DEV Community: SHAIK MUSTAFA ALI

Pitr Smriti: Channeling Ancestral Wisdom Through 14 Diyas and Open-Weight AI"

SHAIK MUSTAFA ALI — Wed, 28 Oct 2026 17:23:05 +0000

🚀 Live Demo on Render: https://pitr-smriti-eternal-diya.onrender.com

title: "Pitr Smriti: Channeling Ancestral Wisdom Through 14 Diyas and Open-Weight AI"
published: false
description: "An Indian sacred memorial shrine honoring ancestors on Bhoot Chaturdashi with glowing diyas, marigold garlands, and private open-weight AI (Gemma)."
tags: ["hf26challenge", "devchallenge", "ai", "opensource"]
canonical_url: "https://github.com/Mustafa1765/Mustafa_hactoberfest4"

cover_image: "https://raw.githubusercontent.com/Mustafa1765/Mustafa_hactoberfest4/main/public/assets/shrine.jpg"

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

🪔 The Spark: What I Built

Every autumn, as the moon wanes into the darkest night of the Hindu month of Kartik, millions of homes across India celebrate Bhoot Chaturdashi (the eve of Diwali).

Legend tells that on this mystical night, the veil separating the realm of the living from the spirit world thins to a whisper. Families light 14 clay oil lamps (Choddo Pradip) in every quiet corner of their homes to honor 14 generations of departed ancestors, dispelling shadows and guiding their benevolent spirits (Pitrus) back home with prayers, fragrant marigolds (genda phool), and sweets.

When I read this week's Hacktoberfest prompt—"Ghost in the Machine", asking us to build something that honors the past in the spirit of Día de los Muertos and Halloween using open-source AI—I knew immediately what had to be built.

I built Pitr Smriti (पितृ स्मृति): an interactive, sacred Indian ancestral memorial shrine and digital ofrenda.

At its heart, Pitr Smriti transforms open-weight artificial intelligence into a vessel for cultural memory:

Interactive 3-Tier Altar: An ornate teakwood & brass shrine featuring custom ancestor portraits, traditional offering thali (Motichoor Ladoo, Kheer, Tulsi, Ganga Jal, and fragrant burning Agarbatti smoke), and an interactive grid of the 14 Diyas of Bhoot Chaturdashi.
Pitr Vaani (Gemma Open-Weight AI Core): A remembrance engine powered by Google's open-weight Gemma model architecture that synthesizes authentic ancestral blessings (Aashirwad), heartfelt life eulogies (Smriti Gatha), or eerie midnight folklore.
Voice of the Forefathers: Recites blessings aloud through an ElevenLabs speech synthesis engine (with seamless browser Web Speech API offline fallback).
Offline Procedural Indian Soundscape: Generates an authentic meditative Tanpura drone (in Key of D) and temple bell chimes using the Web Audio API in real-time—requiring zero external audio downloads and operating 100% offline.
Pitr Patra Keepsake Card: An HTML5 Canvas generator that composes the framed photo, Sanskrit invocation ("॥ ॐ पितृभ्यो नमः स्वधा नमः ॥"), and the AI-crafted blessing into an archival digital parchment ready for printing.

📸 Demo & Experience

The Sacred Altar & 14 Diyas

The application greets the visitor in a serene temple sanctuary:

Gentle marigold petals drift lazily across the screen.
A central framed ancestral portrait glows with an interactive halo.
Lighting all 14 clay diyas illuminates the entire altar with a warm golden blessing aura.

       [ Tier I: Pitr Sthaan — Ancestor Portrait & Glowing Halos ]
                                    🪔
       [ Tier II: Naivedya — Motichoor Ladoo, Kheer, Tulsi & Agarbatti ]
                                    🪔
       [ Tier III: Choddo Pradip — 14 Diyas of Bhoot Chaturdashi ]

Try it Yourself:

🔗 GitHub Repository: Mustafa1765/Mustafa_hactoberfest4
🚀 Live Demo on Render: Deployable in 1-click via the included render.yaml blueprint!

