Daily TL;DR

Team: Code & Chronicle

Members: Jay Layco, Ashley Shaw-Strand, Jack Gess

Daily TL;DR is a personalized, privacy-first news aggregator built for Android. Unlike standard news apps that bombard users with infinite feeds, this app generates a single, finite Daily Digest every morning at 5:00 AM using AI to scrape, summarize, and curate content based on the user's preferred topic.

Key Features

• AI-Powered Summaries: Uses Google Gemini (Flash 2.5) to transform complex articles into concise bullet points and 3-sentence synopses.
• Time Travel Carousel: Users can swipe back to view digests from the past 7 days (Read-Time Aggregation).
• Audio Mode: Native Text-to-Speech (TTS) integration allows users to listen to their news while commuting.
• Dynamic Theming: Full Dark Mode support that persists across app launches.
• Rate Limiting: Server-side protection prevents API quota exhaustion by enforcing a 15-minute cooldown on manual refreshes.

Tech Stack & Architecture

Frontend (Client)

• Framework: React Native (0.76+)
• State Management: React Context API (Auth, Theme, Settings)
• Persistence: Firestore (User Data) + AsyncStorage (Local Preferences)
• Pattern: Modular Firebase SDK (Functional Programming style)

Backend (Serverless)

• Platform: Firebase Cloud Functions (Gen 2)
• Runtime: Node.js 20
• Scheduling: Google Cloud Scheduler (Cron Jobs)
• External APIs: NewsAPI (Discovery), Firecrawl (ETL/Scraping), Google Gemini (Summarization)

Core Computer Science Concepts

This project implements several advanced distributed system patterns:

1. The "Scatter-Gather" Pattern (Parallel Processing)

Instead of processing news articles sequentially (which would be slow), our backend utilizes concurrency:

1. Scatter: The system spawns 5 parallel promises to scrape (Firecrawl) and summarize (Gemini) articles simultaneously.
2. Gather: using Promise.allSettled, the system collects the results.
3. Fault Tolerance: If one article fails (e.g., paywall), the digest still generates successfully with the remaining data (Fail-Open Design).

2. Read-Time Aggregation (Data Structures)

To solve the "Greatest-N-Per-Group" problem (showing only the latest digest for each of the last 7 days), we shifted compute to the client:

• Query: Fetches the last 7 days of raw data.
• Algorithm: Uses a Hash Set (Set<String>) in the React Native client to filter duplicates in O(1) time complexity, ensuring the UI remains responsive.

3. Determinism in AI

Large Language Models are probabilistic. We enforced Determinism by strictly defining a JSON Schema for the Gemini API. This guarantees that the AI output is always valid JSON structure, preventing UI crashes caused by "hallucinated" formats.

Getting Started

Prerequisites

• Node.js (v18+)
• Java Development Kit (JDK 17)
• Android Studio (for Emulator)

Installation

1. Clone the repository:
   git clone [https://github.com/your-repo/daily-tldr.git](https://github.com/your-repo/daily-tldr.git) cd daily-tldr

2. Install Dependencies:
   npm install
   cd functions && npm install && cd ..

3. Environment Setup:
   Create a .env file in the root:
   NEWS_API_KEY=your_key_here
   FIRECRAWL_API_KEY=your_key_here
   GEMINI_API_KEY=your_key_here

4. Run the App:
   npm run android

*Academic Project for SAIT Software