​

Nutrition tracking

Nutrition tracking is the first core product loop for Moveat. Users log what they ate, Platform stores structured meal entries, and the product shows daily calorie progress.

​

Business goal

The goal is not perfect nutrition science at MVP stage. The goal is useful, consistent tracking with clear estimation and correction paths.

​

Meal entry model

A meal entry represents one food logging event. It can come from:

• Web app.
• Agent text message.
• Agent image interpretation.
• Agent audio interpretation.

A meal entry can contain multiple meal items.

​

Meal item model

Each item stores estimated nutrition values:

• Name.
• Estimated quantity.
• Estimated unit.
• Estimated calories.
• Protein grams.
• Carbs grams.
• Fat grams.

The current model accepts estimated values inline. A future food catalog can improve precision by resolving food density and macros from trusted records.

​

Idempotency

Agent-driven meal logging must be idempotent. If WhatsApp retries a webhook, Platform should not double-count the same meal. For WhatsApp, the webhook message ID should be used as the idempotency key.

​

Daily summary

Platform updates a daily nutrition summary when a meal is logged. The summary stores:

• User.
• Local date.
• Calories consumed.
• Protein consumed.
• Carbs consumed.
• Fat consumed.
• Calorie target snapshot.

The target snapshot preserves what target applied on that day.

​

Why daily summary exists

Daily summaries make dashboard reads cheap and provide stable historical views. They should remain rebuildable from meal entries if needed.

​

Estimation risk

LLMs can estimate food and portions, but they can also be wrong. Product UX should make estimates visible and correctable. A good user-facing pattern is:

1. Agent estimates the meal.
2. User confirms or corrects.
3. Platform stores the final structured entry.
4. Daily summary updates once.