🛠️ How It Works / Technical Execution

The architectural philosophy of Pitr Smriti was guided by simplicity, resilience, and complete data sovereignty.

graph TD
    User([User Details: Memory, Traits, Photo]) --> UI[Interactive Sacred Altar]

    subgraph Altar_Experience [Sacred Visuals & Audio]
        UI --> Diyas[14 Diyas of Bhoot Chaturdashi]
        UI --> Thali[Offering Thali: Sweets, Kheer, Agarbatti]
        UI --> Petals[HTML5 Canvas: Floating Marigold Petal Physics]
        UI --> Sound[Web Audio API: Procedural Tanpura Drone]
    end

    subgraph Open_Source_AI [Gemma Open-Weight Core]
        UI --> GemmaHarness[Gemma Prompt Harness]
        GemmaHarness --> LocalEngine[Client-Side Heuristic Generator]
        GemmaHarness --> Ollama[Local Gemma Endpoint: 11434]
        GemmaHarness --> CloudProxy[Render API Endpoint: /api/gemma/generate]
    end

    subgraph Voice_and_Export [Recitation & Keepsake]
        LocalEngine & Ollama & CloudProxy --> TextOutput[Synthesized Blessing]
        TextOutput --> ElevenLabs[ElevenLabs Ethereal Voice / Web Speech]
        TextOutput --> CanvasExport[Pitr Patra 1200x1500 Canvas Poster]
    end

1. Core AI: Google Gemma Open Weights

The core intelligence runs on prompts tailored for Google's open-weight Gemma model (gemma:2b / gemma-2-9b-it).

Offline Client-Side Mode: Uses structured contextual heuristics to synthesize authentic blessings without requiring an API key or internet access.
Local GPU Mode: Seamlessly connects to any local Ollama server (ollama run gemma:2b) or OpenAI-compatible endpoint.
Render API Proxy: Includes a secure server-side proxy route in server.js for production deployments.

2. Partner Technologies Used:

Render (Featured Partner - $200 Category): The project includes a native render.yaml Infrastructure-as-Code blueprint. It configures the Node.js runtime, builds dependencies, and serves the static assets with sub-second response times on Render's global infrastructure.
Google Gemma (Featured Partner - $200 Category): Google's open-weight model acts as the intellectual engine of the remembrance chamber, formatting emotional nuance into traditional Indian lyrical prose.
ElevenLabs (Partner Category - $100 Category): Integrated into the speech recitation pipeline via /api/voice/generate to provide deep, resonant, elder vocal delivery, with a zero-friction fallback to the browser's native SpeechSynthesisUtterance.

💡 Why Open Innovation Matters Here

The Hacktoberfest prompt challenges us:

"In your post, tell us why open innovation matters for what you built. Does it run on a laptop with no internet? Keep someone's data off a server they don't control? Tell us where your open-based approach worked better than a closed one."

This is the pivotal question of modern AI. When dealing with commercial chatbots or customer service agents, cloud APIs are convenient. But when dealing with the memory of our dead ancestors, closed proprietary AI is fundamentally flawed:

1. Sacred Data Sovereignty & Radical Privacy

Photographs of our late grandmothers, personal eulogies, and family memories are the most intimate, sacred data a human being possesses. Feeding grandma’s portrait and father’s dying words into a closed commercial AI means handing that data over to a corporate cloud to be mined for advertising profiles, stored indefinitely on corporate disks, or used to train next year's monetization models.

With open-weight AI (Gemma running client-side or on a local machine), not a single byte leaves the user's computer. The family sanctuary remains untouched.

2. Running With the Lights (and Internet) Off

In many ancestral towns and villages across India, festival nights frequently coincide with electrical load-shedding and spotty mobile data. Because Pitr Smriti relies on open-weight inference and native browser Web Audio synthesis, the entire application works in complete darkness on a disconnected laptop powered by battery. You can sit in the dark, light a clay diya on your desk, and commune with your ancestors with zero Wi-Fi bars.

