🤖 DecodeBot — Rule-Based AI Chatbot

A deterministic, dictionary-driven chatbot built in Python, developed as part of the DecodeLabs AI Internship — Project 1. Ships with three interfaces — terminal, desktop (Tkinter), and web (Streamlit) — all powered by the exact same rule engine, so behavior never drifts between them.

📌 Project Overview

DecodeBot answers user questions using explicit, hard-coded rules — no machine learning, no external API calls. It's the "logic engine" foundation that comes before generative AI: every response can be traced from input → logic → output with zero hallucination risk.

✨ Features

• Continuous conversation loop with a clean exit command
• Input sanitization (case-insensitive, whitespace-trimmed)
• 20+ intents: greetings, small talk, bot info, jokes, time/date, mood detection, and more
• Two-tier intent matching: exact trigger phrases first, then bag-of-words keyword scoring for paraphrased questions
• Typo tolerance via fuzzy matching (suggests the closest known phrase)
• Built-in calculator ("calculate 12*4") — parsed safely via Python's ast module, never eval(), so it can't execute arbitrary code
• Login system — username/password accounts, salted SHA-256 password hashes, plus a ready-to-use demo account
• Persistent chat history — saved per-user to disk; viewable and clearable from the GUI
• Lightweight session memory (remembers the user's name, message count)
• Fallback response for anything outside its rule set
• Three interfaces: terminal script, Tkinter desktop app, Streamlit web app

🛠️ Technologies Used

• Python 3
• Tkinter (desktop GUI)
• Streamlit (web GUI)
• Standard library: re, random, datetime, difflib, ast, hashlib, json

📂 Project Structure

decodebot/
│
├── chatbot_logic.py   # Core rule engine (intents, matching, replies, calc)
├── calculator.py       # Safe AST-based arithmetic evaluator
├── login.py             # Username/password auth (salted hashes)
├── file_manager.py     # Per-user persistent chat history
├── chatbot.py            # Terminal interface
├── gui.py                 # Tkinter desktop interface (login -> chat)
├── main.py                 # Entry point for the desktop app
├── app.py                    # Streamlit web interface
├── requirements.txt
├── README.md
└── LICENSE

users.json and the history/ folder are created automatically the first time you run the app — they aren't shipped with the project.

🚀 Installation

1. Download this folder.

2. Install dependencies:

   pip install -r requirements.txt

   Tkinter ships with most Python installs. On minimal Linux systems, if import tkinter fails, install it with: sudo apt-get install python3-tk

▶️ How to Run

Interface	Command
Terminal	python chatbot.py
Desktop app (GUI, login, history)	python main.py
Web app (Streamlit)	streamlit run app.py

Desktop app demo account: username demo, password demo123 (auto-created on first run — or click "Create Account" to make your own).

📈 How It Works

1. Input — read from the terminal, Tkinter chat box, or Streamlit chat.
2. Sanitization — lowercased and stripped of extra whitespace.
3. Matching
   • Tier 0: is this a plain math expression? → routed to the calculator.
   • Tier 1: exact trigger phrases (e.g. "how are you").
   • Tier 2: bag-of-words keyword overlap scoring for paraphrased input.
   • Tier 3: fuzzy typo suggestion (e.g. "helo" → "hello").
4. Output — the matched intent's reply (some, like time/date, are computed dynamically).
5. Persistence (desktop app only) — every message is appended to the logged-in user's history file.
6. Loop — continues until an exit command (exit, quit, bye, etc.).

🔒 A Note on the Login System

This is a learning-project login demo, not production-grade auth: passwords are salted and hashed (never stored in plain text), but there's no rate limiting, session expiry, or HTTPS transport involved. Good enough to demonstrate the concept — not something to reuse for a real product without hardening it further.

🎯 Future Improvements

• Add multi-turn conversational context
• Hybrid mode: fall back to an LLM when no rule matches
• Sync chat history across devices (currently local file storage only)
• Deploy the Streamlit version online via Streamlit Community Cloud

👩‍💻 Developed As Part Of

DecodeLabs AI Internship — Project 1: Rule-Based AI Chatbot

📄 License

Licensed under the MIT License — see LICENSE for details.