What if your computer could take care of the boring digital tasks you keep postponing?
Students accumulate PDFs. Freelancers collect invoices. Developers leave their Downloads folder full of files named final_v2_revised_FINAL.pdf. And somehow, organizing everything becomes another task on an already endless to-do list.
That's why I built SlothDesk AI — a local-first AI productivity assistant designed to help people spend less time managing digital clutter and more time doing what actually matters.
What I Built
SlothDesk AI is a personal productivity agent powered by Google Gemma 3 4B, running locally through Ollama.
It helps users automate repetitive digital tasks, including:
- 📁 Organize files: Categorize documents and suggest meaningful filenames.
- 📄 Summarize documents: Extract key information from PDFs and text files.
- 📅 Extract deadlines: Turn dates and action items found in documents into useful tasks.
- ✅ Manage tasks: Keep track of priorities, upcoming deadlines, and completed work.
- ☀️ Generate morning briefings: Summarize your priorities and help you decide what deserves your attention today.
- ↩️ Review and undo file operations: Preview proposed changes and approve them before files are moved or renamed.
The idea is simple: let the AI handle the tedious parts while the user stays in control.
Demo
VIDEO: https://youtu.be/F4XZCnb-ktQ
SlothDesk AI currently runs locally. The repository includes instructions for setting up the application and running Gemma through Ollama.
To see the application in action, demo video accompanying this post.
Code
Check out the complete source code:
The project includes the backend, frontend, automated tests, documentation, and an MIT license.
Contributions, suggestions, and feedback are welcome!
How I Built It
The application combines a modern web interface with a local AI backend.
Technology stack
- AI model: Google Gemma 3 4B Instruct
- Inference: Ollama
- Backend: Python and FastAPI
- Frontend: React, TypeScript, Vite, and Tailwind CSS
- Database: SQLite
- Document processing: PyMuPDF
- Testing: pytest, Vitest, and React Testing Library
Gemma is used for tasks such as document summarization, information extraction, and file-organization proposals. A modular AI provider layer keeps the inference implementation separate from the rest of the application.
The application also uses deterministic fallback logic for selected basic operations when the local model is unavailable. These fallbacks are explicitly distinguished from actual AI-generated results.
Safety was an important part of the implementation. The application validates proposed file operations, prevents unauthorized filesystem access, handles filename collisions, and requires user approval before executing changes.
The reported automated test results are 21 passing backend tests and 5 passing frontend tests.
Why Does Open Innovation Matter?
For SlothDesk AI, open innovation isn't just about choosing a different model. It changes how the product can be used and customized.
1. Your documents can stay on your machine
Documents may contain invoices, academic information, contracts, and other private details. With the local Gemma and Ollama setup, supported AI processing can happen on the user's own computer rather than requiring document contents to be sent to a hosted AI API.
That gives users greater control over their information.
2. No mandatory paid AI API for the core local workflow
The application can use a locally installed open-weight model without paying per request to a commercial inference API.
Of course, local inference still requires hardware and electricity, and downloading or running a model has its own resource costs. But users can avoid usage-based API charges for local inference.
3. The model is replaceable
The AI provider is separated from the application services. This makes it easier to experiment with other compatible models and compare their output, performance, and resource requirements.
The application doesn't have to be permanently tied to a single hosted provider.
4. Users retain more control
Local inference offers more control over where data is processed, which model is used, and how the application behaves.
A closed API can be convenient, but a local-first approach is particularly attractive when privacy, predictable usage costs, and customization matter.
For this project, open-weight AI made it possible to build a practical productivity assistant around local inference rather than making a remote AI service a mandatory dependency.
The Lazy Payoff
The goal is to reduce the small, repetitive tasks that quietly consume a person's day.
Instead of manually opening documents, figuring out their contents, inventing filenames, and copying deadlines into a task manager, a user can let SlothDesk prepare suggestions and review them in one place.
I haven't established a reliable benchmark for the exact number of minutes saved, so I won't invent one. The next step is to compare the time required to complete these tasks manually against the time required with SlothDesk AI.
For now, the payoff is a simpler workflow: less repetitive administration, fewer forgotten details, and more time for the work that matters.
Prize Categories
Gemma — Best Use of Gemma
SlothDesk AI uses Google's open-weight Gemma 3 4B Instruct model through local Ollama inference for its AI-powered productivity workflows.
This is the primary partner category I'm targeting.
Team Credits
Built by @rudraksh_karmakar_28.
Team credit: @harry_92e400f33306e4ba7bb.
Final Thoughts
I built SlothDesk AI around a straightforward idea: AI should not only answer questions. It should also help people get rid of the repetitive work they don't want to do in the first place.
Open-weight models make that vision more accessible by giving developers the freedom to experiment with local inference, build privacy-conscious workflows, and customize how their applications work.
Sometimes the most useful AI isn't the one that does something extraordinary.
It's the one that quietly takes a few annoying tasks off your plate.
SlothDesk AI: Do less busywork. Get more of your day back. 🦥
If you have ideas for making SlothDesk AI more useful, I'd love to hear your feedback!
Repository: https://github.com/rudrakshk25060-csds/slothdesk-ai
Top comments (0)