A smart, AI-powered WhatsApp assistant built for Simranjeet Singh to automatically answer incoming messages using Gemini 3.0 Flash.
Unlike standard bots that rely on fragile keyword-matching, or simple RAG setups that answer blindly from raw documents, this bot uses Karpathy's LLM Wiki Architecture. It reads from a curated, internal "knowledge wiki" to synthesize personalized, highly accurate responses with embedded direct links.
- LLM Wiki AI Engine: The bot uses a combination of Intent Classification & Vectorless Retrieval to match user queries with structured knowledge stored in a central markdown wiki.
- Smart Routing: It automatically guides users to the right resources based on their questions:
- Technical deep-dives (RAG, LLMs)? → Links to exact YouTube playlists and GitHub repos.
- General chat/Networking? → Links to Topmate Quick Connect.
- Mentorship requests? → Links to Topmate 1:1 Guidance.
- Self-Documenting Log: Every interaction is timestamped and recorded directly into the
log.mdpage of the wiki for transparency. - Built-in Fallback: If the Gemini API is down, rate-limited, or disabled, the bot seamlessly falls back to 10 solid keyword-matching rules so no message goes unanswered.
- Admin Dashboard: A beautiful web UI to monitor active rules, total messages processed, AI vs rule-based response ratio, and system logs.
The bot's entire knowledge base lives locally inside server/wiki/. This means you can update the bot's knowledge simply by editing markdown files—no database or vector embeddings required!
| Markdown File | Purpose |
|---|---|
about_simranjeet.md |
Background, expertise, social profile links. |
github_projects.md |
Over 10+ GitHub repos mapped to specific AI domains. |
youtube_content.md |
8 dedicated Playlists to answer coding/tutorial questions. |
concepts_knowledge.md |
Explanations around RAG, LangChain, CrewAI, AutoGen, etc. |
topmate_booking.md |
Direct guidance on when to offer which Topmate session. |
medium_blogs.md |
List of publication topics written by Simranjeet. |
SCHEMA.md |
The strict system prompt holding tone guidelines and constraints. |
- Backend: Node.js, Express,
whatsapp-web.js - AI Engine:
@google/generative-ai(Gemini-2.0-Flash) - Database: SQLite (for chat logs and fallback rules)
- Frontend Dashboard: React, Vite, Tailwind CSS (optional)
- Node.js (v18+)
- NPM / Yarn
- A WhatsApp account (preferably a business / secondary number for the bot).
- A Gemini API Key (Free Tier works fine).
Clone the repository, then split into two terminals:
cd server
npm install
# Setup your API key
cp .env.example .env
# Edit .env and paste your GEMINI_API_KEY inside!
npm run devWait for the terminal to generate a QR Code. Scan it with your WhatsApp mobile app (Linked Devices) to authenticate the bot.
cd frontend
npm install
npm run devOpen http://localhost:5173 in your browser to view the monitoring dashboard.
User: "I want to build an AI agent, do you have any code for that?" Bot (AI): "🤖 Yes! I run an entire masterclass on Agentic AI. You can find all 15+ projects inside my Awsome_AI_Agents repository (155⭐). If you prefer video, watch the entire playlist here: 📺 Agentic AI 14 Projects. Let me know if you get stuck!"
User: "Can we jump on a call for career mentorship?" Bot (AI): "🎯 I'd love to help out! For personalized 1:1 guidance on getting into AI/ML, please book a session with me here below:\n🚀 https://topmate.io/simranjeet97/145435"
To update what the bot knows:
Open any file in the server/wiki/ directory and update the markdown. Changes will take effect immediately.
To change the AI personality/rules:
Edit server/wiki/SCHEMA.md.
To view recent chats:
Check the web dashboard or server/wiki/log.md.
