v0.1.4

Much cleaner and better coordinated dialog
faster (except I keep adding new llm calls)
Ability to specify add_generation_prompt in exl_server, hf_server
requirements.txt. (draft, left out openai and anthropic pkgs on pupose...)

V0.1.3

Improved dialog and formatting.

v0.1.0

Multi-character storylines.
local server (exl-server or tabbyAPI or ollama)

V0.1.0

MUCH improved character Action Architecture.
Still pretty lame past the 'scene' level.
Also skeleton Llama.cpp client and server. Not sure why this isn't working well (poor quality text).
Would appreciate help from anyone more knowledgeable about such things.

Haven't tested chiefOfStaff, it probably needs updates.
lost.py is best reference for defining your own plays.

v0.1.2

Characters now have multiple levels of motivation:
self.drives - set at initialization for scenario
self.states - dynamically derived from drives at micro-scene level
self.priorities - (actually current open tasks or goals) - dynamically derived from states
self.intentions - next action options, generated each turn from priorities and dialog interactions

This structure, plus environment sim and revisions to posting into character history of other character actions, seems to make interaction a bit more fluid. Still breaks down after a dozen or so rounds.