What I Built
Manual data entry is the ultimate enemy of a slothful lifestyle. Opening a budgeting app to log a lunch expense, switching to a to-do list to add a task, and then opening a notes app to journal the day requires too much tapping, categorizing, and screen time.
I built SlothSync: a fully offline, voice-driven desktop application that does all the sorting work for you. You press a single button, mumble your day's unstructured thoughts into the microphone, and a local AI automatically parses and categorizes your rambling into a clean dashboard of Expenses, Tasks, and Journal entries. It saves the user from the friction of context-switching between productivity apps.
How I Built It
SlothSync relies entirely on open-weight models and local inference to function:
The AI/Models: I used faster-whisper for instant, offline Speech-to-Text transcription. The transcript is then routed to a local instance of Gemma 2B running via Ollama. A strict system prompt forces Gemma to act as a data extractor, outputting a clean JSON payload.
The Frontend: React wrapped in Tauri v2 for a lightweight, cross-platform native desktop experience.
The Backend: FastAPI (Python) handles the audio file ingestion, local model routing, and JSON structuring.
Why Does Open Innovation Matter?
Building an AI life-logger requires handling highly intimate daily data—personal finances, private tasks, and unstructured thoughts.
Absolute Privacy: Sending a raw, unfiltered brain-dump to an external, closed-source API is a massive privacy risk. By utilizing open-weight models like Gemma and Whisper running purely on local hardware, my personal data never touches a server I don't control.
The Offline Sloth: Automation shouldn't break just because you lose Wi-Fi. Because the open models run entirely locally, I can log my thoughts while on a flight or deep in a library with zero internet connection.
Zero Recurring Costs: Commercial voice-to-text and LLM APIs charge per token and per minute of audio. Open-source infrastructure allows you to build this automation once and run it endlessly for free.
The Lazy Payoff
Instead of spending 5 minutes every evening opening a finance tracker, a task manager, and a diary, my daily logging takes exactly 5 seconds. I hold a button, say exactly what is on my mind in one breath, and close the app. The AI handles the entire administrative burden of sorting my life out.
Prize Categories
Open-Source AI Challenge Week 3: Slow and Steady
CODE
Top comments (0)