AI Database Chat Assistant

This project is an intelligent chat interface for exploring and analyzing retail sales data using advanced search and AI-powered tool calling. Built with React, TypeScript, Next.js, and Prisma, it leverages Orama for fast search indexing and OpenAI's GPT for natural language understanding and tool selection.

Features

• Conversational Search: Ask questions about products, sales, customers, and more in natural language.
• Semantic & Filtered Search: Uses Orama for both keyword and semantic (vector) search, enabling fast and relevant results.
• Tool Calling: The AI automatically selects the best tool for your query—whether it's searching products, analyzing sales, or summarizing statistics.
• Interactive Results: View results as tables, lists, charts, or text summaries directly in the chat.

How It Works

• Orama Indexing: Retail records are indexed with Orama for instant search and semantic queries.
• Tool Calling: OpenAI's GPT chooses from a set of tools (like search_products, get_top_selling_products, or vector_search) to answer your questions.
• Database Integration: Uses Prisma ORM to manage and query the PostgreSQL database.

Example: Orama Search & Tool Calling

// Indexing a retail record with Orama
await orama.insert({
  invoice: row.Invoice,
  stockCode: row.StockCode,
  description: row.Description,
  quantity: Number(row.Quantity),
  invoiceDate: new Date(row.InvoiceDate).toISOString(),
  price: Number(row.Price),
  customerId: row['Customer ID'],
  country: row.Country,
});

// Tool calling in the API route
const toolResults = await toolExecutor.executeTool({
  name: 'search_products',
  parameters: { country: 'United Kingdom' },
  id: 'tool-call-1'
});

Data Source

Retail data is from Online Retail II UCI, produced under the CC0: Public Domain license.

---

Created by Brian Mwangi - AI Engineer

---