DEV Community: Adpirs

You got tricked hehe :)

Adpirs — Tue, 27 Oct 2026 17:11:20 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

My Agent Session

The Spirit of It

Prize Categories

lazy to build something :(

Adpirs — Fri, 23 Oct 2026 16:58:33 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

My Agent Session

The Lazy Payoff

Prize Categories

Great Sage (Raphael AI): The Universal, 100% Private Desktop Voice AI Companion for Everyone!

Adpirs — Sun, 18 Oct 2026 07:31:36 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

Great Sage (Raphael AI) is an open-source, autonomous desktop voice AI companion and personal operating copilot built for anyone and everyone who uses a computer—whether you are a student, gamer, content creator, developer, or daily PC user.

Inspired by the omniscient Great Sage / Raphael (Lord of Wisdom) from That Time I Got Reincarnated as a Slime and the engineering brilliance of JARVIS, Great Sage turns your personal computer into an intuitive, voice-controlled command center that runs 100% locally on your machine with zero subscription fees.

Who Is This For & Who Would Pay For It?

Most modern AI platforms force users into perpetual $20–$100+/month subscriptions and upload personal screen activity to corporate cloud servers. Great Sage flips this paradigm on its head by delivering an Everyday Desktop Companion that anyone would pay for as a one-time product

Everyday Multitaskers & PC Users: Anyone tired of juggling dozens of windows and endlessly clicking around. Just say "Play some chill music on YouTube" or "Turn the volume down", and Great Sage handles it hands-free without touching the mouse or keyboard.
Gamers & Streamers: Players in full-screen games who want to switch tracks, mute audio, launch Discord, or look up a tip without Alt-Tabbing or crashing their game.
Students & Learners: Anyone studying who wants a private, offline tutor and note-keeper that doesn't track their search history or distract them with web ads.
Developers & Creators: Anyone handling confidential work, proprietary code, or personal documents who demands 100% offline privacy and zero monthly API bills.

For a one-time purchase of $15–$29 (or free for the open-source community), everyday users get their own loyal, offline AI companion for life on their laptop.

🌟 Complete System Capabilities:

1. Instant Holographic Magic-Circle HUD

Global Summon Hotkey: Press Ctrl + Shift + S (or Ctrl + Alt + S) anywhere on Windows, and the HUD instantly snaps to the center of your screen in milliseconds (<5ms) without flashing any Command Prompt windows.
Audio-Reactive Anime Interface: Translucent, floating borderless window featuring authentic rotating concentric magic-circle rune rings and real-time audio waveform visualizers.
Hands-Free Continuous Listening: Wake-word activated ("Daikenja", "Great Sage", "Raphael"), complete with instant voice barge-in interruption (speak while it talks to interrupt immediately).
Bilingual Japanese/English Output: Synthesizes spoken Japanese in Great Sage’s signature cadence (『告。』 / 『解。』) via high-fidelity neural audio, while rendering English translations on the display card.

2. Hands-Free Desktop & Application Control

Launch Any App: Seamlessly opens apps by voice (Spotify, Discord, VS Code, Chrome, Edge, Steam, VirtualBox, Terminal, Calculator, Notepad, OBS Studio, Blender, MS Word, Excel, PowerPoint, Paint).
Multi-Command Chaining: Understands compound directives in a single sentence (e.g. "Open VS Code and launch Spotify" or "Open Notepad and type my shopping list").
App Termination: Gracefully closes or terminates any background process or application on command.
Voice Dictation / Auto-Typing: Dictates text directly into target apps like Notepad or chat windows hands-free.

3. Media & Audio Entertainment Hub

Direct YouTube Video Autoplay: Scrapes YouTube video IDs in real time and automatically launches the exact top video with autoplay enabled.
Deep Spotify Control: Searches and plays songs, albums, artists, or playlists directly through the native Spotify protocol.
Hardware Volume & Media Keys: Adjusts master system volume ("Volume up", "Volume down", "Set volume to 40%", "Mute"), and controls playback (Play, Pause, Next Track, Previous Track).

