Skip to content

CORE_LOGIC.md

Latest commit

 

History

History
62 lines (44 loc) · 6.02 KB

CORE_LOGIC.md

File metadata and controls

62 lines (44 loc) · 6.02 KB
Raw

🧠 AIVA: Core Logic & Processing Pipeline

This document defines the exact priority order and fallback mechanisms that AIVA uses to process every single voice or text command from the user.

AIVA is designed to be blisteringly fast and cost-efficient. To achieve this, she uses a strict "Waterfall" priority system. She checks the fastest, cheapest, and most specific local methods first, and only travels down the pipeline to the heavy AI Brain if the local methods cannot handle the request.

Here is the exact step-by-step logic AIVA follows the moment you hit "Send" or stop speaking:

🏎️ Priority 1: Local Offline Actions (0ms Latency)

Before AIVA even checks the internet, she checks her local system.

  • 🕒 Time & Date: If you ask "What time is it?" or "What is the date?", AIVA bypasses the AI completely and calculates the exact local time using the Node.js backend.
  • 🙋‍♀️ Identity: If you ask "Who are you?" or "Who made you?", she triggers a hard-coded introduction.
  • 🗣️ Voice Controls: If you say "Change voice to British" or "Speak in Hindi", she intercepts this command instantly, switches her Text-To-Speech engine, and stops processing.
  • 👋 Greetings & Small Talk: She scans the backend/data/responses.json offline lexicon. If you say "Hello", "Good morning", or "How are you?" in English, Hindi, Bengali, or several other languages, she immediately pulls a randomly generated, native response from the JSON file and replies instantly without ever waking up the AI.

💻 Priority 2: Native Desktop Controls (Windows Only)

If the command isn't a simple greeting or time check, AIVA checks if you want to control your computer.

  • 📧 Smart Email Drafting: If you say "Draft an email to my boss", AIVA intercepts the intent. She makes a quick call to the AI to write the body of the email, and then natively triggers your computer's default mail client (like Outlook or Gmail Web) with the entire email pre-typed and ready to send.
  • 🎛️ Windows OS Controls: If the command contains phrases like "Lock screen", "Mute volume", "Empty recycle bin", or "Sleep computer", AIVA executes native Windows PowerShell or RunDLL commands to physically control your PC.

🌥️ Priority 3: Weather Tracking (Dual API Fallback)

If the user is asking about the weather ("What is the weather in Delhi?"), AIVA isolates the city name and begins a dedicated API workflow:

  1. 🥇 Primary Engine (WeatherAPI.com): AIVA pings this API first for highly accurate, minute-by-minute forecasting.
  2. 🥈 Fallback Engine (OpenWeatherMap): If WeatherAPI is down, rate-limited, or fails to find the city, AIVA automatically catches the error and instantly switches to OpenWeatherMap.
  3. If both APIs completely freeze or crash, she returns a graceful failure message: "Weather services are currently unreachable."

🏏 Priority 4: Live Sports Integration

If you ask about ongoing matches ("What is the live score?"), AIVA looks for specific trigger words:

  • Cricket: If she hears "cricket", "IPL", or "World cup", she intercepts the query and pings the CricketData.org API to fetch live, ball-by-ball updates.
  • Football/Soccer: If she hears "football", "soccer", "Premier League", etc., she routes the query to the API-Football tracker to fetch live goal updates.
  • Note: If you ask about a match that happened in the past or will happen in the future, the Sports APIs will return back empty. AIVA detects this, dynamically realizes it wasn't a "live match" question, and pushes the query down to Priority 6 (Web Search + AI) to find the answer on the internet!

📰 Priority 5: Real-Time News

If you ask "What are the top 5 news headlines?", AIVA triggers her news sequence:

  1. 🥇 Primary Engine (GNews.io): She intercepts the command, extracting the requested number of headlines, and pings GNews.io, sorting by publishedAt to guarantee fresh breaking news.
  2. 🥈 Fallback Engine (NewsAPI.org): If GNews fails or hits its rate limit, AIVA instantly falls back to NewsAPI.org to ensure she never fails to deliver the world's top headlines.

🧠 Priority 6: The AI Brain + Search Fallback (Ultimate Fallback)

If the command was NOT a greeting, NOT a weather check, NOT a live sports score, and NOT a desktop control... it must be a complex question (e.g., "Summarize quantum physics" or "Who won the US Election?"). AIVA now brings out the heavy artillery.

  1. 🥇 Primary AI Generation (Gemini 2.5 Flash): AIVA calls Gemini. It is incredibly fast and parses standard intelligence queries with a brilliant, context-aware answer.
  2. 🥈 Secondary AI Backup (Groq Llama 3.3 70B): If Gemini reaches its API limits or goes offline, AIVA flawlessly catches the error and instantly re-routes the exact same prompt to Groq Llama 3.3 70B as her bulletproof backup brain.
  3. 🥉 Tertiary Web Search (Tavily AI): If both AI models fail, or if the system specifically falls down the intelligence hierarchy, AIVA pings Tavily AI Search to pull highly structured, real-time internet context directly to the user.
  4. 🎖️ Quaternary Web Search (Google Programmable Search): If Tavily is unreachable, she invokes a final failsafe, using Google Custom Search Engine to scrape directly from Google's index.
  5. ⚠️ Error Handling: If all four redundant intelligence and search layers experience failure, AIVA safely catches the error and tells the user: "I'm having trouble connecting to my cloud brain and search engines right now..."

🎓 Summary of the Flow:

  1. Local JSON & Rules (0ms - Free) -> Matches? YES -> Talk!
  2. OS/Email Controls (Fast - Free) -> Matches? YES -> Execute!
  3. Targeted APIs (Weather/News/Live Sports) -> Dual-layer Fallbacks -> Matches? YES -> Talk!
  4. Primary AI (Gemini 2.5 Flash) -> Generates intelligent response.
  5. Secondary AI (Groq Llama 3.3) -> Fallback text generation.
  6. Web Search Fallbacks (Tavily -> Google) -> Final failsafe for external queries.