Psyfighter2

* no " or * just treat as spoken words

* chromadb perist to db

* collect name should contain embedding so switching between chromadb configurations doesn't brick your scenes

* fix save-as long term memory transfer

* add chroma

* director agent refactor

* tweak director command, prompt reset, ux display

* tweak director message ux

* allow clearing of prompt log

* remove auto adding of quotes if neither quote or * are present

* command to reset long term memory for the scene

* improve summarization template as it would cause some llms to add extra details

* rebuilding history will now also rebuild long term memory

* direct scene template

* fix scene time reset

* dialogue template tweaks

* better dialog format fixing

* some dialogue template adjustments

* adjust default values of director agent

* keep track of scene saved/unsaved status and confirm loading a different scene if current scene is unsaved

* prompt fixes

* remove the collection on recommitting the seen to memory, as the embeddings may have changed

* change to the official python api for the openai client and make it async

* prompt tweaks

* world state prompt parsing fixes

* improve handling of json responses

* 0 seconds ago changed to moments ago

* move memory context closer to scene

* token counts for openai client

* narrator agent option: narrate passage of time

* gitignore

* remove memory id

* refactor world state with persistence to chromadb (wip)

* remove world state update instructions

* dont display blank emotion in world state

* openai gpt-4 turbo support

* conversation agent extra instructions

* track prompt response times

* Yi and UtopiaXL

* long term memory retrieval improvements during conversations

* narrate scene tweaks

* conversation ltm augment tweaks

* hide subconfig if parent config isnt enabled

* ai assisted memory recall during conversation default to off

* openai json_object coersion only on model that supports it

openai client emit prompt processing time

* 0.12.0

* remove prompt number from prompt debug list

* add prompt number back in but shift it to the upper row

* narrate time passage hard content limit restriction for now as gpt-4
would just write a whole chapter.

* relock

.gitignore

@@ -6,3 +6,8 @@
 *.internal*
 *_internal*
 talemate_env
+chroma
+scenes
+config.yaml
+!scenes/infinity-quest/assets
+!scenes/infinity-quest/infinity-quest.json

install.bat

@@ -7,10 +7,10 @@ REM activate the virtual environment
 call talemate_env\Scripts\activate
 
 REM install poetry
-python -m pip install poetry "rapidfuzz>=3" -U
+python -m pip install "poetry==1.7.1" "rapidfuzz>=3" -U
 
 REM use poetry to install dependencies
-poetry install
+python -m poetry install
 
 REM copy config.example.yaml to config.yaml only if config.yaml doesn't exist
 IF NOT EXIST config.yaml copy config.example.yaml config.yaml

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.masonry.api"
 
 [tool.poetry]
 name = "talemate"
-version = "0.11.1"
+version = "0.13.0"
 description = "AI-backed roleplay and narrative tools"
 authors = ["FinalWombat"]
 license = "GNU Affero General Public License v3.0"
@@ -17,7 +17,7 @@ black = "*"
 rope = "^0.22"
 isort = "^5.10"
 jinja2 = "^3.0"
-openai = "*"
+openai = ">=1"
 requests = "^2.26"
 colorama = ">=0.4.6"
 Pillow = "^9.5"
@@ -28,7 +28,6 @@ typing_extensions = "^4.5.0"
 uvicorn = "^0.23"
 blinker = "^1.6.2"
 pydantic = "<3"
-langchain = ">0.0.213"
 beautifulsoup4 = "^4.12.2"
 python-dotenv = "^1.0.0"
 websockets = "^11.0.3"
@@ -37,11 +36,12 @@ runpod = "==1.2.0"
 nest_asyncio = "^1.5.7"
 isodate = ">=0.6.1"
 thefuzz = ">=0.20.0"
+tiktoken = ">=0.5.1"
 
 # ChromaDB
-chromadb = ">=0.4,<1"
+chromadb = ">=0.4.17,<1"
 InstructorEmbedding = "^1.0.1"
-torch = ">=2.0.0, !=2.0.1"
+torch = ">=2.1.0"
 sentence-transformers="^2.2.2"
 
 [tool.poetry.dev-dependencies]

reinstall.bat

@@ -9,7 +9,7 @@ REM activate the virtual environment
 call talemate_env\Scripts\activate
 
