DEV Community: Dien

SlothDigest AI — Rewarding Laziness with Local Open-Weight Triage & Wind-Down

Dien — Tue, 20 Oct 2026 00:37:23 +0000

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

What I Built

October 20 is International Sloth Day. Sloths have mastered life: they conserve energy, move slowly, and let the world come to them.

Modern developers do the exact opposite. We drown in 500-unread Slack channels, endless email chains, cognitive code overload, and looming dreaded tasks that we procrastinate for weeks.

I built SlothDigest AI: a 100% local-first, zero-dependency assistant that rewards healthy laziness by automating the busywork and protecting your mental bandwidth.

Key Capabilities:

💬 Chat & Thread Distiller: Paste chaotic chat logs, PR arguments, or multi-threaded emails. SlothDigest strips polite greetings, noise, and emojis, isolates direct @me action items, and generates a punchy 2-line sloth summary via Google Gemma AI.
🦥 The Lazy Payoff Calculator: Computes exact word count reductions (e.g. 800 words down to 40 words) and displays your percentage of reading effort spared and minutes saved.
🌙 Offline Sloth Mode & Ambient Brown Noise: Prescribes tailored decompression routines for eye fatigue and decision overload, powered by an in-browser Web Audio brown-noise synthesizer that runs entirely offline with zero media files.
⚡ 2-Minute Task Shrinker: Takes daunting tasks ("organize 1,000 files in Downloads", "file quarterly taxes") and shrinks them into an embarrassingly simple 2-minute starter step.

The Lazy Payoff: How Much Effort It Actually Saves

90% Reading Effort Spared: Reading a 15-minute chaotic thread drops to a 10-second glance at isolated action items.
Zero Procrastination Friction: The hardest part of dread is starting; shrinking tasks to 2-minute micro-actions eliminates executive dysfunction.
Zero Cloud Subscriptions: No $20/month SaaS billing; runs freely on your own machine.

Demo

Live Web Application (Hosted on Render): https://hacktoberfest-2026.onrender.com/week-3/
Unified Hacktoberfest Portal: https://hacktoberfest-2026.onrender.com/

You can also run it locally on zero external dependencies:

git clone https://github.com/dieennn/hacktoberfest-2026.git
cd hacktoberfest-2026
python main.py
# Open http://localhost:8080/week-3/

Code

The full project is open-source and part of my unified Hacktoberfest 2026 monorepo:

dieennn / hacktoberfest-2026

🎃 Hacktoberfest 2026 — DEV Challenges Monorepo

Repository containing project submissions for the Hacktoberfest 2026 DEV Challenges ("AI belongs to everyone").
Built by @dieennn.

📁 Repository Structure

Round	Challenge Theme	Folder / Project	Status
Launch Weekend (Oct 2 - Oct 5)	Build for a Friend	`00-weekend-allergy-guard`	✅ Completed
Week 1 (Oct 5 - Oct 12)	Touch Grass	`01-week-1`	✅ Completed
Week 2 (Oct 12 - Oct 19)	Side Hustle	`02-week-2`	✅ Completed
Week 3 (Oct 19 - Oct 26)	Slow and Steady	`03-week-3`	✅ Completed
Week 4 (Oct 26 - Nov 1)	Announced Oct 26	`04-week-4`	⏳ Upcoming

🚀 Projects Overview

00-weekend-allergy-guard — Allergy & Diet Guard (SafeBite AI)

Theme: Build for a Friend
Problem: Protect friend Sarah (severe peanut allergy, celiac disease, and lactose intolerance) from hidden allergens (arachis oil, seitan, casein) and cross-contamination when eating out.
Tech: Local Python 3…

View on GitHub

Directory: 03-week-3/

How I Built It

1. Pure Standard Library (True Developer Laziness)

The most resilient code is the code you never have to install or maintain. SlothDigest requires zero pip install commands. Built purely on Python 3's built-in http.server, urllib, re, and json, it starts up instantly on any system without virtual environment hassles.

2. Google Gemma Open-Weight AI Bridge

SlothDigest integrates with Google's Gemma open-weight models (gemma2:2b) running locally via Ollama (http://localhost:11434). It generates two-line summaries and contextual 2-minute task simplifications with sub-second local latency.

