chatbot-implementation

Chatbot Logic

Overview

A specialized RAG (Retrieval Augmented Generation) chatbot that helps users learn from the textbook content.

Backend

• Route: app/api/chat/route.ts
• Logic:
  1. Receives query and history.
  2. Embeds query using Gemini or OpenAI embedding model.
  3. Searches Qdrant (vector DB) for relevant textbook chunks.
  4. Constructs context from matches.
  5. Generates response using Gemini Flash/Pro.

Vector Search (Qdrant)

We use Qdrant for storing embeddings of the textbook.

• Collection: textbook_chunks (or similar).
• Fields: text, source, chunk_id.

UI Component

• Location: textbook/src/components/Chatbot/index.tsx.
• Features:
  • Floating chat window.
  • Size controls (Small, Medium, Large).
  • Markdown rendering of responses.
  • Context selection (highlight text to ask about it).
  • Mobile responsive design.
  • Auth awareness (personalizes answer based on user profile).

Styling

• CSS: styles.module.css (Premium animations, shadow effects).
• Themes: Dark/Light mode compatible (using --ifm variables).