DEV Community: NelsonLAN

How a non-CS student approaches "Haunting Engine While You Slept"

NelsonLAN — Thu, 29 Oct 2026 16:03:03 +0000

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

What I Built

While You Slept — a minimal, privacy-focused open-source web application that converts overnight smart-home sensor logs into creepy, personalized second-person ghost stories.

The Inspiration & Theme Interpretation

Smart home sensors (motion detectors, thermostats, smart locks, camera face-recognition routines) are constantly measuring quiet physical spaces.

Every minor glitch—a sudden temperature drop, a flickering light, or an unassigned face detection confidence score—is usually dismissed as background noise.

For the "Ghost in the Machine" challenge, I flipped this dynamic: what if every recorded sensor event isn't an error, but clear physical evidence of an unseen entity moving through your home while you sleep?

While You Slept parses logbook outputs from Home Assistant (or custom pasted logs) and misinterprets mundane ambient device activity through a supernatural narrative lens.

How I Built It

The project is architected around local-first inference, open-weight language models, and zero build-step front-end standards:

Why Does Open Innovation Matter?

Local Data Privacy: Smart home logbooks contain highly intimate data—when you wake up, when you go to sleep, and room-by-room activity. A closed-source API requires transmitting sensitive private telemetry to third-party cloud servers. Open innovation enables running high-performance local open-weight models (like Llama 3.1) entirely on private hardware without sending home logs over the internet.

Uncensored Atmospheric Creative Control: Closed commercial LLM APIs often apply aggressive content moderation safety filters that can flag dark, eerie, or uncanny horror storytelling as unsafe or disturbing. Running open-weight models locally allows developer control over sampling parameters, system prompts, and creative tone.

Demo

Live Interactive Page: Open index.html locally in any browser or visit your self-hosted instance.

Try Sample Logs: Click "Load example log" on the home screen to instantly analyze a sample 8-hour log, adjust the temperature parameter (creativity level), and generate the horror narrative.

Home Assistant Integration: Connect your local Home Assistant instance by updating CONFIG.HA_URL and CONFIG.HA_TOKEN in app.js to automatically fetch last night's logs via /api/logbook.

Code

nelsonlan01 / Hacktoberfest-2026-Haunting-Engine-While-You-Slept

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

Hacktoberfest-2026-Haunting-Engine-While-You-Slept

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

View on GitHub

How a non-CS student views this lazy message reader problem

NelsonLAN — Thu, 22 Oct 2026 15:13:57 +0000

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

What I Built

How a non-CS student views this lazy message reader problem

Demo

How I Built It

Discord side: Create a bot in the Discord developer portal, enable Message Content Intent, and invite it to the server. Use the bot token, not your own account token

Then paste in the channel IDs you want read (turn on Developer Mode in Discord to copy them).

In its place are a model server URL and a model name, prefilled with Ollama’s defaults. It uses the OpenAI-style /v1/chat/completions format, so Ollama, LM Studio, llama.cpp server and vLLM should all work.

How a non-CS student approaches the topic of "A menu translator for a local restaurant that handles their specific dishes better than a generic tool"

NelsonLAN — Thu, 15 Oct 2026 17:38:35 +0000

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

What I Built

Topic:

A menu translator for a local restaurant that handles their specific dishes better than a generic tool.

Research Aim:

To examine how a non-CS student (Digital Media, with a low-code background) can implement this topic.

Why Does Open Innovation Matter?

【A. Pain Points / Why Generic Translators Fall Short】

Literal Naming: Tools like Google Translate convert regional slang, or speciality cooking methods word-by-word, leaving you with confusing names like "husband and wife lung slices" or parts that make no sense.
No Preparation Info: Generic apps rarely explain how a dish is cooked (e.g., braised, stir-fried, clay-pot) or what flavour profile to expect (sour, numbing-spicy, sweet-and-sour).

【B. What Specialised Menu Tools Do Differently】

Users and owners have UX concerns.

Dish-Level Context: Apps built for food translation explain the actual ingredients, regional background, and flavour profiles.
Allergy and Dietary Filters: Dedicated tools let you flag common allergens (shellfish, peanuts, gluten, dairy) or dietary restrictions to warn you or fade out unsafe options.
Restaurant-Facing Publishing Tools: If you are a restaurant owner looking to publish your own custom menu, specialised workflows.

【Creative Coding Technologies】

Applied some JSON structure to structure it to indicate how the dish is prepared, what animal (+parts) is in it, actual allergy flags, etc other concerns.
Used minimal HTML, CSS, and JS as local web-based apps; can be expanded into Server-Client hosting by exporting into HTML and JSON structure storage.

Artefact / Demo / Code

nelsonlan01 / Hacktoberfest-2026-Project-Menu

Hacktoberfest 2026 Menu Translator Demonstration Repo

Topic : [Example Topic] A menu translator for a local restaurant that handles their specific dishes better than a generic tool.

【Why Generic Translators Fall Short】

Literal Naming: Tools like Google Translate convert regional slang, local protein cuts, or speciality cooking methods word-by-word, leaving you with confusing names like "husband and wife lung slices" or literal animal parts that make no sense.
No Preparation Info: Generic apps rarely explain how a dish is cooked (e.g., braised, stir-fried, clay-pot) or what flavour profile to expect (sour, numbing-spicy, sweet-and-sour).

【What Specialised Menu Tools Do Differently】

Dish-Level Context: Apps built for food translation explain the actual ingredients, regional background, and flavour profiles.
Allergy and Dietary Filters: Dedicated tools let you flag common allergens (shellfish, peanuts, gluten, dairy) or dietary restrictions to warn you or fade out unsafe options.
Restaurant-Facing Publishing Tools: If you are a restaurant owner looking to publish your own custom…

View on GitHub

https://nelsonlan01.github.io/Hacktoberfest-2026-Project-Menu/