Skip to content

README.md

Latest commit

 

History

History
169 lines (109 loc) · 3.47 KB

README.md

File metadata and controls

169 lines (109 loc) · 3.47 KB

Intake logo

🥗 Intake API

Track food, not numbers.
Backend API for an AI-powered food diary focused on simplicity, daily use, and long-term clarity.

📌 Overview

Intake API is the backend service for Intake — a minimalist, AI-powered food diary.

The idea is intentionally simple:

You describe what you ate — the system does the rest.

Users log food using natural language.
The backend parses the input with AI, validates it, stores results by day, and exposes a clean, predictable API for tracking and statistics.

This repository contains only the backend API.
Frontend lives here 👉 https://github.com/bohdan-strilets/Intake-web

🎯 Project Status

  • ✅ MVP completed
  • 🧪 Pet project
  • 💼 Portfolio-ready
  • 🔒 API contracts frozen

The backend is considered stable.
Future work is expected mostly on the frontend and UX.

🧠 Core Principles

  • Backend is the single source of truth
  • No calorie or macro calculations on frontend
  • Derived data is never persisted
  • Explicit, readable, predictable code
  • Minimal API surface
  • No premature optimizations

🧠 Design Decisions

Backend as the source of truth

All calculations are handled on the backend. The frontend only renders validated data.

Derived data is never persisted

Statistics are always calculated from daily records. This keeps the data model simple and consistent.

Day as an aggregate root

A Day represents a single daily aggregate. Food entries are atomic and always belong to a Day.

AI behind a strict boundary

AI output is never trusted directly. All responses are validated before persistence.

Minimal API surface

Only endpoints required for daily usage are exposed. No speculative or “just in case” features.

No food catalogs or barcodes

The product avoids branded databases and barcode scanning. Focus is on awareness, not nutritional micromanagement.

Readability over abstraction

Explicit code is preferred over clever patterns. Long-term maintainability is the priority.

🧱 Architecture

Tech Stack

  • Framework: NestJS
  • Language: TypeScript (strict)
  • Database: MongoDB (Mongoose)
  • Auth: JWT + Refresh Tokens
  • AI: OpenAI (structured JSON output)
  • Validation: class-validator + Zod
  • Docs: Swagger (OpenAPI)

📂 Main Modules

  • Auth — registration, login, refresh, logout
  • Users — profile & body parameters
  • Days — calendar & daily totals
  • Food — add/delete food entries
  • AI — parse food descriptions (JSON-only)
  • Stats — derived statistics (read-only)

🚀 Getting Started

Prerequisites

  • Node.js ≥ 18
  • MongoDB (local or Atlas)

Install

npm install

Development

npm run start:dev

Production

npm run start:prod

🔗 Related

Frontend (Web):
https://github.com/bohdan-strilets/Intake-web

📄 License

MIT License

✍️ Author

Bohdan Strilets
Software engineer

Portfolio project — Intake