Polymind puts a single, coherent workspace over three competing model providers: OpenAI, Anthropic and Google. A request arrives, an intent-based routing layer picks the right model for the job, and the response streams back to the browser over WebSockets token by token. An arena mode streams two or three models' replies to the same prompt side by side.
Every model lives in one model registry that drives the API, the router, the biller and the client's model picker, with each vendor's quirks isolated behind a provider adapter. Adding a model is a single entry, and a request can never be billed at another model's rate.
Because every provider call costs real money, billing is not an afterthought. The platform runs a Decimal-exact credit wallet with Stripe and Razorpay checkout: every balance change is one atomic conditional SQL statement, and payment webhooks settle idempotently under row locks.
Beyond chat, heavier work is pushed off the request path entirely. Celery workers backed by Redis handle image generation, text-to-speech and AI avatar video creation, with generated media stored on Cloudflare R2 and credits refunded automatically when a job fails. The stack ships through Docker and GitHub Actions with health-gated deploys.