Updated the AI explanation system prompt with stricter INS rules

feat: water hardness mode with sodium citrate recommendation via Codex

fix parsing of prompts

fix parsing of prompts

Revert "feat: water hardness mode with sodium citrate recommendation via Codex"

feat: water hardness mode with sodium citrate recommendation via Codex

feat: water hardness mode with sodium citrate recommendation via Codex

fix: parser handles em-dash in optimise input

fix: resolve conflicting improvement rules in agentic optimiser

feat: agentic recipe optimiser - multi-step improvement loop via Codex

fix: oil aliases parser, superfat display, stop follow-up questions

fix: shea butter parser, superfat display, stop hallucination tokens

feat: superfat slider - live lye recalculation

fix: checkbox pointer-events restored

fix: UI polish - cream header, full recipe cards, wider layout

fix: UI polish - cream header, full recipe cards, wider layout

fix: UI polish - cream header, full recipe cards, wider layout

feat: custom UI redesign via Codex - Off-Brand badge

feat: custom UI redesign via Codex - Off-Brand badge

fix: upgrade to Qwen2.5-3B, increase max tokens

fix: clean app with Nebius provider for Qwen2.5-0.5B

fix: clean app with Nebius provider for Qwen2.5-0.5B

fix: use SmolLM2-1.7B via HF Inference API - Tiny Titan eligible

fix: use SmolLM2-1.7B via HF Inference API - Tiny Titan eligible

fix: use SmolLM2-1.7B via HF Inference API - Tiny Titan eligible

fix: remove ollama import

fix: use SmolLM2-1.7B via HF Inference API - Tiny Titan eligible

Initial commit - fixed README YAML configuration