Aixle Enhanced Script Package: Real-World Testing Results & Community Improvements (March 2026)

[biscuitbloke-lab]

This document presents a comprehensive enhancement package for Aixle, the open-source AI cycling coach that generates personalized .zwo workouts based on Intervals.icu data.

🎯 What's Included
Section Description
✅ Complete Enhanced Script Ready-to-paste code with working free models and critical fixes
🔧 Setup Instructions Step-by-step guide from API keys to first workout
🚀 Improvement Recommendations 5 priority enhancements for the developer
📖 Maintenance Guide Weekly/monthly tasks to keep Aixle running
🔍 Troubleshooting Guide Common errors and quick fixes
⚡ Key Improvements Made
4 working free models identified and verified (arcee-ai/trinity-large-preview:free, stepfun/step-3.5-flash:free, z-ai/glm-4.5-air:free, nvidia/nemotron-3-nano-30b-a3b:free)

Critical duration fix – explicit seconds instruction (prevents 86-second workouts)

Multi-model fallback – tries 4 models in sequence

Better error logging – actionable messages for 401/404/429/503 codes

Test functions – testAvailableModels() and testOpenRouterConnection()

🎓 Testing Context
Date: March 1, 2026

Duration: 3+ hours active debugging

Platform: Google Apps Script

API Provider: OpenRouter (free tier)

Integration: Intervals.icu → MyWhoosh

📬 Purpose
This package is intended for:

End users who want a working Aixle setup

The original developer (kuukuu) to incorporate into future versions

Added wellness data to the script to adjust the workouts based on Garmin and Healthsyc integration.

aixle.rework.and.wellbeing.data.txt

[biscuitbloke-lab]

Aixle – AI Cycling Coach

Aixle is an intelligent, autonomous training system for cyclists. It generates daily AI‑powered workouts based on your actual fitness and wellness data, and adapts over time using progressive overload and failure analysis.

✨ Features

• AI‑generated workouts – personalised for your goals and current fitness
• Wellness sync – automatically locks HRV, sleep, weight, RHR from Intervals.icu
• Workout history tracking – avoids repetition and learns from your performance
• Progressive overload – increases intensity when you're ready, decreases when you struggle
• Failure analysis – detects when workouts are too hard and adjusts accordingly
• Rest / recovery logic – suggests rest days or recovery rides when fatigued
• FTP prediction – estimates FTP from normal rides, no tests required
• Difficulty prediction – labels workouts as Easy → Very Hard
• Conversational AI coach – ask questions and get answers based on your data
• Daily email – includes workout, wellness snapshot, and training progress
• Multi‑language – English, Japanese, Spanish, French

🔧 Prerequisites

• A Google Account (for Google Apps Script)
• An Intervals.icu account (free tier works)
• An OpenRouter account (free tier works)
• A Google Sheet to log your activities (create a blank sheet and copy its ID)

🚀 Installation

1. Create a new Google Apps Script project

   • Go to script.google.com
   • Click + New project

2. Paste the code

   • Delete any default code
   • Copy the entire Code.gs from this repository and paste it in

3. Configure your settings

   • Replace all placeholders in the USER_SETTINGS object:
     • YOUR_TRAINING_GOAL_HERE
     • YYYY-MM-DD (your target event date)
     • YOUR_GOOGLE_SHEET_ID
     • your.email@example.com
   • Replace the API keys in API_KEYS with your own

4. Add the Gmail service (if you want emails)

   • In the left sidebar, click the + (Services)
   • Add Gmail API

5. Initialize the adaptive state

   • Run the function initializeAdaptiveState() once

6. Set up triggers

   • Click the clock icon in the left sidebar
   • Add two triggers:
     • fetchAndLogActivities at 2 AM
     • generateOptimalZwiftWorkoutsAutoByGemini at 6 AM

7. Test your setup

   • Run testOpenRouterConnection() to verify your API key
   • Run testWellnessSync() to check wellness data sync
   • Run testAllNewFeatures() to verify the adaptive logic

📊 How It Works

Daily Workflow

1. 2:00 AM – fetchAndLogActivities() downloads your recent rides from Intervals.icu
2. 6:00 AM – generateOptimalZwiftWorkoutsAutoByGemini():
   • Syncs your wellness data (HRV, sleep, RHR)
   • Analyzes fatigue (TSB) and decides rest / recovery / normal day
   • Suggests a workout type based on training phase, wellness, and history
   • Calls OpenRouter AI to generate a .zwo workout file
   • Saves the file to Google Drive
   • Uploads the workout to your Intervals.icu calendar
   • Sends you an email with the workout + wellness snapshot

Adaptive Logic

• Base Phase → mostly endurance (Zone 2 / Tempo)
• Build Phase → mostly threshold / sweet spot
• Peak Phase → mostly VO2max / high intensity
• Fatigued (TSB < -15) → recovery ride
• Very fatigued (TSB < -25) → rest day
• High completion rate + low difficulty → intensity +5%
• Low completion rate → intensity -10%

🧪 Test Functions

Function	What it does
testOpenRouterConnection()	Verifies your OpenRouter API key
testAvailableModels()	Checks which free models are working
testWellnessSync()	Tests wellness data sync
initializeAdaptiveState()	Resets the adaptive state to defaults
checkAdaptiveState()	Displays current adaptive state
testAddSampleHistory()	Adds mock workout history (for testing)
testAnalyzeFailures()	Tests failure analysis on current data
testWorkoutSuggestions()	Tests workout type selection logic
testRecoveryLogic()	Tests rest/recovery decision logic
testFTPPrediction()	Tests FTP prediction
testDifficultyPrediction()	Tests workout difficulty labels
testAskCoach()	Tests the conversational AI coach
testAllNewFeatures()	Runs all new feature tests

🔐 Security Notes

• Your API keys and personal data are stored only in your Google Apps Script project.
• Never commit your actual keys to a public repository.
• Use the placeholder versions provided here.

📝 License

This project is licensed under the MIT License – see the LICENSE file for details.

🙏 Acknowledgements

• Intervals.icu for the amazing training platform
• OpenRouter for providing access to multiple AI models
• The cycling community for inspiration and feedback

🚴‍♂️ Happy Training!