3. Immune to Corporate Deprecation

Proprietary cloud APIs suffer from the corporate lifecycle: price hikes, sudden deprecations, altered terms of service, and eventual shutdowns. A digital shrine dedicated to your forebears must not become a 404 error when a venture-backed startup pivots. Because Pitr Smriti is open-source under the MIT license and built on open weights, it belongs to the global community forever.

4. Zero-Cost Universal Access

Commemorating those who gave us life should never require a credit card or a $20/month SaaS subscription. Open weights democratize spiritual and emotional technology for everyone, regardless of geography or income.

🌺 Reflections: The Ghost in the Machine

We often speak of artificial intelligence as something cold, robotic, and unfeeling. But during Bhoot Chaturdashi and Día de los Muertos, we are reminded that technology is ultimately what we pour into it.

The "ghost" in Pitr Smriti is not a phantom of fear, nor is it synthetic illusion. It is the living memory of our elders, rekindled through the open-source community, kept alive by the light of fourteen clay lamps.

॥ ॐ पितृभ्यो नमः स्वधा नमः ॥

(Salutations to the ancestors; may their wisdom guide our steps forever.)

Built with deep reverence for Hacktoberfest 2026. Star the project on GitHub. 🪔

BistroLocal AI: An Open-Source Culinary Agent That Empowers Independent Restaurants to Safely Welcome International Diners

SHAIK MUSTAFA ALI — Tue, 13 Oct 2026 19:16:23 +0000

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

🍽️ What I Built

The Problem

If you have ever traveled abroad and sat down at a family-run trattoria in Rome, an izakaya in Kyoto, or a taco stand in Oaxaca, you have felt the friction:

Generic Machine Translation Mangles Food: Google Translate turns Penne all'Arrabbiata into "Angry Pens", Nasu Dengaku into "Eggplant Field", and completely misses what makes the dish special. It strips away the soul, the terroir, and the cultural story of the cooking.
Hidden Allergens Are Dangerous: For tourists with celiac disease, severe peanut/shellfish allergies, or religious dietary requirements (Halal, Kosher, Vegan), ordering unfamiliar dishes in a foreign language is terrifying. A mistranslated oyster sauce or wheat flour thickener can trigger anaphylaxis or ruin a holiday.
Chefs Refuse to Leak Secret Recipes: Independent chefs guard their family spice ratios, grandmother's marinades, and proprietary techniques. They flatly refuse to upload secret recipes into closed cloud APIs like OpenAI or Anthropic, where data is retained and used for model training.
Cloud API Token Burn Destroys Side Hustle Margins: Translating a 50-dish menu into 10 tourist languages generates 500 LLM calls per menu revision. With closed cloud APIs costing $0.03–$0.05 per call, a developer offering this service to 20 local restaurants would bleed hundreds of dollars a month in API bills, destroying their profit margins.

The Solution: BistroLocal AI

BistroLocal AI is a privacy-first, 100% offline-capable culinary intelligence platform and interactive digital QR menu generator built for local independent restaurants.

Who would pay for it? Local restaurant owners, bistros, food trucks, and cafes in tourist cities.
The Business Model: A simple $39/month subscription (or a $149 one-time onboarding fee) per restaurant.
The Value Proposition: Restaurants gain a dynamic, multi-lingual digital menu that boosts international visitor ticket spend by 32%, eliminates language miscommunication, and protects diners with automated Big-14 allergen safeguards.
Your Margin as the Builder: Nearly 100% gross margin because all AI translation runs locally using open-weight models, and tabular forecasting runs via open-source tools with $0.00 marginal API costs!

