DEV Community: Aditi Patel

The Last Echo: An Offline AI That Keeps Memories Alive

Aditi Patel — Sat, 31 Oct 2026 06:29:39 +0000

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

What I Built

The Last Echo – An Offline AI Memory Keeper

What if the stories of the people we love never had to disappear?

For Ghost in the Machine, I wanted to explore something beyond traditional horror. I wanted to build something that feels a little mysterious but carries a deeper human meaning.

The Last Echo is an AI-powered digital memory keeper designed to preserve the stories, memories, photographs, and personal experiences of loved ones.

Users can add memories, write personal stories, preserve meaningful conversations, and organize important moments. They can then interact with an AI assistant that retrieves relevant memories and responds using the information they have preserved.

Inspired by Día de los Muertos, the project is built around one simple idea:

Technology should not just remember information. It should help us preserve the stories that make us human.

The AI doesn't claim to bring people back or actually recreate someone's consciousness. Instead, it helps people revisit the memories and stories they've chosen to preserve.

Demo

Coming soon!

A demonstration will showcase:

Adding and preserving personal memories.
Organizing stories into a searchable memory collection.
Asking questions about saved memories.
Receiving context-aware responses from a locally running AI model.
Accessing the experience without an internet connection after initial setup.

Code

GitHub repository: Coming soon.

The project source code, installation instructions, architecture, and implementation details will be published here.

How I Built It

The core idea behind The Last Echo is to make open-source AI the engine of the entire experience rather than simply connecting to a commercial AI API.

Planned technology stack:

Ollama: Local inference and model management.
Llama or another suitable open-weight model: Powers the conversational memory assistant.
Python: Backend logic and memory processing.
LangChain: Helps orchestrate retrieval and language-model interactions.
ChromaDB: Stores and retrieves memory embeddings.
Sentence Transformers: Generates embeddings for semantic memory search.
Streamlit: Provides an interactive user interface.

The architecture follows a Retrieval-Augmented Generation (RAG) approach.

When someone saves a memory, the application converts it into an embedding and stores it locally. When a user asks a question, the system retrieves relevant memories and provides them as context to the local language model.

This helps the assistant generate responses grounded in the information the user actually provided.

The long-term goal is a completely local experience where personal memories never need to be sent to an external AI service.

Why Does Open Innovation Matter?

For a project dealing with personal memories, privacy isn't an optional feature. It's the foundation.

Imagine preserving your family's photographs, personal stories, and private conversations in a system that sends everything to an external AI provider.

That creates a difficult question: Who really owns those memories?

Open-source AI gives us another path.

By using open-weight models and local inference, The Last Echo can be designed to:

Keep personal memories on the user's own machine.
Work without an internet connection after downloading the required models and dependencies.
Allow users to choose, replace, or customize their language models.
Avoid depending on a single commercial AI provider.
Give developers control over how personal information is processed.

A closed API might make the initial development easier, but local open AI makes the project's privacy-first philosophy possible.

For me, that's the real value of open innovation: having the freedom to build technology around people's needs instead of forcing people to adapt to a service's limitations.

My Agent Session

Coming soon!

I plan to share the development process and agent session through DevRelay.

The Spirit of It

There is something haunting about forgetting.

A photograph can fade. A voice recording can get lost. A story that someone tells us today might disappear if nobody remembers to tell it again.

Día de los Muertos reminds us that remembering someone is also a way of keeping their presence in our lives.

The Last Echo explores that idea through technology.

Imagine asking:

"What was Grandpa's favorite story about his childhood?"

And receiving an answer based on a story that your family personally preserved.

It isn't the person speaking again.

It's their story finding its way back to you.

Some echoes deserve to last forever.

Prize Categories

Primary challenge entry:

Hacktoberfest Open-Source AI Challenge – Week 4: Ghost in the Machine.

Partner categories: To be confirmed based on the final implementation and actual partner technology used.

Built with the belief that the best use of AI isn't always creating something new. Sometimes, it's helping us hold on to what matters. -->

PocketCFO

Aditi Patel — Mon, 19 Oct 2026 06:31:27 +0000

*# PocketCFO AI: A Private AI Finance Assistant for Freelancers

Built for Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

💡 What I Built

Freelancers often have to manage invoices, track payments, organize expenses, and prepare financial reports on their own. Many existing tools can be expensive, complicated, or require uploading sensitive financial documents to external services.

PocketCFO AI is a local-first AI finance assistant designed to make bookkeeping simpler for freelancers and independent professionals.

It helps users extract information from invoices, manage invoice records, track income and expenses, monitor outstanding payments, and export financial data for further analysis.

The goal is simple: give freelancers a more accessible way to understand their finances without needing a dedicated finance team.

Who might pay for it?

Freelancers managing multiple clients and invoices.
Independent consultants tracking payments and expenses.
Solo founders who need lightweight financial organization.
Small service businesses looking for a simpler bookkeeping workflow.

🎬 Demo

Live Demo: [https://github.com/gurjaraditipatel-blip/pocketcfo-ai.git]

Video Demo: [Add your demo video link, if available]

The application currently runs locally using Streamlit and Ollama.

💻 Code

GitHub Repository: [https://github.com/gurjaraditipatel-blip/pocketcfo-ai.git]

The project is built around a practical idea: AI-powered financial assistance should be useful, accessible, and privacy-conscious.

🛠️ How I Built It

PocketCFO AI combines open-weight AI with a lightweight Python application.

