Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jaimegayo/KERNDOCUMENTATION/llms.txt

Use this file to discover all available pages before exploring further.

When you complete the onboarding quiz, KERN does not randomly assign exercises — it runs a deterministic selection algorithm that maps your chosen routine type to specific muscle groups and movement patterns, then picks exercises that match both dimensions from the gym exercise catalog.

​
Routine types

KERN supports three routine structures. Each is defined in the ROUTINE_CONFIGS dictionary inside app.py and maps to a list of {muscle, pattern} targets. 
ROUTINE_CONFIGS = {
    "Full Body": [
        {"muscle": "Pecho",           "pattern": "Empuje horizontal"},
        {"muscle": "Espalda",         "pattern": "Tracción vertical"},
        {"muscle": "Cuádriceps",      "pattern": "Sentadilla"},
        {"muscle": "Hombros",         "pattern": "Empuje vertical"},
        {"muscle": "Isquiotibiales",  "pattern": "Bisagra de cadera"}
    ],
    "PPL (Push Pull Leg)": [
        # Push
        {"muscle": "Pecho",           "pattern": "Empuje horizontal"},
        {"muscle": "Pecho superior",  "pattern": "Empuje horizontal"},
        {"muscle": "Hombros",         "pattern": "Empuje vertical"},
        # Pull
        {"muscle": "Espalda",         "pattern": "Tracción vertical"},
        {"muscle": "Espalda media",   "pattern": "Tracción horizontal"},
        {"muscle": "Bíceps",          "pattern": "Aislamiento"},
        # Leg
        {"muscle": "Cuádriceps",      "pattern": "Empuje de piernas"},
        {"muscle": "Isquiotibiales",  "pattern": "Bisagra de cadera"}
    ],
    "Torso-Pierna": [
        {"muscle": "Pecho",           "pattern": "Empuje horizontal"},
        {"muscle": "Pecho superior",  "pattern": "Empuje horizontal"},
        {"muscle": "Espalda",         "pattern": "Tracción vertical"},
        {"muscle": "Espalda media",   "pattern": "Tracción horizontal"},
        {"muscle": "Cuádriceps",      "pattern": "Sentadilla"},
        {"muscle": "Isquiotibiales",  "pattern": "Bisagra de cadera"}
    ]
}
Targets one exercise from each of five fundamental patterns in a single session: horizontal push (chest), vertical pull (back), squat (quads), vertical push (shoulders), and hip hinge (hamstrings). Designed for training 3 days per week.
Splits the week into push, pull, and leg days. The push slot covers chest (flat and incline) and shoulders. The pull slot covers vertical and horizontal back plus biceps isolation. The leg slot covers quad-dominant and hamstring-dominant patterns. Designed for 6-day training frequency.
An upper/lower split that alternates torso (chest, upper chest, back, mid-back) and leg (quad squat, hamstring hinge) sessions. Suitable for 4-day training frequency.

​
How exercise selection works

The generate_routine_exercises function in app.py iterates over each {muscle, pattern} target in the chosen config and selects one exercise per slot.
1

Load the exercise catalog

At startup, KERN reads exercises_gym_id.json into memory. Every entry has a primary_muscle and a movement_pattern field.
2

Match by muscle and pattern

For each config slot, the algorithm looks for exercises where both primary_muscle and movement_pattern match the target (case-insensitive).
candidates = [
    ex for ex in all_exercises
    if ex["primary_muscle"].lower() == target_muscle
    and ex["movement_pattern"].lower() == target_pattern
]
3

Fallback to muscle-only

If no exercise matches both criteria, the algorithm widens the search to muscle alone. This ensures every slot is always filled.
if not candidates:
    candidates = [
        ex for ex in all_exercises
        if ex["primary_muscle"].lower() == target_muscle
    ]
4

Random selection

One exercise is chosen at random from the candidate list using random.choice(). This introduces variety across users with the same routine type.
5

Default series assigned

Each selected exercise is initialized with three sets at 0 kg and 10 reps. These are the starting values the user will fill in during their first session.
{
    "nombre": chosen["name"],
    "series": [
        {"numSerie": 1, "kilos": 0, "reps": 10},
        {"numSerie": 2, "kilos": 0, "reps": 10},
        {"numSerie": 3, "kilos": 0, "reps": 10}
    ]
}

​
Triggering generation

Routine generation is triggered automatically when the user completes the quiz. The POST /users/complete-quiz endpoint receives the selected routine name, validates it against ROUTINE_CONFIGS, runs generate_routine_exercises, and persists the result as a new Routine with linked RoutineExercise rows. 
POST /users/complete-quiz
Authorization: Bearer <token>
Content-Type: application/json

{
  "assigned_routine": "Full Body"
}
If the assigned_routine value does not match any key in ROUTINE_CONFIGS, no exercises are generated. Accepted values are "Full Body", "PPL (Push Pull Leg)", and "Torso-Pierna" — exact string match required.

Build docs developers (and LLMs) love

Get started for freeTalk to us