AI Governance for Fitness

The language that governs AI-generated wellness plans

▸ Constrain LLM output so plans are safe, correct, and hallucination-free
▸ Cover workouts, cardio, nutrition, meditation, recovery, and habits in one validated structure
▸ Personalize plans automatically with a built-in rules engine

How WPL works

Plans follow a clear hierarchy: phases contain weeks, weeks contain days, days contain blocks, and blocks contain activities.

Plan Structure

📋 Plan "4-Week Weight Loss"

🔄 Phase "Foundation"

📅 Week 1

💪 Day 1: Upper Body 45 min

🟢 Warmup Block

🔵 Main Block 3 circuits

• Push-ups (3×10)

• Dumbbell Rows (3×12)

🟣 Cooldown Block

🧘 Day 2: Recovery 30 min

• Yoga Flow

• Meditation

plan.wpl.json

{
  "plan": {
    "name": "4-Week Weight Loss",
    "phases": [{
      "name": "Foundation",
      "weeks": [{
        "days": [{
          "name": "Upper Body",
          "blocks": [{
            "type": "main",
            "structure": "circuit",
            "rounds": 3,
            "activities": [...]
          }]
        }]
      }]
    }]
  }
}

Why WPL?

AI governance meets wellness. Structure that keeps LLMs honest.

Holistic Wellness

One plan, six activity types: workouts, nutrition, meditation, recovery, habits, and cardio. WPL treats wellness as a whole, not just exercise.

Smart Personalization

Built-in rules engine adapts plans to each person. Injuries, equipment, fitness level, and preferences drive automatic modifications.

Progress Built In

Checkpoints, milestones, streaks, and points are part of the plan definition. Track adherence and outcomes from day one.

Platform Agnostic

JSON-based format parseable by any language, renderable on any platform. Build once, deploy everywhere.

AI Governance

WPL constrains LLM output so AI-generated plans are safe, correct, and free from hallucination. A dedicated DSL for generation, compiled to validated JSON.

Template System

Reusable, parameterized templates let trainers create once and customize infinitely. Plans become composable building blocks.

Built for AI generation

AI generates plans in the concise WPL-AI DSL. The system compiles it to WPL JSON for storage, validation, and rendering.

WPL-AI DSL

AI generates this — concise, reliable, easy to validate

upper-body.wpl

PLAN "Upper Body Strength"
TYPE workout
DIFFICULTY intermediate
DURATION 4 weeks

GOALS
  GOAL primary strength:
    target weight 5 kg relative

PHASES
  PHASE "Foundation" (4 weeks):
    WEEK 1:
      DAY Monday training 45m:
        main straight_sets:
          bench_press 4x8 rpe 7 rest 90s tempo 3-1-2-0
          overhead_press 3x10 rpe 6 rest 60s
          push_up 3x12..15 rest 45s

WPL JSON

Compiled output — stored in DB, rendered by apps

bench-press.wpl.json

{
  "id": "exercise_1",
  "type": "exercise",
  "exercise_ref": "bench_press",
  "name": "Bench Press",
  "prescription": {
    "type": "sets_reps",
    "sets": 4,
    "reps": { "target": 8 },
    "tempo": "3-1-2-0",
    "rest": {
      "value": 90,
      "unit": "seconds"
    },
    "intensity": {
      "type": "rpe",
      "value": 7
    }
  },
  "progression": {
    "type": "linear",
    "increment": {
      "value": 2.5,
      "unit": "kg",
      "per": "week"
    }
  }
}

Personalization in action

Client reports a knee injury. WPL automatically replaces unsafe exercises across the entire plan.

Before

Original plan

// Original plan activity
{
  "id": "ex_1",
  "exercise_ref": "squat",
  "name": "Barbell Squats",
  "prescription": {
    "sets": 4,
    "reps": { "target": 10 }
  }
}

Rule

Knee injury detected

// Personalization rule
{
  "condition": {
    "field": "client_injuries",
    "op": "contains",
    "value": "knee"
  },
  "actions": [{
    "type": "replace_exercise",
    "from": "squat",
    "to": "wall_sit"
  }]
}

After

Safe alternative

// Personalized result
{
  "id": "ex_1",
  "exercise_ref": "wall_sit",
  "name": "Wall Sit",
  "prescription": {
    "sets": 4,
    "reps": { "target": 10 }
  }
}

Squats → Wall Sits. Lunges → Step-ups. Jump Squats → removed. All automatic, all safe.

Who it's for

WPL bridges the gap between coaching expertise, technical implementation, and platform needs

💪

Trainers & Coaches

Create structured, reusable plans with a visual constructor. Scale your coaching with templates and automatic personalization. Spend time coaching, not rewriting plans.

✓ Visual plan builder
✓ Reusable templates
✓ Auto-personalization

⚡

Developers

Parse, validate, and render plans in any language. JSON-based format with a clear schema. Build apps, integrations, and tools on a standard foundation.

✓ JSON schema
✓ Any-language parsing
✓ API-ready

🏗️

Platforms

Govern AI-generated plans across your product. Every LLM output is validated against the WPL schema before reaching users. Track outcomes with consistent, structured data.

✓ AI governance
✓ Schema validation
✓ Outcome tracking

Created & maintained by

Gymbile

Building the AI governance layer for fitness and wellness.

Visit gymbile.com