Skip to content

Architecture Overview

ChatKcal is an "Intelligence First" calorie tracker. Instead of maintaining a food database, it leverages the user's existing AI (ChatGPT, Claude, Gemini) to analyze food and structured data.

1. Product Philosophy

  • The Anti-Database: We don't want a search bar; we want a conversation.
  • The Sidecar Model: ChatKcal acts as the "Long-Term Memory" for ephemeral AI chats.
  • Zero Friction: The transition from "Chat" to "Log" must be instantaneous via Deep Links.
  • Minimalist UI: The app ingests structured data and visualizes daily progress.

2. User Experience (UX) Flow

  1. Setup: User installs a "System Prompt" into their AI's custom instructions.
  2. Conversation: User sends a photo of food to their AI.
  3. Link Generation: The AI analyzes the food and generates a deep link (e.g., chatkcal.app/log?meal_json=...).
  4. Ingestion: User clicks the link. ChatKcal parses the JSON, logs it to DynamoDB, and updates the dashboard.

3. Technical Stack

  • Frontend: React + Vite + Tailwind CSS (hosted on AWS Amplify).
  • Auth: AWS Cognito User Pools.
  • Backend: AWS AppSync (GraphQL) with JavaScript Resolvers.
  • Database: AWS DynamoDB (Single Table Design).
  • Infrastructure: AWS SAM (Serverless Application Model).

4. Documentation Map