 REM install poetry
-python -m pip install poetry "rapidfuzz>=3" -U
+python -m pip install "poetry==1.7.1" "rapidfuzz>=3" -U
 
 REM use poetry to install dependencies
 python -m poetry install

src/talemate/__init__.py

@@ -2,4 +2,4 @@ from .agents import Agent
 from .client import TextGeneratorWebuiClient
 from .tale_mate import *
 
-VERSION = "0.11.1"
+VERSION = "0.13.0"

src/talemate/agents/base.py

@@ -10,22 +10,29 @@ from blinker import signal
 import talemate.instance as instance
 import talemate.util as util
 from talemate.emit import emit
+from talemate.events import GameLoopStartEvent
+import talemate.emit.async_signals
 import dataclasses
 import pydantic
+import structlog
 
 __all__ = [
     "Agent",
     "set_processing",
 ]
 
+log = structlog.get_logger("talemate.agents.base")
+
 class AgentActionConfig(pydantic.BaseModel):
     type: str
     label: str
     description: str = ""
     value: Union[int, float, str, bool]
+    default_value: Union[int, float, str, bool] = None
     max: Union[int, float, None] = None
     min: Union[int, float, None] = None
     step: Union[int, float, None] = None
+    scope: str = "global"
 
 class AgentAction(pydantic.BaseModel):
     enabled: bool = True
@@ -160,7 +167,34 @@ class Agent(ABC):
                     config.value = kwargs.get("actions", {}).get(action_key, {}).get("config", {}).get(config_key, {}).get("value", config.value)
                 except AttributeError:
                     pass
-                
+    
+    async def on_game_loop_start(self, event:GameLoopStartEvent):
+        
+        """
+        Finds all ActionConfigs that have a scope of "scene" and resets them to their default values
+        """
+        
+        if not getattr(self, "actions", None):
+            return
+        
+        for _, action in self.actions.items():
+            if not action.config:
+                continue
+            
+            for _, config in action.config.items():
+                if config.scope == "scene":
+                    # if default_value is None, just use the `type` of the current 
+                    # value
+                    if config.default_value is None:
+                        default_value = type(config.value)()
+                    else:
+                        default_value = config.default_value
+                    
+                    log.debug("resetting config", config=config, default_value=default_value)
+                    config.value = default_value
+        
+        await self.emit_status()
+              
     async def emit_status(self, processing: bool = None):
         
         # should keep a count of processing requests, and when the
@@ -195,6 +229,8 @@ class Agent(ABC):
 
     def connect(self, scene):
         self.scene = scene
+        talemate.emit.async_signals.get("game_loop_start").connect(self.on_game_loop_start)
+
 
     def clean_result(self, result):
         if "#" in result:

src/talemate/agents/conversation.py

@@ -6,6 +6,7 @@ from datetime import datetime
 from typing import TYPE_CHECKING, Optional, Union
 
 import talemate.client as client
+import talemate.instance as instance
 import talemate.util as util
 import structlog
 from talemate.emit import emit
@@ -30,6 +31,7 @@ class ConversationAgentEmission(AgentEmission):
     generation: list[str]
     
 talemate.emit.async_signals.register(
+    "agent.conversation.before_generate",
     "agent.conversation.generated"
 )
 
@@ -79,6 +81,12 @@ class ConversationAgent(Agent):
                         min=32,
                         max=512,
                         step=32,
+                    ),#
+                    "instructions": AgentActionConfig(
+                        type="text",
+                        label="Instructions",
+                        value="1-3 sentences.",
+                        description="Extra instructions to give the AI for dialog generatrion.",
                     ),
                     "jiggle": AgentActionConfig(
                         type="number",
@@ -116,6 +124,19 @@ class ConversationAgent(Agent):
                     ),
                 }
             ),
+            "use_long_term_memory": AgentAction(
+                enabled = True,
+                label = "Long Term Memory",
+                description = "Will augment the conversation prompt with long term memory.",
+                config = {
+                    "ai_selected": AgentActionConfig(
+                        type="bool",
+                        label="AI Selected",
+                        description="If enabled, the AI will select the long term memory to use. (will increase how long it takes to generate a response)",
+                        value=False,
+                    ),
+                }
+            ),  
         }
 
     def connect(self, scene):
@@ -301,10 +322,7 @@ class ConversationAgent(Agent):
             insert_bot_token=10
         )
         