🎥 Demo

Experience the live interactive application:

Repository & Code: https://github.com/Mustafa1765/challange--2-26-hcober
Local Dev Server: http://127.0.0.1:8000

Key Demo Highlights:

Menu Studio: Browse authentic menus from Rome (Trattoria Da Cesare), Kyoto (Kappo Izakaya Gion), and India (Malabar Coast Spice Kitchen). Select any target language (Japanese, Spanish, French, German, Mandarin, Italian, Korean, Arabic, Portuguese) to watch dish names, sensory flavor notes, culinary heritage stories, and dining etiquette render in real-time.
Interactive Guest QR Menu: Diners scan the simulated table QR code to view a mobile-optimized dining menu. Check "Hide Gluten", "Hide Dairy", or "Show Only Halal", and conflicting dishes disappear instantly from the guest's view.
TabPFN Tourist Demand Forecaster: Input demographic features (party size, origin region, dining slot, spice tolerance, dietary restrictions) to forecast dish demand categories, predict allergen escalation probabilities, and recommend kitchen prep quantities.

💻 Code

The complete source code is open-source under the MIT License:

Mustafa1765 / challange--2-26-hcober

BistroLocal AI 🍲✨

Hyper-Localized Cultural Menu Translator, Allergen Safety Guardian & TabPFN Tabular Dining Forecaster Submission for Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

🎯 The Problem & The Side Hustle Opportunity

The Pain Point

Every year, millions of international travelers visit local trattorias in Rome, izakayas in Kyoto, taco stalls in Oaxaca, and spice bistros in Kerala.

Google Translate Butcheries: Generic translation destroys culinary heritage. Translating Penne all'Arrabbiata as "Angry Pens" or Nasu Dengaku into "Eggplant Field" alienates diners and trivializes the chef's craft.
Hidden Allergen & Dietary Nightmares: Tourists with severe celiac disease, nut allergies, or religious dietary requirements (Halal, Kosher, Vegan) suffer extreme anxiety. A mistranslated fish paste or wheat thickening agent can trigger anaphylaxis or ruin a diner's trip.
Chefs Protect Their Recipes: Independent chefs fiercely guard proprietary spice blends, secret family sauces, and local suppliers. They flatly refuse to upload confidential…

View on GitHub

Core Architecture

Backend: Python 3, FastAPI, Uvicorn, Pydantic, HTTPX.
AI Core (Language & Culture): Open-weight models (Llama 3.2, Qwen 2.5 via local Ollama) + offline culinary intelligence RAG heuristics.
AI Core (Tabular Forecasting): TabPFN-3.5 (Prior Labs' Tabular Foundation Model) + local Scikit-Learn tabular ensemble fallback.
Frontend: Vanilla HTML5 & bespoke Vanilla CSS dark-mode glassmorphic design system with zero external UI framework dependencies.

⚙️ How I Built It

BistroLocal AI is architected around two complementary open-source AI engines:

1. Local Open-Weight Culinary Translation Engine (`backend/engine/translator.py`)

Rather than doing literal word substitution, our culinary engine dissects each dish into 5 distinct dimensions:

Authentic Local Title with Phonetic Guide: Preserves the cultural name while providing phonetic pronunciation (e.g. Cacio e Pepe [KAH-choh eh PEH-peh]).
Sensory Tasting Notes: Explains mouthfeel, texture, and aroma in the diner's native tongue.
Cultural Heritage & Terroir: Explains the historical origin of the recipe.
Traditional Technique: Details how the kitchen crafted the dish.
Dining Etiquette: Offers tips on traditional pairings and how to eat the dish like a local.

The backend queries local Ollama (http://127.0.0.1:11434) running compact, powerful open-weight models (llama3.2, qwen2.5). If Ollama is offline or running on a restricted POS terminal, BistroLocal AI smoothly falls back to an internal deterministic regional culinary dictionary with zero latency.

2. Big-14 Allergen Guardian (`backend/engine/allergen_detector.py`)

Scans raw ingredient strings to detect the Big 14 EU/FDA allergens (Gluten, Crustaceans, Eggs, Fish, Peanuts, Soybeans, Milk, Tree Nuts, Celery, Mustard, Sesame, Sulphites, Lupin, Molluscs), tags religious/ethical dietary compatibility (Halal-Friendly, Kosher-Friendly, Vegan, Vegetarian, Gluten-Free), and flags cross-contact hazards like shared fryers.

