/mochi-flashback

A half-baked clone of Recall

Primary LanguagePythonMIT LicenseMIT

Mochi Flashback

A half-baked clone of Microsoft Recall, built in Python, for fun.

Mochi Flashback was created as part of a live coding session on Twitch. The goal was to build a simple clone of Microsoft and see how far i could get in a few hours across a week.
I have no intention to maintain this project.

Microsoft Recall is a new feature in the coming Windows that allows users to search across any content they've seen on their PC.
Microsoft takes regular screenshots and has them available for searching with a natural language interface.

🎥 Streams

Follow me on YouTube or Twitch.
New content coming soon...

Day 1 to 5 of Mochi Flashback streams playlist
Follow the development of the code available in this repo.
Mochi Flashback Playlist: Day 1 to 5

🙏 Thanks to

https://solara.dev/ for developing this nice framework!

🛠️ Installation

There are no promises made this would run and I have only tested on macOS (Macbook Air M1) with python 3.11.8 I did not freeze the requirements either.

Create the virtual environment:

# Recommended
$ python -m venv .venv

Install the dependencies:

$ pip install solara pillow openai

🏃 Running

1. Add screenshots 🌉

Screenshots should go into the data folder as seen in:
Screenshot Folder

⚠️ Unfortunately I did not get to taking the screenshots automatically.

2. Process data 🧮

Process the data:

$ python process_screenshots.py

This will create datastores with the processed data in either local_stores or openai_stores folders.

3. Run recall 💬

Here's were we use Solara:

$ solara run recall.py

4. Interact with the UI 🧑‍💻

Here are sample results:
Results

License

LICENSE