-        memory = await self.build_prompt_default_memory(
-            scene, long_term_memory_budget, 
-            scene_and_dialogue + [f"{character.name}: {character.description}" for character in scene.get_characters()]
-        )
+        memory = await self.build_prompt_default_memory(character)
 
         main_character = scene.main_character.character
 
@@ -326,6 +344,10 @@ class ConversationAgent(Agent):
             director_message = isinstance(scene_and_dialogue[-1], DirectorMessage)
         except IndexError:
             director_message = False
+            
+        extra_instructions = ""
+        if self.actions["generation_override"].enabled:
+            extra_instructions = self.actions["generation_override"].config["instructions"].value
 
         prompt = Prompt.get("conversation.dialogue", vars={
             "scene": scene,
@@ -339,12 +361,13 @@ class ConversationAgent(Agent):
             "talking_character": character,
             "partial_message": char_message,
             "director_message": director_message,
+            "extra_instructions": extra_instructions,
         })
         
         return str(prompt)
     
     async def build_prompt_default_memory(
-        self, scene: Scene, budget: int, existing_context: list
+        self, character: Character
     ):
         """
         Builds long term memory for the conversation prompt
@@ -357,29 +380,35 @@ class ConversationAgent(Agent):
         Also it will only add information that is not already in the existing context.
         """
 
-        memory = scene.get_helper("memory").agent
-
-        if not memory:
+        if not self.actions["use_long_term_memory"].enabled:
             return []
 
+
         if self.current_memory_context:
             return self.current_memory_context
 
-        self.current_memory_context = []
+        self.current_memory_context = ""
+        
 
-        # feed the last 3 history message into multi_query
-        history_length = len(scene.history)
-        i = history_length - 1
-        while i >= 0 and i >= len(scene.history) - 3:
-            self.current_memory_context += await memory.multi_query(
-                [scene.history[i]],
-                filter=lambda x: x
-                not in self.current_memory_context + existing_context,
+        if self.actions["use_long_term_memory"].config["ai_selected"].value:
+            history = self.scene.context_history(min_dialogue=3, max_dialogue=15, keep_director=False, sections=False, add_archieved_history=False)
+            text = "\n".join(history)
+            world_state = instance.get_agent("world_state")
+            log.debug("conversation_agent.build_prompt_default_memory", direct=False)
+            self.current_memory_context = await world_state.analyze_text_and_extract_context(
+                text, f"continue the conversation as {character.name}"
             )
-            i -= 1
 
+        else:
+            history = self.scene.context_history(min_dialogue=3, max_dialogue=3, keep_director=False, sections=False, add_archieved_history=False)
+            log.debug("conversation_agent.build_prompt_default_memory", history=history, direct=True)
+            memory = instance.get_agent("memory")
+            
+            context = await memory.multi_query(history, max_tokens=500, iterate=5)
+             
+            self.current_memory_context = "\n".join(context)
+        
         return self.current_memory_context
-    
 
     async def build_prompt(self, character, char_message: str = ""):
         fn = self.build_prompt_default
@@ -423,6 +452,9 @@ class ConversationAgent(Agent):
 
         character = actor.character
         
+        emission = ConversationAgentEmission(agent=self, generation="", actor=actor, character=character)
+        await talemate.emit.async_signals.get("agent.conversation.before_generate").send(emission)
+        
         self.set_generation_overrides()
 
         result = await self.client.send_prompt(await self.build_prompt(character))
@@ -473,9 +505,6 @@ class ConversationAgent(Agent):
         # Remove "{character.name}:" - all occurences
         total_result = total_result.replace(f"{character.name}:", "")
 