3. Resilient Offline Graceful Degradation

When offline or on low battery, the system seamlessly falls back to a deterministic regex noise filter and cognitive behavioral habit routines. It never crashes or leaves the user hanging.

4. Synthesizer Without Bloat (Web Audio API)

Instead of bundling heavy MP3 audio assets or calling streaming APIs, the ambient brown noise generator is synthesized mathematically in JavaScript using the browser's native Web Audio API in under 20 lines of code.

Why Open Innovation Matters

Private Chat & Inbox Security: Sensitive company chats, client names, and confidential conversations should never be uploaded to proprietary cloud summarization APIs. Open-weight models running on your machine keep private data private.
Permanent Autonomy: Open models don't introduce rate limits, surge pricing, or sudden API deprecations.
Low-Power Efficiency: Small open models like Gemma 2B deliver fast local inference on ordinary laptops without needing cloud compute clusters.

Prize Categories

Hacktoberfest Open-Source AI Challenge: Week 3 (Theme: Slow and Steady)
Best Use of Render: Deployed seamlessly as part of a single-dyno unified Python gateway running on Render's free tier.
Best Use of Gemma: Integrated Google's open-weight Gemma model to deliver concise conversational distillation and task simplification.

HustleLedger AI — Privacy-First Financial Sentinel & Scope Defender for Side Hustlers

Dien — Tue, 13 Oct 2026 01:35:48 +0000

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

What I Built

Every side hustler and freelancer faces the same triple threat:

Expensive accounting SaaS fees ($30–$50/mo) eating up precious early margin.
Leaking confidential client rates and bank transactions to proprietary third-party cloud data-miners.
Losing money to unpaid scope-creep and overdue invoices because drafting firm boundary emails feels awkward and confrontational.

I built HustleLedger AI: a 100% local-first, zero-dependency financial sentinel and negotiation companion designed for freelancers, creators, and micro-consultants.

Key Capabilities:

🧾 Expense Tax Shield: Instantly categorizes receipts (cloud hosting, hardware gear, client coffee meetings, home utilities) into recognized freelance tax deduction categories (100% vs 50% write-offs) and calculates real cash kept in your pocket.
📄 Instant Client Invoice Builder: Generates structured, verified client billing statements with automatic itemization and standard Net 14 terms with zero subscription overhead.
🛡️ Scope-Creep & Overdue Defender: Interfaces with Google's Gemma open-weight AI to draft diplomatic yet rock-firm overdue invoice reminders and professional boundaries when clients ask for extra unpaid work.

Demo

Live Web Application (Hosted on Render): https://hacktoberfest-2026.onrender.com/week-2/
Unified Hacktoberfest Portal: https://hacktoberfest-2026.onrender.com/

You can also run it locally with zero pip installs:

git clone https://github.com/dieennn/hacktoberfest-2026.git
cd hacktoberfest-2026
python main.py
# Open http://localhost:8080/week-2/

Code

The full project is open-source and part of my unified Hacktoberfest 2026 monorepo:

dieennn / hacktoberfest-2026

🎃 Hacktoberfest 2026 — DEV Challenges Monorepo

Repository containing project submissions for the Hacktoberfest 2026 DEV Challenges ("AI belongs to everyone").
Built by @dieennn.

📁 Repository Structure

Round	Challenge Theme	Folder / Project	Status
Launch Weekend (Oct 2 - Oct 5)	Build for a Friend	`00-weekend-allergy-guard`	✅ Completed
Week 1 (Oct 5 - Oct 12)	Touch Grass	`01-week-1`	✅ Completed
Week 2 (Oct 12 - Oct 19)	Side Hustle	`02-week-2`	✅ Completed
Week 3 (Oct 19 - Oct 26)	Announced Oct 19	`03-week-3`	⏳ Upcoming
Week 4 (Oct 26 - Nov 1)	Announced Oct 26	`04-week-4`	⏳ Upcoming

🚀 Projects Overview

00-weekend-allergy-guard — Allergy & Diet Guard (SafeBite AI)

Theme: Build for a Friend
Problem: Protect friend Sarah (severe peanut allergy, celiac disease, and lactose intolerance) from hidden allergens (arachis oil, seitan, casein) and cross-contamination when eating out.
Tech: Local Python 3…

View on GitHub

Directory: 02-week-2/

How I Built It

1. Zero External Dependencies (Standard Library Architecture)

When running a side hustle, every dollar and millisecond matters. HustleLedger relies strictly on the Python 3 standard library (http.server, socketserver, json, urllib). No bloated virtual environments, no node_modules, and instant startup on any machine.