4. Active Workspace & Context Awareness

"Help Me With What I'm Doing": When invoked, Great Sage inspects your active focused application (e.g. IDE, browser tab, or document) and reads your current clipboard context to give instant, tailored advice.
Instant Full-Screen Capture: Captures desktop screenshots on voice command and saves them directly to your storage.

5. Self-Improving Long-Term Memory (Knowledge Bank)

Automatic Preference Learning: Learns your daily habits, preferred programming languages, favorite music, study routines, and corrections after every conversation.
Persistent SQLite Brain: All memories are stored locally in data/raphael_memory.db and injected into future cognitive reasoning.
Knowledge Bank UI: Built-in visual memory inspection drawer where you can view, edit, search, and delete learned rules.

6. Power-User, Virtualization & Developer Tools

VirtualBox VM Automation: Starts virtual machines in GUI or headless mode ("Start Kali Linux VM"), halts VMs, and takes or restores snapshots by name.
Docker Management: Checks container status, starts, stops, or restarts Docker containers hands-free.
Git Voice Actions: Checks repository status, stages files, commits with voice messages, and pushes to remote branches.
Silent Shell Commands: Safely executes background terminal and PowerShell commands without pop-up terminal clutter.

7. Tensura "Appraisal" (Kantei / Kaiseki) & Error Diagnosis

Deep Code Appraisal: Analyzes files, functions, or architectures and reports security vulnerabilities, performance bottlenecks, and refactoring tips.
Stack Trace Diagnosis: Paste or dictate error traces, and Great Sage analyzes the exact failure point and provides working fixes.

8. Timers, Alarms & Tactical Presets

Live Countdown Timers: Set timers with custom labels ("Set a 15-minute study timer"); renders visual countdown pills directly inside the HUD.
Tactical Workspace Presets: Switch desktop modes in one command ("Focus Mode", "Gaming Mode", "Work Mode"), minimizing clutter and preparing your apps.

9. Hardware Telemetry & Autonomous Guardian

Live System Telemetry: Voice query CPU usage, RAM utilization, GPU temperature, disk space, network activity, and laptop battery health.
Autonomous Guardian Sentinel: An active background sentinel thread that auto-verifies local neural engines (Ollama) and heals services automatically if interrupted.

10. 100% Offline, Zero-Subscription Intelligence

Local Gemma 3 (4B) via Ollama: Runs entirely on your GPU VRAM with zero API costs, zero data sent to external servers, and unlimited daily usage without internet.
Optional Frontier Cloud Fallback: Integrated Google Gemini Flash-Lite router for when deep cloud reasoning is preferred.

Demo

Youtube: https://youtu.be/F6hp9alkW6U

Example Commands Anyone Can Use:

"Great Sage, open Spotify and turn the volume up."
"Play relaxing lofi on YouTube."
"Help me with what I'm doing." (Reads foreground window context and clipboard to assist immediately).
"Remember that I prefer dark mode and study late at night." (Persisted to memory).
"Mute audio and lock workstation."

Code

GitHub repo: https://github.com/SripadaError101/Great_Sage

How I Built It

Local Neural Inference: I integrated Ollama running Google's open-weight Gemma 3 (4B). The model is pinned directly in GPU VRAM with a keep_alive daemon for instant <100ms response times.
Hybrid Cloud Fallback: Integrated an optional router for Google's Gemini Flash-Lite when users want cloud frontier capabilities while traveling.
Desktop Automation Harness: Engineered a Python FastAPI backend connected to native Win32 APIs via ctypes and subprocess, enabling instant desktop control without external bloat.
Zero-Infobar Edge App HUD: Designed a custom lightweight Chromium app profile that sets permanent native microphone permissions and screen geometry, opening a clean, borderless floating window in milliseconds.

Why Does Open Innovation Matter?

