Satyam
The Core Problem: AI safety tools are fundamentally useless if they require the cloud to save a life.
Core trick
Runs a local Gemma model to score hazards offline and queues SOS messages in local storage until cell signal returns, then sends them.
Description
The Core Problem: AI safety tools are fundamentally useless if they require the cloud to save a life. Traditional apps break the second a trekker loses cell service. GPS maps can show you where you are, but they cannot reason about unfolding dangers.
The Solution: Pahad Sathi shifts intelligence to the edge. By running gemma4:e2b locally, the trekker has a 2-Billion parameter reasoning engine in their pocket that constantly analyzes terrain, weather, and risks offline. If an emergency happens in a dead zone, it doesn't crash—it queues the SOS locally, waiting to fire the exact millisecond the phone hits a cell tower.
It is the perfect marriage of local reliability (Gemma 4) and cloud superpowers (Gemini Vision/TTS) when signal returns!
README excerpt
# Pahad Sathi (Mountain Companion)
Pahad Sathi is an intelligent, offline-first safety companion application designed for trekkers and mountaineers. Built for the **"Best Use of Gemma 4 - Local-First Agents on Gemma"** track, it provides real-time hazard analysis, dynamic safety instructions, and a resilient emergency response system without requiring an internet connection.
## 🌟 Key Capabilities
### 1. Offline-First Gemma 4 Sense-Decide-Act-Check Loop
Mountains rarely have good cell service. Pahad Sathi's core safety reasoning loop runs **100% locally on-device** using the `gemma4:e2b` model via Ollama.
- The local engine continuously evaluates terrain, elevation, seasonal weather patterns, user experience level, and session history to classify the safety status into distinct tiers:
- ✅ **On Track**: Everything is proceeding safely.
- ⚠️ **Caution Required**: The user needs to alter their approach.
- 🚨 **Emergency Risk**: An immediate danger is detected. Automatically triggers the SOS protocol.
- 👤 **Needs Human Input**: The system encounters an ambiguous situation and defers to the user to assess and proceed (e.g., when position data doesn't match any known waypoint, the agent asks the user to confirm location rather than guessing).
- **Local Calibration**: The local engine maintains session state and calibrates its own caution level based on user feedback — if you mark an alert as unnecessary, similar terrain later in the session is reassessed according