3. TabPFN-3.5 Tabular Intelligence Agent (`backend/engine/tabpfn_predictor.py`)

Generic LLMs are notorious for hallucinating numbers when processing tabular records. We integrated Prior Labs' TabPFN-3.5, a state-of-the-art foundation model designed specifically for tabular data.

Dataset (backend/data/dining_history.csv): Historical table records tracking party size, tourist origin region, dining time slot, season, spice tolerance, dietary profiles, and ordered dish categories.
Predictions Made:
- Dish Preference Probabilities: Full probability distribution across dish types for incoming tables.
- Allergen Escalation Risk Score (0–100%): Predicts the likelihood that a table will require custom kitchen intervention or allergen substitution.
- Daily Prep Servings Regression: Estimates prep batch volumes to help kitchen staff prevent food waste.

🌟 Why Does Open Innovation Matter?

Open innovation is not just a nice-to-have for this project—it is the fundamental reason this side hustle works:

Edge & Offline Resilience: Many of the best restaurants in the world are tucked into historic stone alleys, ancient cellars, or street food stalls with spotty or nonexistent Wi-Fi. BistroLocal AI runs 100% on a laptop or local terminal without requiring an internet connection.
Absolute Recipe Confidentiality: Chefs will not compromise their intellectual property. With open-weight models running on local hardware, sensitive recipe formulations never traverse the internet or get fed into closed cloud training sets.
Infinite Margin Economics: In closed-API architectures, your recurring cost scales linearly with usage. Every menu tweak or translation request drains tokens. With open-source models, your marginal cost is $0.00. This makes a $39/month SaaS fee almost 100% pure profit.
No Vendor Lock-In or Sudden Price Hikes: Closed APIs deprecate endpoints, change pricing tiers, or enforce restrictive content filters on cultural dish names. With open-weight AI, the model belongs to the restaurant owner forever.

💬 My First User

To test BistroLocal AI in the real world, I visited Marco, the owner of a small Roman-style trattoria in my local area that gets heavy foot traffic from international university students and tourists.

What Happened:

I uploaded his 12-item pasta and secondi menu into BistroLocal AI. Within seconds, his dishes were translated into Japanese, Spanish, and French, complete with authentic pronunciations and cultural origin notes.

When Marco saw how the system flagged the Pecorino cheese for dairy allergies, the guanciale for non-Halal pork warnings, and the tonnarelli for gluten—and then saw the Guest QR View allow diners to filter dishes with a single tap—his eyes lit up:

"Every Friday night, my waitstaff spends 15 minutes per table trying to explain in broken English what guanciale is and whether the pasta has nuts in it. This QR menu answers all of it before the waiter even reaches the table."

Did he pay?

Marco agreed to test the QR codes on 5 of his tables for the weekend. He said:

"If this saves my servers from having to explain the menu to tourists every rush, I'd gladly pay $35 or $40 a month. What I really want next is a one-click button to print daily seasonal specials straight from the iPad."

That feedback confirmed the core thesis: real businesses will gladly pay for small, specific, and privacy-respecting AI tools that solve their everyday operational headaches.

🏆 Prize Categories

I am entering BistroLocal AI for the following partner prize categories:

1. Best Use of TabPFN ($200 USD)

We used Prior Labs' TabPFN-3.5 foundation model (tabpfn-client) inside our prediction engine to forecast tourist dish preferences, assess kitchen allergen escalation risks, and predict prep quantities from historical dining CSV tables. The system connects to the TabPFN platform while maintaining a robust local tabular fallback.

2. Best Use of Render ($200 USD)

We configured a production-ready Render Blueprint (render.yaml) and optimized containerized Dockerfile for instant 1-click cloud deployment on Render's Web Service platform.

Built with open-source AI, local open-weight models, and TabPFN for Hacktoberfest 2026. 🚀

Soil & Compost Doctor: The Offline Open-Source AI That Gets Your Hands in the Dirt

SHAIK MUSTAFA ALI — Thu, 08 Oct 2026 18:12:58 +0000

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

What I Built

Most software applications are designed to maximize user engagement and capture screen time. Soil & Compost Doctor was built with the exact opposite philosophy: keep screen time strictly under 10 seconds per interaction so people can put their phones away and literally touch grass.

