I used a shared context between the text agent and the voice
Used RAG to pre-train the model on the Company's data
Used Wadi-Degla Development as an example