Claude Twilio Chatbot & Dashboard

I need a production-ready SMS support bot built around the Anthropic Claude API. Incoming messages will arrive through Twilio, pass through a small Redis cache to save prompt/response pairs, and then be answered by Claude. All of it should be containerised and deployed on Railway so I can scale without extra DevOps effort.

Alongside the bot, please create a Streamlit dashboard that lets me review any conversation and manually edit the bot’s replies before they are sent if I toggle “human in the loop.” For v1 the essential dashboard feature is Response editing; keep the codebase flexible so adding a user query log or performance metrics later is straightforward.

Core tasks
• Wire up Twilio SMS webhook → Redis cache → Claude completion → SMS reply
• Handle conversation context, rate limits, retries, and graceful fall-back messages
• Build a clean Streamlit app that lists live conversations and offers an inline editor for each pending response
• Protect the dashboard with simple auth (e.g., basic password or Twilio Verify)
• Dockerise everything and deploy on Railway; environment variables managed in Railway’s UI
• Provide a README covering setup, env keys, and how to extend features

Acceptance criteria
1. I can text the Twilio number and receive context-aware replies powered by Claude within 3 s on average.
2. Editing a pending response in Streamlit overrides the automated reply.
3. The codebase runs locally with `docker compose up` and deploys to my Railway account without modification.
4. Redis stores at least the last 20 messages of each thread and expires them after 24 h.

Please use Python (FastAPI or Flask is fine), `redis-py`, `twilio` SDK, and the official Anthropic client. Push the code to a private GitHub repo and share deploy instructions when done.