 1. It guides users through steps on the watch and can restore a MobileGestalt backup. This backup is useful for restoring data before deleting the app.

https://github.com/pookjw/HermesEnabler

#zig #ai #ai_tools #llm #llm_inference #llm_tools

TensorFold runs language models locally on Apple Silicon Macs and NVIDIA GPUs without requiring Python or MLX for serving. It downloads and manages supported models and provides an API for chat and other requests. A local web app lets users pick models and chat with them, while a Mac desktop app adds model downloads, generation metrics, coding tools, and document editing. This makes it useful for running and using AI models on local hardware.

https://github.com/ashhart/TensorFold

#typescript #ag_ui #agent_governance #ai_agents #browser_automation #copilotkit #generative_ui #mcp

OpenBot is a workspace for using AI assistants on your own computer or infrastructure. You can choose AI models, use built-in assistants, or connect your own agents. Assistants can browse the web, work with files and connected apps, and run scheduled tasks. You can watch their work, take control when needed, and set permissions and approvals. This helps individuals and teams automate work while keeping control over what their assistants can do.

https://github.com/CopilotKit/OpenBot

#python

Claude Trading Skills is an open-source toolkit that uses Claude to help individual investors organize their trading and investing process. It supports market and portfolio reviews, risk management, trade planning, journaling, and performance reviews. It is designed for people who focus on long-term investments, ETFs, and dividend stocks but may also use swing trading. Its workflows and checklists help users make more disciplined decisions and learn from past trades. It does not make buy or sell decisions for users or promise profits.

https://github.com/tradermonty/claude-trading-skills

#typescript #agents #claude_code #code_review #codex #obsidian #opencode #pi_mono #plan_mode #skills

Plannotator is a local, browser-based tool for reviewing work from AI coding agents. It lets users annotate plans, documents, messages, and HTML, review code changes and pull requests, and send feedback directly to an agent. Its inbox brings together questions, files, and review requests from multiple agents while they continue working. This helps users stay involved in agent work, make decisions, and give clear feedback. Plans, annotations, and other review data stay local by default.

https://github.com/backnotprop/plannotator

#cplusplus

Veyra is a Windows 11 application that uses NVIDIA or AMD graphics cards to improve video in real time. It can sharpen and upscale video, reduce noise, adjust colors, add HDR effects, and generate extra frames for smoother playback. It works with local files, capture cards, screen capture, and game streams from PCs, PS5, and Xbox. Users can customize effects, save presets, take screenshots, record processed video, and export edited files. It is useful for improving the look of videos and game streams while watching them.

https://github.com/Likely7/Veyra-NRVideo