You are helping me build the following AI agent workflow.

## Goal
French Fitness Coach on Telegram: Chat in French on Telegram for fitness coaching with voice notes, diet logs, and workout scheduling.

## Specification
- What it does: Get a personal French-speaking fitness and nutrition coach directly inside your Telegram chat that handles both text and voice messages. The system transcribes your audio notes, remembers your training history and diet logs from Google Sheets, and generates tailored coaching advice to help you reach your goals. You can also ask it to schedule specific workout sessions on your calendar for later in the week.
- Trigger: Triggered by an event (Event · on new Telegram message)
- Autonomy: Fully hands-off
- Expected setup effort: about an afternoon
- Tools/services involved:
  - Telegram: receiving text and voice messages from users
  - OpenAI: transcribing voice notes into text
  - Anthropic Chat Model: generating French fitness and nutrition coaching responses
  - Google Sheets: logging workouts, meals, and storing user profiles
  - Google Calendar: scheduling workout sessions based on user requests

## Known pitfalls, handle each one explicitly in your implementation
1. Audio transcription often fails on noisy environments, so add a fallback prompt asking users to re-record clearly if confidence is low.
2. Calendar creation requires strict permission scopes, so ensure the service account has 'https://www.googleapis.com/auth/calendar' explicitly granted.

## Reference implementation
https://n8n.io/workflows/16610 (workflow template)
Fetch and inspect this before building. If it matches my stack, adapt it;
if not, rebuild the pattern with my tools.

## Process requirements
1. Before building: ask me which of the listed tools I actually use and
   what my platform is (n8n / Make / code / other). Do not assume.
2. Adapt the pattern to my answers; do not force the reference stack.
3. Address every pitfall above; tell me how you handled each.
4. Provide a test plan I can run before letting this touch real data.
5. Ask before any step that sends messages, modifies data, or spends money.

Source: https://usecasesforagents.com/use-case/french-fitness-coach-on-telegram/ via usecasesforagents.com