👻 GhostWrite: A Local AI Ghost Story Generator

Akshat Maheshwari — Sat, 31 Oct 2026 19:44:43 +0000

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

What I Built

👻 GhostWrite — A Local AI Ghost Story Generator

GhostWrite is a small creative AI application that turns simple ideas into atmospheric ghost stories.

Users can customize three elements:

Setting: Where the story takes place
Main character: Who the story follows
Mysterious object: An object that drives the mystery

Based on these inputs, GhostWrite generates a unique ghost story that users can read and download.

I wanted to explore how open-source AI could be used to create something fun and interactive without depending on paid AI APIs.

Demo

Generating a Story

GhostWrite allows users to customize their story before generation.

GhostWrite generating a story using custom user inputs.

Generated Story

The application generates an atmospheric story based on the selected inputs, with an option to download the result.

A completed AI-generated ghost story, ready to download.

Code

GitHub Repository: GhostWrite

The repository contains the application source code, dependencies, and instructions for running the project locally.

How I Built It

I built GhostWrite using Python and Streamlit, integrating the TinyLlama 1.1B Chat model through Hugging Face Transformers.

The application follows a simple workflow:

Collect the user's story inputs.
Combine them into a structured prompt.
Generate a story using the locally running language model.
Display the generated story in the interface.
Allow users to download their story as a text file.

The main technologies used were:

Python
Streamlit
Hugging Face Transformers
TinyLlama 1.1B Chat

The model runs locally, so the application does not require an external AI API key.

Why Does Open Innovation Matter?

Open innovation makes AI experimentation more accessible.

Building GhostWrite with an open-weight model allowed me to explore AI-powered text generation without relying on proprietary APIs or paid inference services.

It also gave me the opportunity to experiment with prompts, generation parameters, and local inference while keeping the application simple enough for others to run and modify.

For me, this project was a small but meaningful step toward understanding how open-source AI can make creative technology more accessible to developers.

The Spirit of It

GhostWrite started with a simple idea: what if a few ordinary details could become the beginning of an unsettling story?

A broken radio, an abandoned railway station, and a night-shift engineer are enough to get the imagination going.

The goal was to make storytelling feel interactive, letting users create the ingredients of their own little nightmare.

Thanks

Built for the Hacktoberfest 2026 Open-Source AI Challenge — Week 4: Ghost in the Machine.

AI Resume Summarizer – Built for a Friend (#hf26challenge)

Akshat Maheshwari — Mon, 05 Oct 2026 06:06:44 +0000

🔗 Project Links

Live Demo: resume-summarizer.streamlit.app
GitHub Repository: https://github.com/AkshatMaheshwari27/resume-summarizer

1. What I Built & Who It's For

I built AI Resume Summarizer for a friend and fellow student who was struggling to condense their lengthy academic resume into a sharp, 2–3 sentence summary for job applications and recruiter cold-outreach emails.

The app takes raw resume text and uses open-source AI to generate a professional overview covering candidate background, technical skills, and core strengths.

2. Why Open Innovation Matters

Resumes contain sensitive personal information (phone numbers, project links, personal history). Using an open-weight model like meta-llama/Llama-3.1-8B-Instruct through Hugging Face open inference keeps processing transparent and avoids sharing confidential career data with closed, proprietary API providers.

3. How It Works & Tech Stack

AI Core: meta-llama/Llama-3.1-8B-Instruct via huggingface_hub InferenceClient
UI & Web Framework: Streamlit
Deployment: Streamlit Community Cloud
Secrets & Security: API keys handled securely via st.secrets without hardcoding.

4. What My Friend Thought

After testing it with their resume, my friend went from spending 20 minutes tweaking summary bullet points to getting a recruiter-ready executive summary in seconds.

DEV Community: Akshat Maheshwari