-        if total_result.count("*") % 2 == 1:
-            total_result += "*"
-
         # Check if total_result starts with character name, if not, prepend it
         if not total_result.startswith(character.name):
             total_result = f"{character.name}: {total_result}"

src/talemate/agents/director.py

@@ -8,10 +8,12 @@ from typing import TYPE_CHECKING, Callable, List, Optional, Union
 
 import talemate.util as util
 from talemate.emit import wait_for_input, emit
+import talemate.emit.async_signals
 from talemate.prompts import Prompt
 from talemate.scene_message import NarratorMessage, DirectorMessage
 from talemate.automated_action import AutomatedAction
 import talemate.automated_action as automated_action
+from talemate.agents.conversation import ConversationAgentEmission
 from .registry import register
 from .base import set_processing, AgentAction, AgentActionConfig, Agent
 
@@ -26,11 +28,13 @@ class DirectorAgent(Agent):
     verbose_name = "Director"
     
     def __init__(self, client, **kwargs):
-        self.is_enabled = True
+        self.is_enabled = False
         self.client = client
+        self.next_direct = 0
         self.actions = {
-            "direct": AgentAction(enabled=False, label="Direct", description="Will attempt to direct the scene. Runs automatically after AI dialogue (n turns).", config={
-                "turns": AgentActionConfig(type="number", label="Turns", description="Number of turns to wait before directing the sceen", value=10, min=1, max=100, step=1)
+            "direct": AgentAction(enabled=True, label="Direct", description="Will attempt to direct the scene. Runs automatically after AI dialogue (n turns).", config={
+                "turns": AgentActionConfig(type="number", label="Turns", description="Number of turns to wait before directing the sceen", value=5, min=1, max=100, step=1),
+                "prompt": AgentActionConfig(type="text", label="Instructions", description="Instructions to the director", value="", scope="scene")
             }),
         }
         
@@ -46,316 +50,57 @@ class DirectorAgent(Agent):
     def experimental(self):
         return True
     
-    def get_base_prompt(self, character: Character, budget:int):
-        return [character.description, character.base_attributes.get("scenario_context", "")] + self.scene.context_history(budget=budget, keep_director=False)
-
-    
-    async def decide_action(self, character: Character, goal_override:str=None):
+    def connect(self, scene):
+        super().connect(scene)
+        talemate.emit.async_signals.get("agent.conversation.before_generate").connect(self.on_conversation_before_generate)
         
-        """
-        Pick an action to perform to move the story towards the current story goal
-        """
+    async def on_conversation_before_generate(self, event:ConversationAgentEmission):
+        log.info("on_conversation_before_generate", director_enabled=self.enabled)
+        if not self.enabled:
+            return
         
-        current_goal = goal_override or await self.select_goal(self.scene)
-        current_goal = f"Current story goal: {current_goal}" if current_goal else current_goal
+        await self.direct_scene(event.character)
         
-        response, action_eval, prompt = await self.decide_action_analyze(character, current_goal)
-        # action_eval will hold {'narrate': N, 'direct': N, 'watch': N, ...}
-        # where N is a number, action with the highest number wins, default action is watch
-        # if there is no clear winner
+    async def direct_scene(self, character: Character):
         
-        watch_action = action_eval.get("watch", 0)
-        action = max(action_eval, key=action_eval.get)
+        if not self.actions["direct"].enabled:
+            log.info("direct_scene", skip=True, enabled=self.actions["direct"].enabled)
+            return
         
-        if action_eval[action] <= watch_action:
-            action = "watch"
+        prompt = self.actions["direct"].config["prompt"].value
         
-        log.info("decide_action", action=action, action_eval=action_eval)
+        if not prompt:
+            log.info("direct_scene", skip=True, prompt=prompt)
+            return
         
-        return response, current_goal, action
+        if self.next_direct % self.actions["direct"].config["turns"].value != 0 or self.next_direct == 0:
+            
+            log.info("direct_scene", skip=True, next_direct=self.next_direct)
+            self.next_direct += 1
+            return
         
+        self.next_direct = 0
         
-    async def decide_action_analyze(self, character: Character, goal:str):
-
-        prompt = Prompt.get("director.decide-action-analyze", vars={
+        await self.direct_character(character, prompt)
+        
+    @set_processing
+    async def direct_character(self, character: Character, prompt:str):
+        
+        response = await Prompt.request("director.direct-scene", self.client, "director", vars={
             "max_tokens": self.client.max_token_length,
             "scene": self.scene,
-            "current_goal": goal,
+            "prompt": prompt,
             "character": character,
         })
-
-        response, evaluation = await prompt.send(self.client, kind="director")
         
