Add Context-Aware Customer Support Agent with Langbase (Bottom-Up Learning)

[gwho]

🤖 Context-Aware Customer Support Agent with Langbase

Overview

This PR adds a comprehensive educational project that teaches RAG (Retrieval Augmented Generation) from first principles using Langbase primitives. Built with a bottom-up learning approach - no black boxes, just explicit SDK calls and well-commented code.

🎯 What This Adds

A complete learning project that builds an AI support agent step-by-step:

📚 Core Learning Scripts (4 Progressive Steps)

1. 1-memory-creation.ts - Memory Primitive (Data Layer)

   • Demonstrates parsing, chunking, and embedding
   • Uploads FAQ knowledge base to vector database
   • Explains how semantic search works

2. 2-retrieval-test.ts - Retrieval Testing (Logic Layer)

   • Tests semantic search WITHOUT the LLM
   • Shows raw chunks and similarity scores
   • Teaches top_k parameter tuning

3. 3-pipe-creation.ts - Pipe Primitive (Cognition Layer)

   • Creates AI agent with system prompt
   • Configures model, temperature, and memory attachment
   • Demonstrates the full cognition layer

4. main.ts - Full Orchestration

   • Production-ready RAG pipeline
   • Interactive and single-query modes
   • Debug mode to see retrieval details

🧪 Mini-Projects (4 Hands-On Experiments)

1. Personality Swap - Test 5 different personas (pirate, sarcastic, Shakespeare, etc.)
2. Knowledge Injection - Add new documents without code changes (Quantum Spaghetti recipe!)
3. Accuracy Tuner - Optimize top_k retrieval parameter (1 vs 20 chunks)
4. Multi-Format Challenge - Test CSV, structured text, and other formats

📖 Documentation

• README.md - Comprehensive learning guide with architecture diagrams
• QUICKSTART.md - 5-minute setup guide
• PROJECT-SUMMARY.md - Learning outcomes and recommended path

🗄️ Sample Data

• FAQ.txt - 5,000+ word customer support knowledge base covering 30+ topics

🎓 Learning Philosophy

This project teaches:

• ✅ WHY each component exists (not just how to use it)
• ✅ HOW components work together (explicit SDK calls)
• ✅ WHEN to use different configurations (trade-offs explained)
• ✅ WHERE things can go wrong (debugging guidance)

🏗️ Architecture