# Can I Run It? - Local AI Hardware API

This project provides deterministic hardware compatibility data for local open-source LLM usage.

## Data Endpoints

- Raw JSON database: `/data/models.json`
- Format: JSON object containing metadata, quantization options, hardware presets, model parameters, VRAM estimates, Ollama tags, capability notes, source links, confidence fields, and fallback recommendations.
- Current V1 dataset: 15 curated model entries and 8 hardware presets.

## Dynamic Insights

Model memory estimates use quantization bits and a default 1.2x KV-cache multiplier. Compatibility status changes with GPU VRAM, system RAM, and quantization level.

## Caveat

The data is designed for planning and comparison. Real runtime memory and speed depend on backend, context length, KV cache, driver stack, quantization file, and offloading settings.

## Policy

Hardware and cloud recommendations must be clearly labeled. Compatibility status must remain data-based and must not be changed by affiliate or sponsor relationships.