2. Google Gemma Open-Weight AI Integration

HustleLedger bridges to Google's Gemma open-weight models (gemma2:2b) running via local Ollama inference (http://localhost:11434/api/generate). When negotiating sensitive client boundaries or collecting overdue payments, Gemma drafts polite, assertively professional messages tailored to the exact dollar amount and context.

3. Bulletproof Offline Fallback

If Ollama is not active, the system automatically degrades to deterministic, tested communication heuristics and tax deduction tables (EXPENSE_TAXONOMY). The application functions continuously offline without cloud latency or failure points.

Why Open Innovation Matters for Side Hustlers

Complete Financial Privacy: Client agreements and billing numbers never leave your local machine or trusted deployment.
Zero Recurring Tool Tax: Solopreneurs keep 100% of their earnings instead of paying recurring subscription fees to SaaS platforms.
Resilience & Autonomy: Open-weight models empower independent makers to own their intellectual property and business workflows without vendor lock-in.

Prize Categories

Hacktoberfest Open-Source AI Challenge: Week 2 (Theme: Side Hustle)
Best Use of Render: Deployed as an integrated micro-route on Render's free tier gateway monorepo.
Best Use of Gemma: Integrated Google's open-weight Gemma model to assist freelancers in business negotiation and scope protection.

TrailFlora & Garden AI — Breaking Screen Fatigue with Offline Open-Weight Botanical Intelligence

Dien — Tue, 06 Oct 2026 01:40:18 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Modern life traps developers and creators behind glowing screens for 10–14 hours a day. The prompt for Week 1 — "Touch Grass" — challenges us to build an open-source AI companion that makes the screen the shortest part of the user journey, nudging people outside into sunlight, woods, and backyard gardens.

I built TrailFlora & Garden AI: an offline-first botanical scout and autumn planting sentinel designed for trail hikers, backyard gardeners, and outdoor walkers.

Key Capabilities:

Trail Flora & Toxic Weed Sentinel: Identifies wild species from simple descriptive observations (leaf shape, berry clusters, stem geometry) and alerts hikers against contact dermatitis hazards (Poison Ivy, Poison Oak) or lethal ingestions (Deadly Nightshade/Belladonna).
Backyard Garden & Frost Predictor: Calculates zone-specific autumn planting schedules and warns when ambient evening temperatures threaten frost-sensitive crops.
Touch Grass & Sunlight Scout: Evaluates real-time ambient temperature and cloud coverage to compute an "Outdoor Vitality Score", suggesting optimal trail walk windows that reset digital eye fatigue and circadian rhythm.

Demo

Live Web Application (Hosted on Render): https://hacktoberfest-2026.onrender.com/week-1/
Unified Hacktoberfest Portal: https://hacktoberfest-2026.onrender.com/

You can also run it locally on zero pip dependencies:

git clone https://github.com/dieennn/hacktoberfest-2026.git
cd hacktoberfest-2026
python main.py
# Open http://localhost:8080/week-1/

Code

The full project is open-source and part of my unified Hacktoberfest 2026 monorepo:

dieennn / hacktoberfest-2026

🎃 Hacktoberfest 2026 — DEV Challenges Monorepo

Repository containing project submissions for the Hacktoberfest 2026 DEV Challenges ("AI belongs to everyone").
Built by @dieennn.

