how to build online point to point video chat web application, what tech stack any library? especially in python

I'm trying to build a web application like chatroulette, but with text chatting and user identification and user authorization functionality. Is there any suggestion on what tech stack I should use? Would be better if it's python. The most baffling part for me so far is I have no idea how to make two user video chat with each other directly without server transferring the video stream, how to make them be able to find each other? Especially when they are under sub-net. My friend mentioned use SIP to make them be able to talk to each other, any other options? Any library or framework suggested? Thanks...