👻 ECHO

The Living Institutional Memory Engine for Universities

"Every batch leaves. Make sure their knowledge didn't take everything with them."

---

📌 The Problem

At modern engineering colleges and universities, senior wisdom is trapped in ephemeral WhatsApp groups, late-night DMs, and hallway conversations.

• Which professor is strict with midterms?
• How do you balance an IIT Madras dual degree with 48-hour project sprints?
• How did past juniors crack international internships in Singapore or the Apple Academy?

When senior batches graduate, years of institutional survival knowledge vanish forever. Junior batches repeatedly make the same mistakes, ask the same unanswered questions, and struggle with the same unwritten rules.

---

💡 What is ECHO?

ECHO is an AI-powered institutional memory engine that continuously captures, organizes, and synthesizes campus knowledge from both voice recordings on the web and real-time conversations across WhatsApp group chats.

       ┌────────────────────────────────────────────────────────┐
       │                 CAMPUS KNOWLEDGE SOURCES                │
       └────────────────────────────────────────────────────────┘
                 │                                    │
    🎙️ Voice Notes & Web Submissions        💬 WhatsApp Group Chats
                 │                                    │
                 ▼                                    ▼
       [ Groq Whisper Large v3 ]             [ Baileys Event Stream ]
        • Speech-to-Text Transcription        • Multi-lingual Intent Detection
                 │                            • Q&A Verification (Groq/Gemini)
                 ▼                                    │
       [ Gemini Embeddings Engine ] ◄─────────────────┘
        • 3072-Dimensional Vector Generation
        • Cross-Group Deduplication & Consensus Merging
                 │
                 ▼
       ┌────────────────────────────────────────────────────────┐
       │             ECHO MULTI-CHANNEL CORE (SQLite)           │
       │    • Persistent Knowledge Store   • Time-Decay Engine  │
       │    • Knowledge Gaps Tracker       • Consensus Metrics  │
       └────────────────────────────────────────────────────────┘
                 │                                    │
                 ▼                                    ▼
       🌐 WEB APP DISCOVERY                  💬 WHATSAPP BOT REPLIES
        • Semantic Cosine Search              • Automated Instant Answers
        • Senior Consensus Synthesis          • DM Fallback for Locked Groups
        • Interactive Knowledge Radar         • 👻 Silent Confirmation Reactions

---

🚀 Key Features

1. 💬 Autonomous WhatsApp Knowledge Bot

• Multi-Lingual Question Intent Detection: Understands questions in plain English, Hindi, and Hinglish (e.g. "Pushkar sir kaisa padhate h", "attendance criteria kya hai"), with or without question marks.
• Instant Auto-Replies: When a student asks a question in any group, Echo semantically matches the knowledge base and delivers a verified answer in seconds.
• Locked Announcement Channel Support: If the bot lacks permission to post in a restricted announcement channel, it privately direct-messages (DM) the student the verified answer.
• Passive Conversation Capture: When a senior answers a junior's question in chat, Echo verifies the pair via LLM and saves it to the central repository with a 👻 ghost reaction.
• Silent Confirmation Voting: Reactions like 👍, 💯, or messages like +1 / vouch automatically boost the memory's confidence score across groups.

2. 🧠 "Ask the Batch" — Consensus Synthesis

• Instead of returning disconnected search links, Echo analyzes matching senior memories and uses Google Gemini & Groq to generate a single synthesized consensus answer.
• Displays a live Senior Agreement Meter showing the percentage of seniors aligned on the advice.

3. 🎙️ 60-Second Senior Voice Studio

• Built-in browser audio recorder powered by native MediaRecorder and real-time audio equalizers.
• Automatic transcription via Groq Whisper Large v3 Turbo with near-zero latency.
• Tagged by Course Code, Professor, and Category with a real-time live preview card.

4. 📡 The Knowledge Gaps Radar

• Automatically logs every student search query that currently lacks a senior answer.
• Visualizes campus blind spots ranked by question frequency, enabling senior mentors and student councils to address high-demand topics with 1-click answer workflows.

