Customer Support AI Agent

A production-deployed AI support agent built with LangChain, FastAPI, and ChromaDB. Handles customer inquiries via RAG over a knowledge base, looks up orders, creates tickets, and escalates to a human with full conversation context when it can't help.

Live demo: https://csaia.onrender.com

First message may take ~30s if the service is waking from idle (Render free tier).

---

What It Does

• Answers policy and product questions from a semantic knowledge base (RAG)
• Looks up order status via Airtable
• Creates support tickets via HubSpot
• Sends reply emails via Resend
• Escalates to a human via Slack — with full conversation context attached
• Accepts new knowledge base content at runtime via /ingest (no restart needed)

---

Stack

Layer	Technology
Agent	LangChain ReAct (create_react_agent)
LLM	Groq — llama-3.3-70b-versatile
Embeddings	fastembed — BAAI/bge-small-en-v1.5 (ONNX)
Vector store	ChromaDB
API	FastAPI
Deployment	Render free web service

---

API Endpoints

Method	Path	Auth	Description
GET	/health	None	Service status + document count
POST	/chat	x-api-key	Send a message, get a response
POST	/ingest	x-api-key	Add content to the knowledge base

Chat request

{
  "message": "What is your refund policy?",
  "conversation_history": [],
  "user_email": "user@example.com"
}

Chat response

{
  "answer": "According to refund_policy.md, we offer full refunds within 14 days...",
  "sources": [{ "source": "refund_policy.md", "relevance_score": 0.82 }],
  "escalated": false,
  "session_id": "abc-123"
}

---

Running Locally

# 1. Clone and install
git clone https://github.com/k3d07/CSAIA.git
cd CSAIA
python -m venv venv && source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# 2. Set environment variables
cp .env.example .env
# Fill in GROQ_API_KEY and API_KEY at minimum

# 3. Start the server
uvicorn app.main:app --reload --port 8000

The server will auto-ingest ./knowledge_base/ on first startup.

---

Environment Variables

Variable	Required	Description
GROQ_API_KEY	Yes	Groq API key (free at console.groq.com)
API_KEY	Yes	Your chosen key for the x-api-key header
AIRTABLE_API_KEY	Optional	For order lookups
AIRTABLE_BASE_ID	Optional	Airtable base containing orders table
HUBSPOT_API_KEY	Optional	For ticket creation
RESEND_API_KEY	Optional	For sending reply emails
SLACK_WEBHOOK_URL	Optional	For escalation alerts

---

Project Structure

app/
  main.py          # FastAPI app, routes, lifespan
  agent.py         # LangChain ReAct agent
  rag.py           # ChromaDB + fastembed embeddings
  models.py        # Pydantic request/response models
  tools/
    knowledge_base.py   # search_faq — RAG retrieval
    orders.py           # get_order_status — Airtable
    hubspot.py          # create_support_ticket
    email.py            # send_reply_email — Resend
    escalation.py       # escalate_to_human — Slack
  static/
    index.html     # Portfolio chat UI
knowledge_base/    # Markdown docs ingested into ChromaDB
tests/             # RAG + agent test suite

---

Tests

pytest tests/ -v

8 tests covering RAG retrieval, multi-turn memory, escalation trigger, auth, and dynamic ingest.

---

Part of the 30-Day AI Automation Battle Plan — Project 4 of 30.