BADGES.md

🎖️ Parry — badge evidence, one click each

(Duplicated from README.md so it can't be missed: this file, the README table, the in-game "✦ artifacts" links, and the about Gradio API all carry the same links.)

Badge	Evidence
🔌 Off the Grid	All inference runs in YOUR browser: Tactician + imagination on WebGPU (WebLLM), the tuned Analyst on llama.cpp/WASM. GET /healthz → "inference": "in-browser" — pull your Wi-Fi out mid-match.
🎯 Well-Tuned	The published fine-tune runs inside this app as the Analyst. parry-tactician-1.5b-merged — model card has the recipe, the six-generation eval arc (including the failed gates), the CFG γ-sweep table, and evals/ with all 18 raw report JSONs. Adapter: -lora.
🦙 Llama Champion	parry-tactician-1.5b-gguf (Q4_K_M) runs through llama.cpp's WASM runtime (wllama) in-browser. Watch for "✦ tuned analyst (llama.cpp)" in the HUD meta line.
🎨 Off-Brand	The custom pixel-dojo canvas, served by gr.Server — Gradio's engine at /, the trace_digest endpoint through Gradio's queue.
📡 Sharing is Caring	parry-traces — decision-by-decision agent traces with full provenance, replayable against the deterministic engine. Export your own: B → Export trace.
📓 Field Notes	The full build report — six model generations, three honestly-failed pre-registered gates, the engine-physics bug a playtest found, classifier-free guidance for game policies, and the pose-mode saga.