[FEAT]: Hybrid AI Orchestration & Pydantic Validation Layer #408
Description
name: 🚀 Feature Request
about: Implement Hybrid AI Orchestration and Pydantic Data Validation Layer for resilient reporting.
title: " [FEAT]: Hybrid AI Orchestration & Pydantic Validation Layer"
labels: enhancement
assignees: ''
📝 Description
This feature introduces a Hybrid AI Orchestration system that enables FireForm to operate in resource constrained environments by dynamically switching between local SLMs and Cloud APIs. It also adds a Pydantic based validation layer to ensure all AI extracted data is hallucination free and schema compliant.
💡 Rationale
First responders often work in disconnected or low-bandwidth environments. Current AI prototypes often fail without a stable internet connection or high VRAM. This feature ensures resiliency by making the software hardware-aware, allowing it to run locally on mid-range field laptops (8GB RAM) while maintaining 100% data integrity for official reports.
🛠️ Proposed Solution
A brief sketch of how we might implement this.
- Resource Monitoring: Integrate psutil to detect real-time RAM/VRAM availability.
- Dynamic Routing: Logic to trigger Ollama (4-bit quantized models) locally if resources allow, with a seamless fallback to Cloud APIs (OpenAI/Groq) if offline or underpowered.
- Validation: Implementation of Pydantic Models to strictly enforce data schemas on AI outputs before they reach the UI.
- HITL Component: A React-based "Human-in-the-Loop" verification step for final data auditing.
✅ Acceptance Criteria
- The system correctly detects hardware limits and toggles the inference mode.
- The project remains functional within a standard Docker container.
- README updated with "System Requirements" for local execution.
📌 Additional Context
This is a core component of my 350 hour GSoC 2026 proposal. It builds on my background in AI & Data Science to ensure the tool is field ready and remains fully operational in low resource, disconnected environments where data integrity is mission critical.