-        log.info("question_direction", response=response)
-        return response, evaluation, prompt
-        
-    @set_processing
-    async def direct(self, character: Character, goal_override:str=None):
-        
-        analysis, current_goal, action = await self.decide_action(character, goal_override=goal_override)
-        
-        if action == "watch":
-            return None
-        
-        if action == "direct":
-            return await self.direct_character_with_self_reflection(character, analysis, goal_override=current_goal)
-        
-        if action.startswith("narrate"):
-            
-            narration_type = action.split(":")[1] 
-            
-            direct_narrative = await self.direct_narrative(analysis, narration_type=narration_type, goal=current_goal)
-            if direct_narrative:
-                narrator = self.scene.get_helper("narrator").agent
-                narrator_response = await narrator.progress_story(direct_narrative)
-                if not narrator_response:
-                    return None
-                narrator_message = NarratorMessage(narrator_response, source="progress_story")
-                self.scene.push_history(narrator_message)
-                emit("narrator", narrator_message)
-                return True
-            
-            
-    @set_processing
-    async def direct_narrative(self, analysis:str, narration_type:str="progress", goal:str=None):
-        
-        if goal is None:
-            goal = await self.select_goal(self.scene)
-        
-        prompt = Prompt.get("director.direct-narrative", vars={
-            "max_tokens": self.client.max_token_length,
-            "scene": self.scene,
-            "narration_type": narration_type,
-            "analysis": analysis,
-            "current_goal": goal,
-        })
-        
-        response = await prompt.send(self.client, kind="director")
-        response = response.strip().split("\n")[0].strip()
-                
-        if not response:
-            return None
-            
-        return response
-    
-    @set_processing
-    async def direct_character_with_self_reflection(self, character: Character, analysis:str, goal_override:str=None):
-        
-        max_retries = 3
-        num_retries = 0
-        keep_direction = False
-        response = None
-        self_reflection = None
-        
-        while num_retries < max_retries:
-            
-            response, direction_prompt = await self.direct_character(
-                character, 
-                analysis, 
-                goal_override=goal_override, 
-                previous_direction=response,
-                previous_direction_feedback=self_reflection
-            )
-            
-            keep_direction, self_reflection = await self.direct_character_self_reflect(
-                response, character, goal_override, direction_prompt
-            )
-            
-            if keep_direction:
-                break
-            
-            num_retries += 1
-            
-        log.info("direct_character_with_self_reflection", response=response, keep_direction=keep_direction)
-            
-        if not keep_direction:
-            return None
-        
-        #character_agreement = f" *{character.name} agrees with the director and progresses the story accordingly*"
-        #
-        #if "accordingly" not in response:
-        #   response += character_agreement
-        #
-        
-        #response = await self.transform_character_direction_to_inner_monologue(character, response)
-        
-        return response
-        
-    @set_processing
-    async def transform_character_direction_to_inner_monologue(self, character:Character, direction:str):
-        
-        inner_monologue = await Prompt.request(
-            "conversation.direction-to-inner-monologue",
-            self.client,
-            "conversation_long",
-            vars={
-                "max_tokens": self.client.max_token_length,
-                "scene": self.scene,
-                "character": character,
-                "director_instructions": direction,
-            }
-        )
-        
-        return inner_monologue
-        
-        
-    @set_processing
-    async def direct_character(
-        self, 
-        character: Character, 
-        analysis:str, 
-        goal_override:str=None,
-        previous_direction:str=None,
-        previous_direction_feedback:str=None,
-    ):
-        """
-        Direct the scene
-        """
-        
-        if goal_override:
-            current_goal = goal_override
-        else:
-            current_goal = await self.select_goal(self.scene)
-            
-        if current_goal and not current_goal.startswith("Current story goal: "):
-            current_goal = f"Current story goal: {current_goal}"
-        
-        prompt = Prompt.get("director.direct-character", vars={
-            "max_tokens": self.client.max_token_length,
-            "scene": self.scene,
-            "character": character,
-            "current_goal": current_goal,
-            "previous_direction": previous_direction,
-            "previous_direction_feedback": previous_direction_feedback,
-            "analysis": analysis,
-        })
-        
-        response = await prompt.send(self.client, kind="director")
         response = response.strip().split("\n")[0].strip()
         