Open innovation is the single reason a project like Great Sage can belong to everyone, rather than being locked behind big-tech paywalls:

AI as an Asset, Not a Landlord: Cloud AI forces users into perpetual rent-seeking subscriptions ($240+/year for Copilot Pro or ChatGPT Plus). Open-weight models running on consumer hardware prove that powerful AI can be owned outright by everyday users.
Uncompromising Personal Privacy: Everyone has personal data on their PC—family photos, banking tabs, personal journals, and private chats. Running open-weight models locally guarantees that personal life never leaves the user's hard drive.
Inclusive & Offline: Big-tech AI stops working the second your Wi-Fi drops or your mobile hotspot runs out. Great Sage operates on airplanes, trains, during internet blackouts, or in remote regions.
Zero Censorship or Vendor Depreciation: Users can swap between Gemma, Llama, Mistral, or specialized models with a single line in config.py without worrying about an API provider changing terms of service.

Grounded: The Zero-Signal Outdoor Companion Powered by Local Gemma 3 🌿 #TouchGrass

Adpirs — Wed, 07 Oct 2026 14:19:34 +0000

🌿 Grounded: TouchGrass AI

Built with Open-Weights Gemma 3 for the "Touch Grass" Challenge

Most modern software is engineered to maximize dwell time—infinite scrolls, algorithmic feeds, and push notifications designed to keep your eyes locked to a glass screen.

I built Grounded for the exact opposite. Its guiding design metric is:

Screen Time ≤ 45 Seconds. Nature Time ≥ 45 Minutes.

🍃 What Grounded Does

Grounded is an air-gapped, zero-signal nature companion designed to get you off the screen and into the real world:

⚡ 30-Second Field Dispatcher: Select your available window (15m to 90m) and intent (Foliage Run, Mindful Forest Walk, Birding Stroll, or Homestead Soil Work). Local Gemma 3 crafts a concise mission card with 1 physical task, 1 auditory mindfulness anchor, and an urgent prompt to lock your device and step outside.
🐦 Zero-Signal Bioacoustic Bird Call ID: Record directly from your microphone or upload audio. A local NumPy/SciPy FFT engine calculates peak dominant frequencies and generates a live spectrogram, while Gemma 3 identifies species candidates, visual field marks, and flocking behavior with zero internet required.
🌱 Frost Date & Soil Garden Action Engine: Computes exact days until first fall frost for your hardiness zone with immediate action items for planting garlic cloves, cover crops, and overwintering greens.
🏃 Foliage & Run Club Route Synthesizer: Plots scenic autumn loops with canopy ratings and exports standard GPX files directly to Garmin, Apple Watch, or Strava—allowing you to leave your phone at home entirely.
☀️ Sunlight & Night-Vision Modes: Features high-contrast black-on-white mode for direct outdoor sunlight and deep crimson red mode for night owling and stargazing.

💡 Why Open-Source AI is What Makes This Work

Proprietary cloud APIs (OpenAI, Anthropic, cloud-hosted LLMs) fail fundamentally in the outdoors for four reasons:

Dead Zones in the Wilderness: The best trails, ridges, and old-growth canopies have zero cellular reception. Grounded runs Google's open-weights Gemma 3:4b locally via Ollama on my laptop's GPU (RTX 3060 CUDA). It works just as fast on airplane mode 15 miles into a state park as it does on fiber.
Radical Foraging & Trail Privacy: Foragers and naturalists protect their secret mushroom patches and raptor nests. In Grounded, observations live in an encrypted local SQLite database. Zero corporate telemetry, zero ad tracking.
Anti-Dwell Time: Commercial chatbots optimize for conversational engagement. Because I control the open model prompt, I instructed Gemma 3 to be radically terse: give the user their mission in 30 words and order them to shut the laptop.
$0.00 Forever: No API meters, credit cards, or subscription tiers. Nature intelligence belongs in the public commons.

🌲 Field Test: Taking It Outside!