📁 Repository Structure

Round	Challenge Theme	Folder / Project	Status
Launch Weekend (Oct 2 - Oct 5)	Build for a Friend	`00-weekend-allergy-guard`	✅ Completed
Week 1 (Oct 5 - Oct 12)	Announced Oct 5	`01-week-1`	⏳ Upcoming
Week 2 (Oct 12 - Oct 19)	Announced Oct 12	`02-week-2`	⏳ Upcoming
Week 3 (Oct 19 - Oct 26)	Announced Oct 19	`03-week-3`	⏳ Upcoming
Week 4 (Oct 26 - Nov 1)	Announced Oct 26	`04-week-4`	⏳ Upcoming

🚀 Projects Overview

00-weekend-allergy-guard — Allergy & Diet Guard (SafeBite AI)

Theme: Build for a Friend
Problem: Protect friend Sarah (severe peanut allergy, celiac disease, and lactose intolerance) from hidden allergens (arachis oil, seitan, casein) and cross-contamination when eating out.
Tech: Local…

View on GitHub

Directory: 01-week-1/

How I Built It

1. Zero External Dependencies (Standard Library Architecture)

Out on a wilderness hiking trail, you have zero cellular data and cannot run pip install on bloated 500 MB cloud SDKs.
TrailFlora is built using pure Python standard library (http.server, urllib, json, re). It loads in milliseconds on any battery-constrained laptop or offline field device.

2. Google Gemma Open-Weight AI Bridge

TrailFlora interfaces with Google's Gemma open-weight models (gemma2:2b / gemma:2b) running via local Ollama inference (http://localhost:11434/api/generate). When hikers encounter ambiguous foliage descriptions, Gemma generates concise 2-sentence outdoor ranger field advice.

3. Resilient Offline Graceful Degradation

If the user is deep in the backcountry without an active local LLM instance, the engine automatically falls back to a deterministic botanical rule matrix and hardiness database (PLANT_TAXONOMY and GARDEN_SCHEDULES). The application never crashes or leaves the user stranded on the trail.

Why Does Open Innovation Matter?

True Off-Grid Capability: Proprietary cloud AI endpoints (e.g., OpenAI or Claude APIs) are completely useless when you lose signal 4 miles into a forest trail. Open-weight models (like Google Gemma) belong to the user, run on local hardware, and never demand an active Wi-Fi connection.
Zero Recurring Infrastructure Costs: By pairing an open-weight model with a lightweight Python standard library gateway, anyone can self-host this application for free on Render's free tier.
Data Privacy in the Wild: Your personal trail coordinates, garden habits, and schedule notes stay strictly on your local device.

Prize Categories

Hacktoberfest Open-Source AI Challenge: Week 1 (Theme: Touch Grass)
Best Use of Render: Deployed as part of a unified multi-application monorepo web gateway running seamlessly on Render's cloud platform.
Best Use of Gemma: Integrated Google's open-weight Gemma model to deliver trail ranger field guidance and botanical safety insights.

Allergy & Diet Guard — Protecting My Friend with Local Open-Source AI

Dien — Fri, 02 Oct 2026 11:20:52 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built Allergy & Diet Guard (SafeBite) for my friend Sarah, who lives with an unforgiving combination of dietary hazards:

Severe Peanut Allergy (life-threatening anaphylactic shock)
Celiac Disease (strict gluten-free requirement)
Lactose Intolerance (no dairy)

Eating out at local restaurants, street food stalls, or night markets with Sarah is usually stressful. Menus and product labels rarely say "Peanut" or "Wheat"; they list obscure culinary terms like "arachis oil", "seitan", "malt extract", "casein", or "modified starch". A single contaminated bite can mean an emergency hospital visit.

Allergy & Diet Guard is a local, privacy-first food safety sentinel. It takes any food description, ingredients list, or menu item and runs a dual-layer open-source safety analysis to:

Detect both direct allergens and obscure hidden derivatives (e.g. flagging arachis oil as a peanut risk).
Highlight high-risk kitchen cross-contamination vectors (shared fryers, woks, cutting boards).
Issue an unmistakable verdict: 🔴 DANGER, 🟡 WARNING, or 🟢 SAFE.
Generate a one-click, bilingual (English & Indonesian) Chef Safety Card that Sarah can show directly to the waiter or kitchen staff.

When I showed the prototype to Sarah, she breathed a huge sigh of relief:

"Finally, something I can use in the basement food court without needing an internet signal or wondering if my medical details are being tracked by an ad network!"