-        log.info(
-            "direct_character", 
-            direction=response, 
-            previous_direction=previous_direction, 
-            previous_direction_feedback=previous_direction_feedback
-        )
-                
-        if not response:
-            return None
-
-        if not response.startswith(prompt.prepared_response):
-            response = prompt.prepared_response + response
-
-        return response, "\n".join(prompt.as_list[:-1])
-    
-    
-    
-    @set_processing
-    async def direct_character_self_reflect(self, direction:str, character: Character, goal:str, direction_prompt:Prompt) -> (bool, str):
+        response += f" (current story goal: {prompt})"
         
-        change_matches = ["change", "retry", "alter", "reconsider"]
+        log.info("direct_scene", response=response)
         
-        prompt = Prompt.get("director.direct-character-self-reflect", vars={
-            "direction_prompt": str(direction_prompt),
-            "direction": direction,
-            "analysis": await self.direct_character_analyze(direction, character, goal, direction_prompt),
-            "character": character,
-            "scene": self.scene,
-            "max_tokens": self.client.max_token_length,
-        })
         
-        response = await prompt.send(self.client, kind="director")
+        message = DirectorMessage(response, source=character.name)
+        emit("director", message, character=character)
         
-        parse_choice = response[len(prompt.prepared_response):].lower().split(" ")[0]
-        
-        keep = not parse_choice in change_matches
-        
-        log.info("direct_character_self_reflect", keep=keep, response=response, parsed=parse_choice)
-        
-        return keep, response
-    
-    
-    @set_processing
-    async def direct_character_analyze(self, direction:str, character: Character, goal:str, direction_prompt:Prompt):
-        
-        prompt = Prompt.get("director.direct-character-analyze", vars={
-            "direction_prompt": str(direction_prompt),
-            "direction": direction,
-            "scene": self.scene,
-            "max_tokens": self.client.max_token_length,
-            "character": character,
-        })
-        
-        analysis = await prompt.send(self.client, kind="director")
-        
-        log.info("direct_character_analyze", analysis=analysis)
-        
-        return analysis
-
-    async def select_goal(self, scene: Scene):
-        
-        if not scene.goals:
-            return ""
-        
-        if isinstance(self.scene.goal, int):
-            # fixes legacy goal format
-            self.scene.goal = self.scene.goals[self.scene.goal]
-
-        while True:
-
-            # get current goal position in goals
-        
-            current_goal = scene.goal
-            current_goal_positon = None
-            if current_goal:
-                try:
-                    current_goal_positon = self.scene.goals.index(current_goal)
-                except ValueError:
-                    pass
-            elif self.scene.goals:
-                current_goal = self.scene.goals[0]
-                current_goal_positon = 0
-            else:
-                return ""
-                
-                
-            # if current goal is set but not found, its a custom goal override
-            
-            custom_goal = (current_goal and current_goal_positon is None)
-        
-            log.info("select_goal", current_goal=current_goal, current_goal_positon=current_goal_positon, custom_goal=custom_goal)
-
-            if current_goal:
-                current_goal_met = await self.goal_analyze(current_goal)
-                
-                log.info("select_goal", current_goal_met=current_goal_met)
-                if current_goal_met is not True:
-                    return current_goal + f"\nThe goal has {current_goal_met})"
-                try:
-                    self.scene.goal = self.scene.goals[current_goal_positon + 1]
-                    continue
-                except IndexError:
-                    return ""
-            
-            else:
-                return ""
-    
-    @set_processing
-    async def goal_analyze(self, goal:str):
-        
-        prompt = Prompt.get("director.goal-analyze", vars={
-            "scene": self.scene,
-            "max_tokens": self.client.max_token_length,
-            "current_goal": goal,
-        })
-        
-        response = await prompt.send(self.client, kind="director")
-    
-        log.info("goal_analyze", response=response)
-        
-        if "not satisfied" in response.lower().strip() or "not been satisfied" in response.lower().strip():
-            goal_met = response
-        else:
-            goal_met = True
-        
-        return goal_met
+        self.scene.push_history(message)

