Juggernaut Agent

Juggernaut Router is a general-purpose AI agent built for AMD Developer Hackathon Track 1. It handles the required eight capability categories: factual knowledge, mathematical reasoning, sentiment classification, text summarisation, named entity recognition, code debugging, logical reasoning, and code generation. The system reads tasks from /input/tasks.json and writes evaluator-compatible answers to /output/results.json. Instead of sending every prompt to the same model, it classifies each task and chooses the most reliable route available: deterministic solvers for safe patterns, optional bundled local GGUF inference for local reasoning, and Fireworks models for harder or riskier tasks. The project focuses on accuracy first, then efficiency. It includes answer normalization, validation, route telemetry, Docker submission checks, public pullability checks, and synthetic evaluation fixtures that mirror the Track 1 task categories. The goal is a practical routing agent that preserves correctness while reducing unnecessary remote token usage.