SafeRL-Lab/AgenticPay ? reverse-engineered prompt

Build me a Python project called AgenticPay that simulates buyer and seller negotiations with LLM powered agents.

I want it to support both simple text only price bargaining and richer contract negotiations where the buyer and seller can agree on things like price, delivery time, return policy, packaging, and other terms. It should also handle multimodal cases, like product images or other visual context, and work across a few commerce scenarios like shopping, food delivery, ride hailing, and apartment rental.

Please make the code easy to run with example scripts, a clean environment setup, and a simple way to register and create new negotiation scenarios. I’d like basic buyer and seller agents, conversation memory, scoring for the final deal, and a Gymnasium style API so it feels easy to plug into experiments.

If you need to check the latest docs for any model backend or dependency, feel free to look it up online.