ChinmayShringi/MicroFish-En — reverse-engineered prompt
Reverse engineered prompt
Build me an English version of MiroFish, a web app where someone can upload seed material like reports, news, drafts, or story text, then describe in normal language what they want to predict or explore.
The app should create a simulated digital world from that material, generate agents with personalities and memories, run a multi agent simulation, and return a clear prediction report. It should also let the user keep interacting afterward, like chatting with the report assistant or with agents inside the simulated world.
Please make it feel polished and usable, with a frontend for uploading materials, entering the prediction request, starting the simulation, watching progress, viewing results, and exploring the generated world. Use the existing Python backend and web frontend approach in this repo, and make it easy to run locally with the needed API keys from an env file. Docker support is fine too.
Keep the first version practical, so it can run smaller simulations without huge cost. Look up current docs online if needed.
Want more depth? Deep Reverse