🧠 Routing Agent with LlamaIndex, Groq, and HuggingFace Embeddings

This project demonstrates a routing agent setup using LlamaIndex, Groq's LLaMA3-70B model, and HuggingFace Embeddings for answering queries from multiple domain-specific documents.

🚀 Features

• ✅ Uses Groq to access LLaMA 3-70B via LlamaIndex
• ✅ HuggingFace bge-small-en-v1.5 model for text embeddings
• ✅ Indexes and queries PDF documents (AeroFlow and EcoSprint)
• ✅ Routes questions automatically to the correct index using RouterQueryEngine
• ✅ Built in Google Colab with interactive Q&A support

📁 Project Structure

├── AeroFlow_Specification_Document.pdf
├── EcoSprint_Specification_Document.pdf
├── routing_agent_llamaindex.ipynb
├── README.md
└── LICENSE.txt

🧱 Requirements

• Python ≥ 3.8
• llama-index
• nest_asyncio
• huggingface_hub
• groq API Key
• Google Colab or local Jupyter environment

🔧 Setup Instructions

1. Upload PDF documents: Place AeroFlow_Specification_Document.pdf and EcoSprint_Specification_Document.pdf in the working directory or Colab file manager.

2. Set up API Keys:

from google.colab import userdata
Settings.llm = Groq(model="llama3-70b-8192", api_key=userdata.get('GROQ_API_KEY'))

3. Configure Embedding & LLM:

from llama_index.embeddings.huggingface import HuggingFaceEmbedding
Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

4. Create Document Indexes:

• Read, chunk, and index each document (AeroFlow and EcoSprint)

• Create individual query engines

5. Route Queries Using Router Agent:

• Automatically select the appropriate document source using RouterQueryEngine

• Example queries:

response = router_agent.query("What colors are available for AeroFlow?")
print(response)

response = router_agent.query("What colors are available for EcoSprint?")
print(response)

🤖 Technologies Used

• LlamaIndex

• Groq (LLaMA3-70B)

• HuggingFace Transformers

• Google Colab

• Python

📌 Sample Use Case

🔍 Ask: "What colors are available for AeroFlow?" ✅ The agent routes the query to the AeroFlow specification index and provides a relevant, grounded answer.

---

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.