src/talemate/agents/memory.py

@@ -6,8 +6,10 @@ from typing import TYPE_CHECKING, Callable, List, Optional, Union
 from chromadb.config import Settings
 import talemate.events as events
 import talemate.util as util
+from talemate.context import scene_is_loading
 from talemate.config import load_config
 import structlog
+import shutil
 
 try:
     import chromadb
@@ -34,6 +36,18 @@ class MemoryAgent(Agent):
     agent_type = "memory"
     verbose_name = "Long-term memory"
 
+    @property
+    def readonly(self):
+        
+        if scene_is_loading.get() and not getattr(self.scene, "_memory_never_persisted", False):
+            return True
+        
+        return False
+
+    @property
+    def db_name(self):
+        raise NotImplementedError()
+
     @classmethod
     def config_options(cls, agent=None):
         return {}
@@ -50,16 +64,24 @@ class MemoryAgent(Agent):
     def close_db(self):
         raise NotImplementedError()
 
+    async def count(self):
+        raise NotImplementedError()
+
     async def add(self, text, character=None, uid=None, ts:str=None, **kwargs):
         if not text:
             return
-
+        if self.readonly:
+            log.debug("memory agent", status="readonly")
+            return
         await self._add(text, character=character, uid=uid, ts=ts, **kwargs)
 
     async def _add(self, text, character=None, ts:str=None, **kwargs):
         raise NotImplementedError()
 
     async def add_many(self, objects: list[dict]):
+        if self.readonly:
+            log.debug("memory agent", status="readonly")
+            return
         await self._add_many(objects)
     
     async def _add_many(self, objects: list[dict]):
@@ -131,13 +153,13 @@ class MemoryAgent(Agent):
                 break
         return memory_context
 
-    async def query(self, query:str, max_tokens:int=1000, filter:Callable=lambda x:True):
+    async def query(self, query:str, max_tokens:int=1000, filter:Callable=lambda x:True, **where):
         """
         Get the character memory context for a given character
         """
 
         try:
-            return (await self.multi_query([query], max_tokens=max_tokens, filter=filter))[0]
+            return (await self.multi_query([query], max_tokens=max_tokens, filter=filter, **where))[0]
         except IndexError:
             return None
 
@@ -158,7 +180,7 @@ class MemoryAgent(Agent):
         memory_context = []
         for query in queries:
             i = 0
-            for memory in await self.get(formatter(query), **where):
+            for memory in await self.get(formatter(query), limit=iterate, **where):
                 if memory in memory_context:
                     continue
 
@@ -238,26 +260,52 @@ class ChromaDBMemoryAgent(MemoryAgent):
     @property
     def USE_INSTRUCTOR(self):
         return self.embeddings == "instructor"
+    
+    @property
+    def db_name(self):
+        return getattr(self, "collection_name", "<unnamed>")
+
+    def make_collection_name(self, scene):
+        
+        if self.USE_OPENAI:
+            suffix = "-openai"
+        elif self.USE_INSTRUCTOR:
+            suffix = "-instructor"
+            model = self.config.get("chromadb").get("instructor_model", "hkunlp/instructor-xl")
+            if "xl" in model:
+                suffix += "-xl"
+            elif "large" in model:
+                suffix += "-large"
+        else:
+            suffix = ""
+        
+        return f"{scene.memory_id}-tm{suffix}"
+
+    async def count(self):
+        await asyncio.sleep(0)
+        return self.db.count()
 
     async def set_db(self):
         await self.emit_status(processing=True)
 
-        if getattr(self, "db", None):
-            try:
-                self.db.delete(where={"source": "talemate"})
-            except ValueError:
-                pass
-            await self.emit_status(processing=False)
-            
-            return
 