We took Grounded into a local woodland park on an overcast autumn afternoon with Wi-Fi disabled:

The Dispatch: Generated a 45-minute "Autumn Foliage Trail Run". Gemma 3 dispatched our mission in 2.1 seconds: collect 5 fallen acorns and pause at minute 20 for 60 seconds of silent auditory anchoring.
The Route: Downloaded the 5.0 km loop GPX to our watch and shut the laptop lid.
The Sighting: Paused at the 2.2 km mark near white pines and captured a 2.5s two-tone whistle. Local FFT and Gemma 3 identified the Black-capped Chickadee (Poecile atricapillus) at 3450 Hz with 93% confidence within 1.2s.
Total Screen Time: Under 45 seconds. Total outdoor time: 48 minutes of trail running, crisp air, and muddy shoes.

Check out the repo, run it locally, and go touch grass! 🍂

GitHub Repo:https://github.com/SripadaError101/grounded-touchgrass

Speech Practice Partner

Adpirs — Sat, 03 Oct 2026 21:44:09 +0000

What I Built

Most voice assistants decide you've finished speaking after about a second of silence. For someone who stammers, is nervous about speaking, or is searching for words (after a stroke, for example), that means being cut off again and again. It makes practising out loud feel worse, not better.

So I built Speech Practice Partner, a patient conversation partner that never interrupts. You pick a topic you enjoy, press one big button and just talk. It waits as long as you need (you choose anywhere from 2 to 15 seconds of quiet), then replies in one or two short, gentle sentences. It never finishes your sentences, never corrects your grammar and never comments on pauses or repetition.

I built it for anyone and everyone, who gets anxious speaking to people and wants a low-pressure place to practise.

Things I cared about:

Calm and simple: one screen, one button, a soft orb that shows whether it is listening, hearing you, thinking or speaking.
Private: everything runs on the laptop and nothing goes to the cloud. Only the written transcript and a few numbers are saved, and your voice is never recorded. One button deletes everything.
Progress without pressure: a Progress page shows personal trends (pause length, pace, words per session). They are framed as trends, not scores, and the numbers are hidden during a session by default.

Demo

Video Link :https://youtu.be/UPSXFH25tTc?si=QE1RDLcW4migRK6D

The app has a --demo mode that previews the whole interface with synthetic sample data and a pretend conversation. Any trend charts in the demo come from that synthetic data and are not real people's results.

How I Built It

The whole pipeline runs locally:

Mic → Silero VAD → faster-whisper → Gemma (via Ollama) → Piper voice

Silero VAD detects when you start and stop talking. I set the end-of-speech silence window myself, which is the heart of the project.
faster-whisper (small.en) transcribes locally. I prompt it to keep repetitions and fillers instead of "cleaning them up".
Gemma (gemma3:4b, open weights, run through Ollama) is the conversation partner. A short system prompt keeps it warm, brief and non-corrective.
Piper speaks the replies, with an adjustable slower voice.
TabPFN (optional, local) forecasts the next session's pause length and speaking rate from past sessions. It is backtested against a naive baseline, and it needs at least 8 real sessions before it will forecast. [Add your backtest result here once you have real sessions, or delete this line.]
Flask + plain HTML/CSS/JS make up the interface. There are no frontend frameworks and no CDN, so it works fully offline. The server only listens on 127.0.0.1.

I started from a terminal-only prototype and turned it into a simple browser app: settings inside the app, friendly error messages (like "Open the Ollama app and try again"), a session you can end at any moment, and dark mode.

Why Does Open Innovation Matter?

Control over the one setting that matters. With a closed voice API, I can't change how long it waits before deciding you're done. With open components I could set that window to match the person, and tune it with them.
Privacy. A person practising something vulnerable shouldn't have to send their voice to someone else's servers. Because every model runs locally, their speech never leaves the laptop.
Free, unlimited practice. Confidence builds with repetition, and nobody has to worry about per-minute API costs.
Tunable to the person. The prompt, voice speed and wait time can all be adjusted after each session.

Prize Categories

Gemma

Intro...

Adpirs — Thu, 01 Oct 2026 17:57:21 +0000

Hi all, I am new to dev and I really like the UI of this app gives you that gamer like feeling when you see the text...lol (^_^)