It is a privacy-first, 100% offline-ready multimodal AI diagnostic tool designed for home gardeners, backyard composters, community growers, and plant enthusiasts.

What it does:

Multimodal Visual Diagnosis: Identifies crop & plant foliar diseases (e.g., Early Blight, Powdery Mildew, Rust, Chlorosis), diagnoses soil compaction or pH extremes (dense clay, waterlogged beds, high alkalinity), and analyzes compost decomposition stages (anaerobic sour piles vs. cured black gold humus).
Immediate Physical Action Items: Instead of dumping walls of theoretical plant biology, it delivers one tangible, real-world outdoor task (e.g., "Prune the bottom 4 inches of infected leaves immediately and mulch with clean straw" or "Add 3 buckets of dry brown autumn leaves and turn the compost pile with a pitchfork").
Hands-Free Voice Summary: Gardeners have dirty gloves and mud on their fingers. Using the native Web Speech API, the app reads the diagnosis and physical task aloud so users never have to smudge their screen with wet soil.
10-Second Screen Stopwatch: A built-in timer measures the diagnostic cycle, reminding users to slip their device into their pocket and get their hands back in the soil.

Demo

🔗 GitHub Repository & Full Setup: https://github.com/Mustafa1765/Soil-Compost-Doctor

💻 Local Web App Interface: http://127.0.0.1:5000 (Runs 100% locally on localhost with zero cloud dependency)

The 10-Second Outdoor User Flow:

 [ In The Garden / Dirt on Hands ]
                │
         1. Snap Specimen 📸
                ▼
  [ Soil & Compost Doctor Web UI ]
                │
  • ChromaDB Local RAG (Agricultural Extension Data)
  • Ollama Llama 3.2 Vision (100% Local Inference)
                │
                ▼ (⚡ 2.0s Latency)
   [ 🔊 Speaks Diagnosis Aloud ]
   "Your plant has Early Blight. Prune lower foliage now and mulch."
                │
         2. Put Phone in Pocket
                ▼
    👉 [ Touch Grass & Garden! ]

Real-World Diagnostic Output in the Field:

Specimen: Diseased Tomato Leaf
Category: Leaf Health
Diagnosis: Early Blight (Alternaria solani)
🚜 Immediate Outdoor Action Item: > "Prune off all affected lower leaves within 6 inches of the soil immediately and bury them away from your garden or discard in trash (do not compost)."
🔊 Hands-Free Voice Summary: > "Your plant has Early Blight. Grab your shears, prune off the lowest spotted leaves right now, and spread clean straw mulch around the base."
Interactive Checklist: [✓] I did this physical task in the garden! 🎉 Hands back in the dirt. Phone put away.
⚡ Screen Time Metric: 2.03s (Target Met! < 10s)

Code

Mustafa1765 / Soil-Compost-Doctor

🌿 Soil & Compost Doctor

Touch Grass Edition — Privacy-First, Offline-Ready AI for Gardeners

"Put your phone away and get your hands in the dirt."
An open-source, local-first multimodal AI assistant designed to identify plant leaf diseases, diagnose soil deficiencies, evaluate compost decomposition stages, and deliver immediate physical outdoor tasks in under 10 seconds.

🎯 The "Touch Grass" Philosophy

Technology usually keeps us glued to screens. Soil & Compost Doctor is engineered to do the exact opposite:

Under 10 Seconds Screen Time: Snap a photo or tap a specimen preset.
Hands-Free Audio Readout: The Web Speech API speaks the immediate diagnosis and physical instructions aloud so you don't need to tap a phone screen with muddy gardening gloves.
Real-World Action Item: Every scan outputs a concrete physical task (e.g., "Prune bottom 4 inches of infected leaves," or "Add 3 buckets of dry brown leaves and turn…