🚀 NeonAI

Local-First Multi-Mode AI System (Experimental)

Mode-Driven Intelligence • Privacy Focused • System > Model

---

🧠 What Is NeonAI?

NeonAI is a local-first AI system designed to run primarily on your own machine using a local LLM pipeline, with optional and controlled internet access for selected features.

What started as an experiment gradually evolved into a complete AI system architecture with multiple operational modes, strict behavior control, and a premium custom UI.

⚠️ This is not a chatbot wrapper. NeonAI is an AI system with modes, rules, confidence gates, memory, and decision pipelines.

---

✨ Core Philosophy

• 🧠 Local LLM First — No mandatory cloud LLM APIs.
• 🔒 Privacy-Focused — Data stays on the user’s machine.
• 🎯 Mode-Driven Intelligence — AI behavior depends on context.
• 🧪 Experimental by Design — Built to explore system ideas.
• 🧩 System > Model — The LLM is a tool, not the decision-maker.

---

🎮 Modes & Functionality

NeonAI is a multi-mode AI assistant that can switch between different roles, each with its own rules and permissions.

Mode	Purpose
NEON CASUAL	General chat using a local LLM with memory.
NEON MOVIES	Movie discovery, recommendations & metadata.
NEON STUDY	PDF-based syllabus learning (Strict Offline, No Hallucinations).

Each mode enforces different constraints, memory usage, and access permissions.

---

🧱 System Architecture

Key Principle: The LLM never directly decides responses. All outputs pass through rules, confidence checks, and mode restrictions.

graph TD;
    User-->Frontend_UI;
    Frontend_UI-->Flask_Backend;
    Flask_Backend-->Brain_Layer;
    Brain_Layer-->Waterfall_Logic;
    Waterfall_Logic-->Confidence_Gate;
    Waterfall_Logic-->Mode_Rules;
    Mode_Rules-->Local_LLM;

Loading

---

🖥️ Frontend (UI)

• Pure HTML, CSS, JavaScript (No frameworks).
• GSAP-powered animations.
• 10+ Neon Themes + Light / Dark mode.
• Physics-based Liquid Toggle.
• Fully responsive (Desktop + Mobile).

Note on Structure:

• templates/: Contains UI files (index.html, styles.css, app.js) served via Flask.
• static/: Currently used for user-uploaded wallpapers.
• The frontend is not an SPA. It is a controlled UI panel, intentionally simple and stable.

---

🎬 NEON MOVIES Mode

• Displays trending movies.
• Uses TMDB API (Optional).
• Auto-scroll carousel with hover/touch pause.
• Learns user genre preferences.
• Offline Fallback: Uses a local movie database if internet is unavailable.

📚 NEON STUDY (Exam Mode)

• RAG Pipeline: Upload a PDF syllabus.
• Indexes content locally using a Vector Database.
• Strict Rule: If the answer is not present in the PDF → The AI Refuses.
• Internet access is Completely Blocked in this mode.
• Designed for exam-safe, hallucination-free learning.

---

🧠 Local LLM Engine

NeonAI does not depend on cloud LLM APIs.

• Powered by local models (via Ollama).
• Fully offline capable.
• Internet access is optional and gated.
• External APIs never override system rules.

---

📂 Project Structure

NeonAI/
│
├── server.py                  # Flask backend (API + routing)
├── START_NEON.bat             # One-click launcher (Windows)
├── requirements.txt           # Python dependencies
├── README.md                  # Project documentation
│
├── brain/                     # Core AI system logic
│   ├── waterfall.py           # Decision flow & routing
│   ├── confidence_gate.py     # Confidence & hallucination control
│   ├── memory.py              # Session & preference memory
│   └── gk_engine.py           # Knowledge & reasoning engine
│
├── models/                    # LLM abstraction layer
│   ├── local_llm.py           # Local LLM interface (offline-first)
│   └── hybrid_llm.py          # Local + optional online logic
│
├── exam/                      # NEON STUDY (Exam / PDF Mode)
│   ├── indexer.py             # PDF indexing
│   ├── retriever.py           # Strict PDF-only retrieval
│   └── uploads/               # User-uploaded PDFs (gitignored)
│
├── movie/                     # NEON MOVIES engine
│   ├── engine.py              # Recommendation logic
│   └── lookup.py              # Movie metadata handling
│
├── web/                       # Controlled web adapters
│   ├── search_adapter.py      # Web search (Tavily / DDG)
│   └── movie_adapter.py       # TMDB integration
│
├── templates/                 # Frontend HTML templates
│   └── index.html
│
├── static/                    # Frontend static assets
│   ├── styles.css
│   ├── app.js
│   └── wallpapers/            # User background images (runtime)
│
├── user_data/                 # Local user state (offline-first)
│   └── profile.json
│
└── .gitignore                 # Ignore runtime & private data

---

▶️ How to Run

1️⃣ Requirements

• Python 3.10+
• Local LLM runtime (e.g., Ollama) installed and running.

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Start NeonAI

Double-click START_NEON.bat Or run manually via terminal:

python server.py

Open in Browser: http://localhost:5000

---

🧪 Project Status

• ✅ Core system functional
• ✅ UI stable & responsive
• ✅ Multi-mode logic working
• ⚠️ Experimental (Architecture locked for iteration)

---

⚠️ Disclaimer

This is an experimental project built for learning, research, and AI system design exploration. It is not a commercial product.

---

🧠 Author

Ansh

B.Tech CSE

Focus Areas:

AI Systems (not just models) • Offline-first Architecture • Controlled AI Design

"NeonAI is not about how smart the model is. It’s about how controlled, safe, and purposeful AI should be."