Top Open Source GitHub Projects — Sonarr, Dokku, neko, Spec Kit, WeKnora & Public APIs #29
Read the Portuguese translation →
Author: ManuAGI - AutoGPT Tutorials · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Top Open Source GitHub Projects — Sonarr, Dokku, neko, Spec Kit, WeKnora & Public APIs #29 · Channel: ManuAGI - AutoGPT Tutorials · 13:04 · Watch on YouTube
Topics: AI agents, Software engineering with AI, AI models, Open source and software ecosystems
Summary
Section titled “Summary”This weekly GitHub project roundup presents 20 open-source tools, ranging from AI coding-agent skills and research workflows to self-hosted media, deployment, and browser platforms. It also covers developer resources such as a public API directory, a 3D texture-painting app, and tools for AI model serving and on-device tool calling. Several projects focus on managing knowledge and agent context, including document-based RAG, persistent memory, and context-window optimization. The video closes with tools for emulator firmware, meeting memes, and visualizing AI agent sessions.
Key points
Section titled “Key points”- I Have ADHD is a skill for coding assistants that encourages concise answers, concrete estimates, and one clear next step.
- God’s Eye View presents public data about aircraft, ships, satellites, wildfires, and cameras on a 3D globe.
- Sonarr automates managing TV libraries, including finding, renaming, and upgrading releases.
- Hyper Research and Spec Kit offer structured workflows for AI-assisted research and specification-driven development.
- Context Mode indexes large tool outputs and gives coding agents pointers to relevant content instead of raw payloads.
- WeKnora turns documents into RAG-based answers, agent workflows, and a self-maintaining wiki.
- Dokku provides a self-hosted Heroku-style deployment path, while Neko streams a shared virtual browser over WebRTC.
- The roundup also features Public APIs, a directory of free APIs, and Needle 2, a small model for tool calling on constrained devices.
- OKF Agent Memory stores agent memory as Git-tracked Markdown, and Bot Crossing visualizes agent sessions as a game-like colony.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Introduction and projects 1–4 | The video introduces its weekly GitHub roundup and covers an ADHD-focused coding skill, a spatial intelligence globe, Armor Paint, and Sonarr. |
| 03:18 | Research, development, and AI tools | The roundup covers Hyper Research, Spec Kit, OpenAI’s plugin examples, Context Mode, WeKnora, GPT Image 2 prompts, and Modular. |
| 06:34 | AI resources and self-hosted applications | Projects include a GPT Image 2 prompt library, Modular’s AI platform, Public APIs, Needle 2, Dokku, and Neko. |
| 09:17 | Emulation, banking, and agent memory | The video presents Neko, Retro Bios, a read-only bank MCP server, a meme overlay for meetings, and Git-native agent memory. |
| 11:59 | Agent memory and session visualization | The final projects are OKF Agent Memory and Bot Crossing, which stores project memory in Git and visualizes agent sessions as a colony. |
Quotes
Section titled “Quotes”It makes the assistant lead with the next action. — 01:07
Instead of prompting an AI agent, you write the specification first, then a plan, a task breakdown, and the implementation. — 03:48
So the model retrieves the sections it needs. — 05:26
It stores memory as plain markdown files with YAML front matter inside your repo. — 11:59
Run it and watch your agents. — 12:32
Concepts and entities
Section titled “Concepts and entities”I Have ADHD, God’s Eye View, Armor Paint, Sonarr, Hyper Research, Spec Kit, OpenAI plugins, Context Mode, WeKnora, Modular, Public APIs, Needle 2, Dokku, neko, Bank MCP, Bot Crossing
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Every week, developers release new open-source projects on GitHub, and this is our weekly GitHub project update video covering the best of them. Top trending open-source GitHub projects. This week, you’ll discover useful and trending developer tools from self-hosted platforms to AI agent skills. Without wasting time, let’s get started. >> Before we jump into today’s project updates, here’s a quick announcement for everyone. We’ve launched a brand new YouTube channel called AI Agent Studio dedicated entirely to AI Agent projects, tutorials, and tools. So, if you’re
00:35 interested in staying up to date with the latest AI Agent open source projects, learning how to build your own agents, or exploring cuttingedge agent frameworks, make sure to check it out. Subscribe now to get weekly videos, in-depth guides, and real-time project breakdowns. The link is right there in the description. Don’t miss it. All right, let’s get into today’s video. >> Project number one, mine. I have ADHD skill that stops coding agents burying the answer. I have ADHD is an open- source skill for AI coding assistants
01:07 that stops them from burying the answer instead of a wall of preamble and hope this helps. It makes the assistant lead with the next action. Number multi-step tasks, cap lists, give concrete time estimates, and end with one clear next step. It’s 10 rules adapted from an 8HD toolkit. You install it into claude code or codeex and invoke it. Install it and get straight to the point. Project number two, God’s Eye View. Live open-source spatial intelligence on a 3D globe. God’s Eye View is an open-source browser-based spatial intelligence