My stack includes:

Python — core application logic.
Streamlit — interactive user interface.
Ollama — local model inference.
Qwen2.5 3B — open-weight language model used for invoice information extraction.
SQLite — local financial data storage.
PyMuPDF — PDF text extraction.
CSV export — portable financial records and reporting.

The workflow is designed to extract useful information from invoice documents, validate the extracted fields, and help users manage invoices and payment records.

I also implemented invoice payment tracking, transaction management, dashboard metrics, and financial reporting workflows.

Rather than building a generic chatbot, I wanted to build something that solves a specific, recurring problem for people who work independently.

🌱 Why Does Open Innovation Matter?

Financial information is sensitive. Freelancers should have more control over how their invoices and financial records are processed.

Open-weight AI and local inference make it possible to build tools that can run on a user's own machine instead of requiring every document to be sent to a paid, closed AI API.

For PocketCFO AI, this approach offers several advantages:

Privacy and control: Local inference can reduce the need to send financial documents to external AI providers.
Cost flexibility: Users can run the model locally without paying a per-request fee to a hosted model API, although local hardware and electricity still have costs.
Less vendor lock-in: The model and inference setup can be changed as the project evolves.
Customization: Developers can inspect, adapt, and extend the application for different bookkeeping workflows.
Accessibility: Open tooling makes it easier for independent developers to experiment with AI-powered products.

Open innovation is not just about making code available. It is about giving builders the freedom to adapt technology to real problems and create products that people can control and improve.

🤖 My Agent Session

[Add your DevRelay agent session link here, if available.]

🏆 Prize Categories

[Add the applicable partner prize categories from the challenge page, or remove this section if none apply.]

PocketCFO AI is an ongoing project, and the next step is to put it in front of real freelancers, collect feedback, and learn which features are valuable enough that people would actually pay for them.

Touch-Grass

Aditi Patel — Sat, 10 Oct 2026 10:24:14 +0000

🌱 Touch-Grass AI — Your Digital Detox Companion

Touch Grass. Touch Life. Feel Better. 🌿

🔗 Live Demo: touchgrass-ai.streamlit.app

TouchGrass AI is a fun, interactive web application designed to encourage people to take a break from their screens, spend more time outdoors, build healthy daily habits, and reconnect with the real world.

Instead of endlessly scrolling, turn your outdoor time into an adventure! Complete daily missions, track your progress, and make every day a little greener. 🌍

✨ Features

🏡 Daily Missions — Discover fun activities that encourage you to step outside.
🎮 Play Together — Explore interactive activities and challenges.
🎯 Daily Goals — Set outdoor time targets and work toward healthier habits.
📊 My Journey — Track your progress with interactive checklists and a 100-point scoring system.
🗺️ Adventure Planner — Get ideas for outdoor activities and plan your next adventure.
🌿 Fun & Interactive UI — Enjoy a colorful, beginner-friendly experience.
📥 Download Your Plan — Save your outdoor adventure plan for later.

🎯 Why TouchGrass AI?

Spending too much time in front of screens can make it easy to forget the world outside. TouchGrass AI encourages small, achievable steps toward a more balanced digital lifestyle.

Whether it's taking a short walk, enjoying nature, or spending time with friends, every little step counts!

🛠️ Tech Stack

Python — Application logic
Streamlit — Interactive web interface
ollama
cloud code
Git & GitHub — Version control and collaboration

🚀 Run Locally

Want to run TouchGrass AI on your own machine?

1. Clone the repository

git clone https://github.com/gurjaraditipatel-blip/touchgrass-ai.git
cd touchgrass-ai

2. Create a virtual environment

python -m venv .venv

3. Activate the environment

Windows PowerShell:

.venv\Scripts\Activate.ps1

macOS / Linux:

source .venv/bin/activate

4. Install dependencies

pip install -r requirements.txt

5. Run the application

streamlit run app.py

The app will open in your browser, usually at http://localhost:8501.

📸 Screenshots

Screenshots of the application can be added here to showcase the interface.

TouchGrass AI Home

🌍 Hacktoberfest & Contributions

Contributions are welcome! 💚

If you'd like to improve TouchGrass AI, you can help by:

🐛 Fixing bugs and improving existing features.
🎨 Enhancing the UI and user experience.
🌱 Adding new outdoor challenges and daily missions.
📈 Improving progress tracking and goal-setting features.
📚 Improving documentation.
♿ Making the app more accessible.
🧪 Adding tests and improving code quality.

💡 Future Improvements

💾 Persistent user progress and saved goals.
🏆 Badges, streaks, and achievement tracking.
🌤️ Weather-aware outdoor activity suggestions.
👥 More multiplayer challenges.
📱 Improved mobile experience.
🤖 Optional AI-powered personalized activity recommendations.

🔒 Privacy & Data

TouchGrass AI is designed to be simple and accessible. The current version uses in-app logic for its activities and scoring. Progress may not persist after a session or app restart.

Avoid entering sensitive personal information into the app.

💚 Our Mission

We believe technology should help us live better—not keep us glued to our screens.

Take a break. Step outside. Touch grass. 🌱

Made with 💚 for a healthier digital lifestyle.

⭐ If you like this project, consider giving it a star on GitHub!

🔗 Try it now: TouchGrass AI — Live App

intro

Aditi Patel — Sun, 04 Oct 2026 11:49:58 +0000

i am aditi patel. i am software engineer. excited to learning new things

Intro

Aditi Patel — Sat, 03 Oct 2026 05:43:33 +0000

I am new to community , happy to be the part of dev .