Pipecat + Twilio Outbound IVR LLM Automation

I have a CSV that, row by row, tells the system which number to dial and which IVR paths to follow. For every call, an LLM running through Pipecat should read the instructions, place the outbound call with Twilio, navigate the IVR in real time, collect both any numeric responses it hears as well as any recorded voice messages, and then write everything into a database table I supply.

LLM: Llama 3.1 70B on Groq

You are free to pick the language (Python or Node.js preferred) and the database engine, but the final solution must be container-ready and clearly documented so I can redeploy it on my own infrastructure. Error handling, retries, and concise logging are important because the CSV may contain hundreds of calls.

Deliverables
• Clean, well-commented source code that consumes the CSV, orchestrates Twilio via Pipecat, invokes the LLM, and persists results in the database
• A schema or migration script for the results table, plus a brief README showing example records for both numeric and voice responses
• A one-command deployment script or Dockerfile, and step-by-step setup notes (API keys, webhooks, environment variables)

I will supply a sample CSV, Twilio credentials, and the target database connection string as soon as the project begins.