[Feature] Scheduled next-day recovery feedback prompt

[shchukins]

Parent

Part of #82 Unified Subjective Feedback Model.

Related to:

• [Feature] Next-day recovery subjective feedback #84 Next-day recovery subjective feedback
• event-driven readiness/recovery workflows

Idea

Automatically send next-day recovery feedback prompts after relevant training days.

The current system supports:

• manual recovery prompt triggering
• Telegram recovery feedback collection
• idempotent recovery feedback persistence

But prompt delivery still requires a manual debug endpoint call.

This feature introduces scheduled orchestration for longitudinal recovery feedback collection.

Problem

Without automated prompt scheduling:

• users forget to provide recovery feedback
• longitudinal subjective dataset becomes sparse
• delayed recovery signal quality decreases
• readiness calibration becomes weaker
• recommendation validation quality decreases

Manual triggering is acceptable for debugging but not for production data collection.

Proposed solution

Introduce a scheduled recovery feedback workflow.

Every morning:

1. identify users with relevant previous-day training load
2. check whether next-day recovery feedback already exists
3. check whether recovery prompt was already sent
4. send lightweight Telegram recovery prompt
5. persist prompt delivery state

Prompt UX

Example:

How recovered do you feel today?

😴 Exhausted
😐 Tired
🙂 Okay
⚡ Fresh
🚀 Very fresh

Core UX principles:

• one tap
• optional participation
• low friction
• high longitudinal consistency

Scheduling behavior

Suggested initial logic:

Run daily scheduler at configurable morning time.

For each user:

Send prompt only if:

• previous day contains relevant training load
• or previous day contains activities
• and no recovery feedback exists yet
• and prompt was not already sent

Requirements

• at most one recovery prompt per day
• idempotent scheduling behavior
• no duplicate Telegram prompts
• timezone-aware scheduling support
• configurable quiet hours/fallback timing
• support future event-driven sync workflows
• persist prompt delivery metadata

Suggested persistence

Possible new table:

subjective_feedback_prompt_log

Suggested fields:

• user_id
• prompt_type
• target_date
• sent_at
• source
• delivery_status
• telegram_message_id
• created_at

Future integration

This scheduler should later integrate with:

• HealthKit freshness-aware morning briefing
• event-driven sync workflows
• recommendation engine calibration
• adaptive prompt timing
• notification fatigue prevention
• multi-channel feedback collection

Acceptance criteria

• scheduler runs automatically
• recovery prompts are sent after training days
• duplicate prompts are prevented
• users without recent load do not receive prompts
• repeated button taps update existing feedback
• prompt delivery state is persisted
• logs and tests cover scheduling/idempotency behavior
• documentation updated

Out of scope

• ML-based adaptive scheduling
• iOS push notifications
• conversational feedback flows
• sport-specific second-tap feedback
• notification prioritization engine

Complexity

Medium.

Value

Very high.

Automated longitudinal recovery feedback collection is critical for:

• readiness calibration
• adaptation modeling
• prediction validation
• future ML dataset quality