Inspiration
File Fit started with a simple problem I faced myself: my Downloads folder was becoming messy, with multiple files having similar names and related files scattered across different formats. I wanted a way to automatically understand which files belonged together instead of manually sorting everything.
What started as a personal tool eventually became a product designed to continuously keep files organized.
What it does
File Fit is a real-time file organizer that groups related files based on their names rather than simply sorting them by file type.
When a file is added or removed, File Fit detects the change and automatically updates the organization. It can place a new file into an existing group or create a new group when necessary.
Its three main advantages are:
- Real-time: It continuously reacts to changes in the folder.
- Offline: It does not require LLMs, external AI APIs, or an internet connection.
- Lightweight: It primarily analyzes file names instead of reading the complete contents of every file.
A PDF, DOCX, PNG, or any other format can belong to the same group if the files are related. File Fit organizes files based on what they represent, not what format they use.
How we built it
File Fit uses TF-IDF vectorization to convert file names into numerical representations and HDBSCAN to identify groups of related files.
We use Watchdog to monitor the filesystem for changes. When a file is created, modified, or removed, File Fit responds to the event and updates the organization accordingly.
This allows the system to perform automatic, unsupervised grouping locally without depending on external AI models or cloud services.
Challenges we ran into
The biggest challenge was making the organization happen in real time without making the application resource-heavy.
We also had to figure out how to produce meaningful groups using only file names. File names can be short, inconsistent, or ambiguous, which makes grouping difficult in some cases.
Another challenge was handling file additions and removals while keeping the existing organization consistent. The system needed to react to changes without repeatedly processing everything from scratch.
Accomplishments that we're proud of
We are proud that a simple personal utility evolved into a working product capable of organizing files continuously in real time.
We built the system without relying on LLMs, external APIs, or an internet connection. Using TF-IDF and HDBSCAN, File Fit can discover relationships between files locally while remaining lightweight.
We are also proud of the way it groups files by their names rather than their extensions. Related files can stay together even when they have completely different formats.
Most importantly, we built something that solved a real problem we personally experienced.
What we learned
File Fit taught us that solving a practical problem does not always require complex AI models. Techniques like TF-IDF and HDBSCAN can be surprisingly effective when applied to the right problem.
We also learned about the challenges of building real-time systems, especially how to respond efficiently to filesystem changes without unnecessary processing.
Most importantly, we learned that some of the best project ideas can come from problems we experience ourselves. A small personal tool can become a much larger product when the problem is common and the solution is useful.
What's next for File Fit
The next major feature we plan to add is a rollback system.
Currently, File Fit reorganizes files in real time, but it does not store enough information about the previous file structure to easily undo those changes. Once files have been grouped, restoring the original organization manually can be difficult.
The rollback system will keep track of File Fit's organizational changes so users can undo changes and restore their previous file structure when needed.
This will make File Fit safer and give users more confidence when allowing it to automatically organize their files.
Log in or sign up for Devpost to join the conversation.