CalCore: AI-Powered Custom Instruction Set for Fat Loss

Welcome to CalCore, an experimental platform that reimagines how we interact with fat-loss tracking systems by combining custom ISA (Instruction Set Architecture) design with AI-driven input interpretation.

This is not just a calculator—it's the foundation of an intelligent, extensible, user-friendly metabolic tracking engine designed for daily use.

---

💡 Project Vision

Most fat-loss apps require painful manual logging of food and exercise. CalCore's vision is to eliminate manual input entirely by enabling users to simply say:

“I weigh 51kg today. I had 3 buns for breakfast, a bowl of noodles for dinner, and danced for 3 hours.”

1. 🔍 Automatically interpret the input using an AI language model
2. ⚙️ Translate it into CalCore DSL commands
3. 🧮 Execute the commands via interpreter
4. 📊 Generate a personalized metabolic report, including:
   • Daily calorie intake
   • Calorie consumption from activities
   • Net deficit or surplus
   • Visual summary (charts, logs, trends)

Eventually, every user will have access to:

• 📆 A per-day auto-generated metabolic summary
• 📈 Long-term weight tracking and curve plotting
• 🧠 AI-driven natural language interaction — no manual tracking needed

---

🔧 Features (In Progress)

• DSL-based calorie and movement command parser
• Support for per-user profile (height, weight, age, body fat %)
• BMR/RMR calculation using Harris-Benedict equation
• Calorie tracking (intake + consumption)
• Weight history logging with duplicate filtering
• Weight trend visualization with matplotlib
• Natural language to DSL parser using LLM
• Auto-generated daily metabolic report
• Web or mobile front-end

---

🚀 Sample Usage

Example CalCore DSL (.cal) input file:

LOG_WEIGHT 2025-06-25 51.7
EAT 300 carb
EAT 500 protein
MOVE DANCE 120
QUERY BMR
QUERY DEFICIT

Run the interpreter:

python interpreter/interpreter.py examples/sample_day.cal

📈 Plotting Weight Trends Generate weight curve graph from JSON log:

python plot/plot_weight.py

Saves to output/weight_curve.png

🧠 Future: Natural Language Interface The next step is to allow:

“I weigh 52kg today. Ate a chicken salad. Biked 1 hour.” to be automatically transformed into:

LOG_WEIGHT 2025-06-25 52.0
EAT 350 protein
MOVE BIKE 60

This will be achieved through LLM-powered input parsing and DSL code generation.

👩‍💻 Author Erin Xu University of Michigan – ECE Major 🔗 GitHub: ErinXU2004