Langflow-Customer-Support-Agent

An LLM-powered, multi-agent chatbot using Langflow, Streamlit, and OpenAI GPT-4o mini, with RAG-based retrieval via AstraDB, a vector database. It dynamically routes queries to specialized agents (FAQAgent, OrderLookupAgent) for accurate, context-aware responses. Includes vector search and file upload for knowledge expansion.

🌟 Features

• Multi-Agent Architecture: Uses specialized agents (FAQAgent, OrderLookupAgent, ManagerAgent) for handling different tasks.
• RAG-powered Retrieval: Integrates AstraDB to fetch relevant information based on user queries.
• Dynamic Routing: The ManagerAgent directs queries to the appropriate agent for accurate responses.
• File Upload for Knowledge Base Expansion: Users can upload files to enrich the FAQ database.
• Built with Streamlit: Provides an interactive and user-friendly frontend.

---

📸 App Interface

---

🎥 Demo Video

demo_video.mp4

---

🛠 Tech Stack

• Langflow – For building multi-agent workflows.
• Streamlit – Web UI framework.
• AstraDB – Vector storage for RAG-based retrieval.
• OpenAI GPT-4o mini – LLM powering the agents.

---

📌 How It Works

1️⃣ Query Handling

• User inputs a question.
• The ManagerAgent determines the best agent to handle the request.

2️⃣ RAG & FAQ Retrieval

• AstraDB stores FAQ entries as vectors.
• Queries are embedded using NV-Embed-QA and matched against stored FAQs.
• Results are parsed and injected into the agent's prompt for context-aware responses.

3️⃣ Order Lookup

• The OrderLookupAgent retrieves product and order details from AstraDB.
• Uses structured queries based on orderNumber and productId.

4️⃣ Multi-Agent Task Delegation

• The ManagerAgent intelligently routes queries:
  • General Queries → FAQAgent
  • Order/Product Queries → OrderLookupAgent
  • Aggregates responses and returns a structured answer.

5️⃣ File Upload for FAQ Expansion

• In the backend, the admin can upload documents to expand the FAQ knowledge base.
• Files are split into chunks and stored as vector embeddings in AstraDB.

---

🚀 Getting Started

Prerequisites

• Python 3.8+
• AstraDB Account (for vector storage)
• OpenAI API Key (for GPT-4o mini)
• Streamlit installed

Installation

git clone https://github.com/AkshatBhat/Langflow-Customer-Support-Agent.git

cd Langflow-Customer-Support-Agent

pip install -r requirements.txt

Configuration

Set the environment variable for AstraDB Application Token in a .env file (refer to the .env.example file).

Running the App

streamlit run app.py

---

📝 Example Queries

• General FAQ: "How can I track my order?"
• Order Lookup: "Where is my order with ID #1001?"
• Product Details: "What are the features of Product #101?"

---

📜 License

MIT License