Backend Transcription abstraction and Vexa provider

[jspada200]

• Build pydantic models to represent transcription segments

• Setup endpoints to query transcription segments
  GET /segments//

• Setup endpoints to dispatch a bot to join a call

• setup vexa provider to work with vexa

• Setup a storage provider which will allow us to access segments from vexa based on version ID.

Unknowns:

• How much data do we put into vexa? how do we assosiate each segment with a version? Can we Push into the vexa meta data? Do we store segment IDs on a version?

Acceptance criteria:

• Be able to interact with vexa via the abstraction/endpoints

Down the line:

• We want to emit websocket events to ensure the transcript is up to date. For now, we are going to "poll" the segments to keep the UI up to date.

Green sections rep the parts of the system: