the first submission of inception (embedding service):

[legaltextai]

• uses cpu for query embedding, and gpu - for everything else;
• i left port 8005 as it ran on my machine but we may need to change to a new one as a default;
• tried to mimic Doctor microservice;
• option to run in a docker;
• the embed_endpoint includes some text preprocessing but we need to decide on whether we want all pre-processing be done in the client's script that will be sending texts for embedding;
• tested on my ubuntu server, v100 gpu;
• 'inception' sub-folder within 'inception' is like a dream within a dream.

[mlissner]

Thanks Sergei. I'll put this on Alberto's backlog for when he's back.