The Story of Elderly Routine Check

What Inspired Me

The inspiration for this project comes from a massive global crisis in eldercare. Millions of independent seniors live alone, and thousands reside in understaffed assisted living facilities. Their global family networks live with constant anxiety regarding sudden emergencies, missed meals, or late check-ins.

Current tech solutions present a terrible choice: either force seniors to wear uncomfortable tracking buttons they forget, or install invasive cloud-based cameras that stream private video onto the public internet. I wanted to build a third option: an acoustic and visual safety shield that runs 100% locally on a private home or facility server, converting raw logs into simple text milestone alerts to respect an elder's ultimate dignity and privacy.

How I Built It

I strictly utilized the permitted Uliya framework capabilities to orchestrate an on-demand, non-invasive validation pipeline:

• Natural Setup Loop: I integrated ug-knowledge -> audio-transcription to convert voice recordings of daily routines into structured schedule checklists.
• On-Demand Auditing: The system reads the baseline using file-agent (cat), calls security-monitor-agent (event-timeline) to scan for historical room-presence activity events, and resolves faces using the read-only identity-resolve database.
• Privacy Verification: The skill isolates data, extracting a single freeze-frame image via event-evidence (snapshot) to attach as safe visual evidence without raw video streaming.
• Inventory Calculations: I designed a diet inventory tracker leveraging audio-transcription and a strict "Waiting for Gateway Approval" status loop via file-agent (echo) to safely edit local asset logs without premature writes.

Challenges I Faced

As a first-time hacker, my biggest hurdle was learning the strict distinction between a theoretical, magical AI concept and a realistic, engineered solution. Initially, I planned a passive system with continuous background listening for alarms and automated emergency phone dialing.

However, cross-referencing my ideas against the strict parameters of the capability catalog revealed critical platform boundaries: Uliya Skills cannot run continuously in the background without user invocation, and there is no outbound SMS or phone capability. Navigating these constraints forced me to pivot into a realistic "Absence of Safety" indicator approach that provides ready-to-send text templates to the user instead of over-claiming real-time device capabilities.

What I Learned

I learned that technical precision beats flashy, impossible promises. I discovered how to handle asynchronous states like waiting_approval and how to process metadata without accessing raw streaming layers.

Furthermore, I learned the importance of structural constraints, realizing that the mathematical calculation of a duration anomaly between bathroom door entry and exit events can be written as:

$$\Delta t = t_{\text{exit}} - t_{\text{entry}}$$

By evaluating if $\Delta t > T_{\text{threshold}}$ entirely on-demand, I learned that I can preserve safety bounds mathematically without needing a live background timer running on the server.

Built With

• javascript
• markdown
• os
• python
• uliya