5. 🧬 Time-Decay Memory Health

• Every piece of advice has a dynamic health score based on time-decay math and confirmation counts:
  • 🟢 Fresh (< 6 months / recently reconfirmed)
  • 🟡 Aging (6–12 months)
  • 🔴 Stale (> 1 year / outdated curriculum)
• Students can click "Mark as Still True" to re-verify older advice and restore freshness.

6. 🏷️ JIT (Just-In-Time) QR Handover

• One-click printable QR code generation for any specific Echo.
• Stick them on hardware lab doors, hostel noticeboards, or professor cabins so juniors can scan and listen to senior notes right where they need them.

---

🛠️ Technology Stack

Domain	Technology	Purpose
Backend	Python 3.11+, Flask	High-throughput REST API and templating engine
Speech-to-Text	Groq Whisper Large v3 Turbo	Blazing-fast voice note transcription
Vector Embeddings	Google Gemini gemini-embedding-001	3072-dimensional semantic vector search
Synthesis & LLM	Google Gemini 3.1 Flash + Groq	Multi-perspective consensus answer synthesis
WhatsApp Layer	Node.js + @whiskeysockets/baileys	Multi-device WebSocket connection to WhatsApp
Database	SQLite + NumPy	Local vector math and institutional storage
Styling & UI	Vanilla CSS3 (Custom Design System)	Glassmorphism, CSS keyframe animations, dark mode
Deployment	Railway	Microservices architecture with persistent volumes

---

🏁 Quick Start (Run Locally)

1. Prerequisites

• Python 3.11 or higher
• Node.js 18+ (for WhatsApp Bot)
• Free API keys from Groq Console and Google AI Studio

2. Backend Setup

# Clone the repository
git clone https://github.com/GyatsoYT/Echo.git
cd Echo

# Install Python dependencies
pip install -r requirements.txt

# Create environment configuration
cp .env.example .env

Edit .env with your API keys:

GROQ_API_KEY=gsk_...
GEMINI_API_KEY=AIzaSy...
SECRET_KEY=your-secret-key

Run the Flask server:

python app.py

Visit http://localhost:5000 in your browser.

---

3. WhatsApp Bot Setup

cd whatsapp-bot
npm install
node bot.js

Scan the QR code printed in the terminal (or at http://localhost:5000/bot/qr) using WhatsApp on your phone (Linked Devices).

---

🏛️ Architecture & Data Model

-- Core Memory Units (Audio, Transcripts, Vector Embeddings)
CREATE TABLE echoes (
    id                INTEGER PRIMARY KEY AUTOINCREMENT,
    course_tag        TEXT    NOT NULL,
    professor_tag     TEXT,
    topic_tag         TEXT,
    transcript        TEXT    NOT NULL,
    audio_path        TEXT    DEFAULT '',
    embedding         BLOB    NOT NULL,
    confirmations     INTEGER DEFAULT 1,
    source            TEXT    DEFAULT 'web',
    question_context  TEXT,
    group_names       TEXT    DEFAULT '[]',
    group_count       INTEGER DEFAULT 1,
    last_confirmed_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    created_at        TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

-- Search Analytics (Powers the Knowledge Gaps Radar)
CREATE TABLE searches (
    id                INTEGER PRIMARY KEY AUTOINCREMENT,
    query_text        TEXT    NOT NULL,
    best_match_score  REAL,
    created_at        TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

---

🌐 Production Deployment (Railway)

ECHO is configured for turnkey multi-service deployment on Railway:

1. Flask Web API: Deployed with Procfile / railway.toml using gunicorn app:app.
2. WhatsApp Bot Worker: Deployed in whatsapp-bot/ using persistent volume storage for WhatsApp session keys (/data/auth_session).
3. Zero-Config Persistent Storage: Configured via DATABASE_PATH and UPLOAD_FOLDER environment variables.

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

_{Built for the Scaler School of Technology (SST) Community · People graduate. Knowledge shouldn't.}