01:40 console. A spy satellite simulator with real data. On a 3D globe, it layers live aircraft, ships, satellites, wildfires, and CCTV. So you move between a global view and a single object. You ride a tracked flight, lock onto anything to see its trail, and hand off to the nearest camera. It runs locally on public data, and by design, it tracks assets and systems, not people. Build it and watch the world. Project number three, Armor Paint. Open-source 3D PBR texture painting software. Armor Paint is an open-source application for 3D PBR
02:14 texture painting powered by the Armory Engine. You load a 3D model and paint physically based materials, color, roughness, metaleness, and normals directly onto its surface. Then export texture maps for your renderer or game engine. Written in hacks on the C and Chromst stack, it runs on Windows, Mac OS, and Linux with plugins that live linked to Blender, Unity, and Unreal. You build it from source for free. Clone it and paint your model. Project number four, the Sonar self-hosted PVR that manages your TV library. Sonar is an
02:48 open-source self-hosted PVR that manages your TV library for Usenet and Bit Torrent. It monitors RSS feeds for new episodes of the shows you follow, then grabs, sorts, and renames releases automatically. It scans your library for missing episodes, watches for better quality releases, and upgrades them DVD to Blu-ray, and retries another release when a download fails. It integrates with SAB DENSBD, NZBG, Plex and Kodi, and runs on Windows, Mac OS, Linux, and
03:18 Raspberry Pi. Deploy it and automate your shows. Project number five, Hyper Research. Turn Claude Code into an audited deep research agent. Hyper Research is an open- source harness that turns Claude Code into a disciplined deep research agent. A tier adaptive 16-step pipeline gathers sources, cross-checks them, and produces an adversarily audited report with provenence. Every citation verified against the sentence it supports before the report ships. Each fetched source lands in a persistent markdown plus
03:48 esquilite vault. Your next session reuses, so research compounds. It chases open access full text for paywalled papers, resumes crashed runs, and scales from a survey to a dissertation. Install it and research anything. Project number six, SpecKit. GitHub’s toolkit for specd-driven development with AI. SpecKit is an open-source toolkit from GitHub for specriven development. Instead of prompting an AI agent, you write the specification first, then a plan, a task breakdown, and the implementation. It specifies CLI
04:20 scaffolds a project and each phase lives as a markdown artifact constitution spec plan and tasks right in your repo driven by slash commands. It works with 30 plus agents including copilot claude code cursor and gemini with a human in the loop. Install it and spec your project. Project number seven bingies openai’s repository of codec plug-in examples. Plugins is OpenAI’s open-source repository of codec plug-in examples. A plug-in is a folder with a manifest that
04:51 packages skills, MCP servers, app integrations, commands, and hooks into a sharable installable bundle that works across the Codeex app, CLI, and IDE extensions. The repo defines the marketplace format the client reads, and it includes plugins curated by OpenAI like Box, Cloudflare, and Figma. Register it as a marketplace or copy one to build your own. Browse it and extend codecs. Project number eight, context mode. MCP server that optimizes an agent’s context window. Context mode is an MCP server that keeps AI coding
05:26 agents from drowning in tool output. When a tool call returns a large payload, it indexes the full content into a local full text database and hands the agent a pointer instead of raw text. So the model retrieves the sections it needs. It persists session memory across compaction and enforces routing through hooks. It works across 17 platforms including clawed code and cursor. Install it and reclaim your context. Project number nine. Wignora. Turn documents into rag agents and autowiki. Wignora is an open-source LLM
05:59 powered knowledge platform from Tencent that turns documents into knowledge three ways. Ragbased Q&A for lookups, a React agent that orchestrates retrieval. MCP tools and web search for multi-step tasks and a wiki mode where agents distill documents into a self-maintaining interlin markdown wiki with a knowledge graph. It ingests 10 plus formats, autosyncs from FishU, Notion and UK. Works with 20 plus LLM providers and swaps vector databases. Self-host it with Docker and query your documents. Project number 10. Wansom GPT
06:34 image 2 prompt as code template library and skill for GPT image 2. WASUM GPT image 2 is an open-source prompt as code library of prompts and templates for OpenAI’s GPT image 2 image model. It reverse engineers 530 plus real examples into 20 plus reusable structured templates each with the full prompt and generation record covering posters, product shots, UI, infographics, brands, characters, and more. The templates are distilled into an agent skill that turns
07:06 your image intent into a productionready prompt installable via npm and a live gallery lets you browse, filter, and copy prompts. Install it and generate better images. Project number 11, Modular Platform. Unified AI platform with the Max Server and Mojo. Modular platform is an open suite of AI libraries and tools for AI development and deployment, bundling the Max framework and the Mojo language. Max is a high performance inference server that exposes an OpenAI compatible endpoint. So you install it with pip and serve
