🧠 MindsDB KB App — Customer Support & Movie Q&A Agent

This project is built for the MindsDB Quest-19: Stress-Test Knowledge Bases challenge. It demonstrates a complete end-to-end pipeline using MindsDB, local Ollama LLMs, and Streamlit UI for:

• 📊 Customer support ticket analysis
• 🎬 Movie-related question answering
• ✅ Accuracy benchmarking

---

📌 Overview

This application performs the following:

✅ Customer Support Agent

• Creates a Knowledge Base (customer_tickets_kb)
• Ingests a CSV dataset (customer_support_tickets.csv) with metadata
• Builds a vector index for semantic search
• Creates an AI model using LLaMA3 via Ollama
• Deploys a retrieval-augmented agent (ticket_support_agent)
• Supports interactive Q&A and summarization of customer issues

✅ Movie Expert Agent

• Loads a summary-based knowledge base (movies_kb)
• Creates an agent (movie_expert_agent) using LLaMA2
• Answers plot, character, and theme-related movie questions

✅ Benchmarking

• Measures ingestion time, semantic query latency, and p95/p99 response delays
• Outputs structured .md reports under benchmarks/

---

📂 Dataset

1. customer_support_tickets.csv

Contains 8,469 support tickets with metadata:

• Ticket ID, Subject, Description
• Priority, Status, Type, Channel

3. imdb_movies_prepared.csv

Contains:

• Movie title, genre, actors, year, rating
• ✅ Original dataset size: 238,256 rows
• ✅ After removing duplicates: 161,765 rows
• ✅ Saved to imdb_movies_prepared.csv for clean ingestion

---

🧪 Project Structure

cd mindsdb-kb

---

🚀 How to Run

1️⃣ Install dependencies (Python ≥ 3.10)

python -m venv venv
source venv/bin/activate  # Or venv\Scripts\activate on Windows
pip install -r requirements.txt

2️⃣ Launch the Streamlit UI

🧠 Customer Support Agent:

streamlit run app.py

🎬 Movie QA Agent:

streamlit run movie.py

3️⃣ Run Benchmarking

python benchmark.py

---

🔍 What’s Measured in Benchmarks?

Each .md report includes:

Category	Metric
⏱️ Ingestion Time	Total seconds + time per 1K rows
⚡ Query Latency	Average, p95, p99 response times

---

Built with ❤️ using MindsDB, Ollama, Streamlit, and Python.