This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady.
Meet SlowMate AI, a local-first productivity assistant designed for people who feel overwhelmed by their to-do lists and struggle to decide where to start.
Instead of asking users to organize everything themselves, SlowMate lets them dump their thoughts, unfinished tasks, or daily responsibilities into one place. They can select their available energy level and time, and the AI turns that input into a small, realistic action plan.
The idea is simple: productivity shouldn't always mean doing more. Sometimes, it means making the next step easier.
🦥 Try SlowMate AI: https://namansrivastava1.github.io/SlowMate-AI/ The website provides the interface for turning a messy task list into a calmer, more manageable plan. AI generation requires Ollama and the configured local model to be running on the user's computer.
💻 GitHub Repository: https://github.com/namansrivastava1/SlowMate-AI The project is built with HTML, CSS, and JavaScript, with no frontend framework required.
Explore the repository to see how the interface, task-planning workflow, and local AI connection fit together.
I built SlowMate AI using three core web technologies:
For the AI layer, SlowMate uses Ollama to run Qwen2.5 3B, an open-weight language model, locally. Rather than sending task descriptions to a hosted AI API, the application connects to the Ollama service running on the user's own computer. The model generates a practical plan based on the task description, energy level, and available time.
This architecture makes the project more than a static productivity dashboard: the open-weight model is responsible for the central task-planning experience.
The frontend is deployed on GitHub Pages, while AI inference runs locally. Users need to install Ollama, download the model, and configure the local browser connection before they can use AI generation.
For SlowMate, open innovation is about control, privacy, and flexibility. Task descriptions can contain personal information about studies, work, or everyday responsibilities. Running inference locally means those descriptions do not need to be sent to a third-party AI provider.
Recent plans are also stored in the user's browser rather than in a remote application database.
Using an open-weight model gives developers more flexibility to experiment with model versions, prompts, and local configurations instead of depending entirely on one closed AI provider. Once the model has been downloaded, local inference does not require a per-request hosted AI API fee. However, it still uses the computer's processing power and electricity, and users need hardware capable of running the model.
After the model and required software have been installed, local inference can work without an internet connection. The hosted website still needs to be loaded or available locally, and the initial setup requires down the necessary files.
For this project, open innovation makes a private, customizable productivity assistant possible without making a hosted AI API the center of the architecture. SlowMate is built around a small but important question:
What if getting started took less effort than putting the task off?
Instead of manually sorting a long list, deciding priorities, and planning every step, users can describe the situation once and let the model suggest a manageable starting point.
The goal isn't to automate every part of life. It's to remove enough friction that taking the first step feels possible.
I haven't assigned a numerical time-saving claim because the actual benefit depends on the task and the quality of the model's output. Measuring the time required to organize a task list manually versus using SlowMate would be a useful next experiment.
Some improvements I'd like to explore include:
SlowMate is a small step toward a different kind of productivity: less pressure, less friction, and more realistic progress.
Slow is still a speed. 🦥