07:39 open models like Llama on Nvidia or AMD GPUs without changing your code. The repo holds open- source pieces, the Mojo standard library, max GPU and CPU kernels, and the serving layer. Install it and serve a model. Project number 12, public APIs, a curated list of free public APIs. Public APIs is a community-maintained open-source list of free APIs for software and web development. It organizes hundreds of public APIs into categories: finance,
08:10 weather, machine learning, geocoding, games, government, and dozens more. So you can find one for your project instead of hunting across the web. Every entry notes what it does, whether it needs authentication, and whether it supports HTTPS and cores with a direct link. Anyone can add an API by pull request. Browse it and find your API. Project number 13, Needle 2, 14 MB tool calling model for tiny devices. Needle 2 is an open-source 45 million parameter model for tool calling, device use, and structured extraction. The whole model
08:44 is a single 14 MB binary that runs a full session in about 28 MB of RAM. So, it fits phones, wearables, and robots. shipped as a Python package. You describe your tools and it returns calls as JSON with a grammar from your schemas constraining every token. Each response carries a confidence score for escalation. Install it and call your tools on device. Project number 14, Daku. Docker powered self-hosted path for your apps. Daku is an open-source Dockerpowered path that helps you build and manage the life cycle of
09:17 applications on your own server. It’s a small self-hosted Heroku alternative. You get push your app and do builds it with build packs or docker files and runs it. Install it on a fresh abundu or Debian VM and deploy your app. Project number 15, Niko, self-hosted virtual browser streamed over WebRTC. Niko is an open-source self-hosted virtual browser that runs in Docker and streams over WebRTC. It gives you a browser inside an isolated container you reach from a web page, so you browse securely and privately from anywhere. Multiple people
09:51 can join the same session and share control. Handy for collaboration, watch parties and presentations. It isn’t limited to a browser. It can stream any Linux app or a desktop and automate it with Playright. Deploy it and share a browser. Project number 16, Retro Bios. Sourceverified BIOS and firmware packs for emulators. Retrobios is an open-source project that assembles source verified BIOS and firmware packs for emulation frontends like RetroArchch, Bata Sierra, Retro Pi, Emo Dec, and MR FPGA. It reads each emulator
10:25 source code to learn which files it opens, then verifies packs the way the platform does, so a game never stops on a missing file. A gap analysis names any gaps. An autodetecting installer places files correctly. The tooling is open source. The firmware is thirdparty software for personal backup. Grab your platform’s pack. Project number 17. Bank MCP. Self-hosted read only MCP server for your bank. Bank MCP is an open-source self-hosted MCP server that
10:55 lets your AI assistant read your bank accounts. It connects through enable banking which wraps 2,700 plus European banks in one PSD2 API then exposes them to any MCP client like claude chatgpt or cursor. It’s read only with no payments and your server holds only the consents and account ids storing no balances or transactions and sending no telemetry. Run it with NPX or a container. Deploy it and ask about your money. Project number 18. It’s giving prick. overlay
11:26 memes on your face in meetings. Its giving prick is a playful open- source Python tool that lets you express yourself in video meetings with memes. It uses Media Pipe to track your face in real time, recognizes about 14 facial expressions, and overlays a matching meme on your webcam feed as you react to avoid fixed thresholds. It normalizes against your own neutral face, so the detection adapts to you. It’s built for internal demos and calls where a little levity helps. run it and mean your meetings. Project number 19, man’s OKF
11:59 agent memory. Git native persistent memory for AI coding agents. OKF agent memory is an open-source git native memory engine that gives AI coding agents persistent project memory. It stores memory as plain markdown files with YAML front matter inside your repo. So you audit and roll back what it remembers with plain git, no vector database. It searches with BM25, uses progressive disclosure to load only the concepts an agent needs, and enforces search before write. It ships as a Go binary with a built-in MCP server.
12:32 Install it and give your agent memory. Project number 20, Bot Crossing, a game that visualizes your AI agent sessions. Bot crossing is a video game for your AI agents, a local dashboard that turns your agent sessions into an Animal Crossing style colony. Every hex tile is an agent and every astronaut is a job running on its own with visual status. A hammer means working. A question mark means it’s waiting on you and red means stuck. It reads each harness’s local session files through an adapter so every installed harness appears in one
13:04 colony. Run it and watch your agents. Thanks for watching. See you in the next update.