Demo

Live Web App (Hosted on Render): https://hacktoberfest-2026.onrender.com/weekend/
Unified Challenge Portal: https://hacktoberfest-2026.onrender.com/

The app runs as a zero-dependency lightweight web dashboard, deployed live as a Python web service on Render, or runnable locally at http://localhost:8080.

Real-world Test Scenarios:

Test 1: Pad Thai with hidden Peanut oil
- Input: "Pad thai noodles with fried egg, bean sprouts, crushed peanuts, scallions, cooked with arachis oil and fish sauce."
- Verdict: 🔴 DANGER: DO NOT CONSUME
- Detected Hazards: Direct peanuts + obscure alias arachis oil.
- Kitchen Risks: Shared woks and cold-pressed peanut oils.
Test 2: Vegan Stew with hidden Gluten
- Input: "Vegetarian plant-based BBQ bowl with roasted seitan chunks, malt extract sauce, steamed corn, and barley pearls."
- Verdict: 🔴 DANGER: Celiac / Gluten Violation
- Detected Hazards: seitan, malt extract, and barley pearls.
Test 3: Clean Atlantic Salmon
- Input: "Pan-seared Atlantic salmon fillet with extra virgin olive oil, sea salt, cracked black pepper, steamed broccoli, and jasmine rice."
- Verdict: 🟢 SAFE
Chef Safety Card Output (English & Indonesian):

⚠️ ALLERGY NOTICE FOR THE CHEF:
Hello, I am ordering for my friend Sarah.
STRICT ALLERGIES: Peanut / Groundnut, Gluten / Wheat / Celiac Risk, Dairy / Milk / Casein
DIETARY PREFERENCES: None
Please ensure food contains NO traces of these ingredients and clean cookware/utensils are used to avoid cross-contamination. Thank you!

Code

The code is completely open-source and structured for immediate zero-dependency execution:

Repository: github.com/dieennn/hacktoberfest-2026/tree/main/00-weekend-allergy-guard

Project Architecture:

app.py: Standalone Python web server & REST API using purely Python 3 standard library (http.server & socketserver). Zero pip install required.
analyzer.py: Open clinical taxonomy engine mapping FDA Big 9 and EU 14 allergens, cross-contamination rules, and local inference bridge.
static/index.html: Responsive, accessible web UI with friend profile presets and instant safety feedback.
test_analyzer.py: Self-contained assert-based automated test suite.

How I Built It

The core engine is built on a hybrid open-source AI architecture:

Embedded Clinical Semantic Ontology (100% Offline & Deterministic): We codified FDA Big 9 and EU 14 allergens into an open clinical knowledge base with dozens of obscure culinary aliases (e.g. arachis, valencias, seitan, farro, spelt, caseinate, surimi, albumen) and cross-contact risk patterns.
Local Open-Weights LLM Integration (Ollama): When Ollama is running locally (llama3.2, gemma2, or qwen2.5), the system passes the matched hazards to the local open model to generate tailored culinary advice and allergen-free ingredient swaps.
Failsafe Design Ladder: Safety cannot rely on probabilistic guesswork. If either the deterministic clinical rules or the local open model identifies a risk, the system escalates to DANGER.

Why Does Open Innovation Matter?

Why not just use a closed proprietary API like ChatGPT or Claude?

Works Completely Offline in Low-Connectivity Environments: Food courts, basement restaurants, rural markets, and international travel frequently suffer from poor or zero internet connectivity. Closed APIs fail completely. Open-source local inference works everywhere, every time.
Health Data Privacy & Sovereignty: A person's medical allergies and health conditions are strictly sensitive personal data. Running open-source models locally guarantees zero data telemetry and zero tracking.
Zero Cost & Zero Rate Limits: Friends shouldn't have to pay subscription fees or worry about API token credits just to know if their dinner is safe to eat.
Deterministic Failsafe Control: Closed AI models change behind closed doors and frequently suffer from hallucinations on complex ingredient derivatives. With open code and local weights, we enforce strict verification and safety boundaries.

My Agent Session

This project was built and validated during our Hacktoberfest 2026 preparation session using the DevRelay harness.

Prize Categories

Hacktoberfest Open-Source AI Challenge
Build for a Friend Theme
Best Use of Render: Deployed the live unified AI food safety application as a cloud web service on Render, serving both the interactive UI and the real-time inference API.