This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
SlothOS (Smart Local Operations & Task Helper Operating System) is a 100% offline, privacy-first AI life-admin assistant that reads sensitive personal documents like utility bills and payment receipts, turning them into actionable tasks without ever sending a single byte to the cloud.
Key features include:
- Evidence-Grounded Extraction: Every extracted deadline or amount must be backed by a verbatim quote from the source document to eliminate hallucinations.
- Cross-Document Intelligence: Automatically links documents via shared Account IDs and flags conflicting records (e.g., unpaid bills vs. paid receipts) as "Needs Review" instead of silently overwriting data.
- Ask SlothOS (Offline Conversational AI): A local chat interface that answers questions about your paperwork using structured context.
- Knowledge Graph: Integrates the Graphify CLI to turn files into a visual map, feeding structured relationship data directly into the chat interface.
Demo
Code
🦥 SlothOS
Your life admin, handled slowly, safely, and 100% offline.
A privacy-first AI assistant that reads your bills and receipts, turns them into actionable tasks, and never sends a single byte to the cloud.
Features · Architecture · Hackathon Tracks · Getting Started
🌿 What is SlothOS?
Most "AI assistants" want your documents in their cloud. SlothOS flips that: it brings a small, open-weight model to your machine, reads your electricity bills, rent receipts and insurance letters, and turns them into a clean list of deadlines and amounts.
And it's deliberately slow and steady. When two documents disagree, SlothOS doesn't quietly pick a winner. It raises its hand, shows you the evidence, and lets you decide.
"A sloth never rushes, never drops the branch, and always checks twice."
🔒 Why SlothOS? (The Privacy Problem)
Life-admin documents are among the most sensitive files you own: account numbers, addresses, balances…
How I Built It
SlothOS is built entirely around local inference and open-weight AI to ensure 100% data privacy.
-
Local Inference & Model: We utilize Google’s open-weight **Gemma 2B (
gemma2:2b) running entirely offline on local hardware via Ollama, eliminating any dependency on cloud APIs, external endpoints, or paid keys. - Agentic Framework & Stack: The application pairs a FastAPI, Python backend with a React, Vite, and Tailwind CSS frontend, backed by a local SQLite database.
- Document & Graph Harnesses: We integrated custom extraction and cross-document validation pipelines to guarantee evidence-grounded parsing (verbatim quote matching), alongside the open-source Graphify CLI to automatically map file relationships into an interactive knowledge graph for offline conversational search ("Ask SlothOS").
Why Does Open Innovation Matter?
Managing personal life-admin paperwork—such as utility bills, lease agreements, and bank statements—requires absolute privacy and trust. Closed, proprietary AI APIs force users to send sensitive financial data to third-party cloud servers, creating significant security risks. Open innovation and open-weight models like Gemma 2B make it possible to run powerful agentic document workflows locally on user hardware. This guarantees complete data sovereignty, ensures zero cloud telemetry, and empowers developers to audit, customize, and extend secure AI tools without vendor lock-in.
The Lazy Payoff
SlothOS saves hours of tedious manual paperwork tracking every single month. Instead of manually cross-referencing utility bills against bank statements and payment receipts to check if a due date was met or if an amount was already paid, users simply drag and drop the files. This week, we tested it with real utility bills and payment receipts (such as tracking Account TEST-1042), and SlothOS instantly automated the heavy lifting—extracting deadlines with verbatim quote backing and automatically flagging financial contradictions as "Needs Review" without a single second spent organising spreadsheets or risking cloud data exposure.
Prize Categories
Best Use of Gemma -SlothOS showcases the true potential of Gemma 2B as a privacy-first edge AI engine:
- 100% Offline Power: Runs entirely local via Ollama with zero cloud dependencies, handling sensitive personal documents securely.
- Evidence-Grounded Extraction: Eliminates hallucinations by forcing Gemma to back every extracted deadline or amount with a verbatim quote from the source file.
- Cross-Document Reasoning: Uses Gemma to connect multiple files (like matching utility bills to payment receipts via Account IDs) to catch financial contradictions automatically.
Thanks for reading my submission all the way to the end! Appreciate your time and support!!
Top comments (0)