-        log.info("chromadb agent", status="setting up db")
-
-        self.db_client = chromadb.Client(Settings(anonymized_telemetry=False))
+        if not getattr(self, "db_client", None):
+            log.info("chromadb agent", status="setting up db client to persistent db")
+            self.db_client = chromadb.PersistentClient(
+                settings=Settings(anonymized_telemetry=False)
+            )
 
         openai_key = self.config.get("openai").get("api_key") or os.environ.get("OPENAI_API_KEY")
-
-        if openai_key and self.USE_OPENAI:
+        
+        self.collection_name = collection_name = self.make_collection_name(self.scene)
+        
+        log.info("chromadb agent", status="setting up db", collection_name=collection_name)
+        
+        if self.USE_OPENAI:
+            
+            if not openai_key:
+                raise ValueError("You must provide an the openai ai key in the config if you want to use it for chromadb embeddings")
+            
             log.info(
                 "crhomadb", status="using openai", openai_key=openai_key[:5] + "..."
             )
@@ -266,7 +314,7 @@ class ChromaDBMemoryAgent(MemoryAgent):
                 model_name="text-embedding-ada-002",
             )
             self.db = self.db_client.get_or_create_collection(
-                "talemate-story", embedding_function=openai_ef
+                collection_name, embedding_function=openai_ef
             )
         elif self.USE_INSTRUCTOR:
             
@@ -280,25 +328,60 @@ class ChromaDBMemoryAgent(MemoryAgent):
                 model_name=instructor_model, device=instructor_device
             )
             
+            log.info("chromadb", status="embedding function ready")
+            
             self.db = self.db_client.get_or_create_collection(
-                "talemate-story", embedding_function=ef
+                collection_name, embedding_function=ef
             )
+            
+            log.info("chromadb", status="instructor db ready")
         else:
             log.info("chromadb", status="using default embeddings")
-            self.db = self.db_client.get_or_create_collection("talemate-story")
+            self.db = self.db_client.get_or_create_collection(collection_name)
+        
+        self.scene._memory_never_persisted = self.db.count() == 0
 
         await self.emit_status(processing=False)
         log.info("chromadb agent", status="db ready")
 
-    def close_db(self):
+    def clear_db(self):
         if not self.db:
             return
         
+        log.info("chromadb agent", status="clearing db", collection_name=self.collection_name)
+        
+        self.db.delete(where={"source": "talemate"})
+        
+    def drop_db(self):
+        if not self.db:
+            return
+        
+        log.info("chromadb agent", status="dropping db", collection_name=self.collection_name)
+        
         try:
-            self.db.delete(where={"source": "talemate"})
-        except ValueError:
-            pass
+            self.db_client.delete_collection(self.collection_name)
+        except ValueError as exc:
+            if "Collection not found" not in str(exc):
+                raise
 
+    def close_db(self, scene):
+        if not self.db:
+            return
+        
+        log.info("chromadb agent", status="closing db", collection_name=self.collection_name)
+        
+        if not scene.saved:
+            # scene was never saved so we can discard the memory
+            collection_name = self.make_collection_name(scene)
+            log.info("chromadb agent", status="discarding memory", collection_name=collection_name)
+            try:
+                self.db_client.delete_collection(collection_name)
+            except ValueError as exc:
+                if "Collection not found" not in str(exc):
+                    raise
+        
+        self.db = None
+        
     async def _add(self, text, character=None, uid=None, ts:str=None, **kwargs):
         metadatas = []
         ids = []
@@ -329,7 +412,7 @@ class ChromaDBMemoryAgent(MemoryAgent):
         log.debug("chromadb agent add", text=text, meta=meta, id=id)
 
         self.db.upsert(documents=[text], metadatas=metadatas, ids=ids)
-
+        
         await self.emit_status(processing=False)
 
     async def _add_many(self, objects: list[dict]):
@@ -354,7 +437,7 @@ class ChromaDBMemoryAgent(MemoryAgent):
 
         await self.emit_status(processing=False)
 
-    async def _get(self, text, character=None, **kwargs):
+    async def _get(self, text, character=None, limit:int=15, **kwargs):
         await self.emit_status(processing=True)
 
         where = {}
@@ -378,7 +461,11 @@ class ChromaDBMemoryAgent(MemoryAgent):
 
         #log.debug("crhomadb agent get", text=text, where=where)
 
-        _results = self.db.query(query_texts=[text], where=where)
+        _results = self.db.query(query_texts=[text], where=where, n_results=limit)
+        
+        #import json
+        #print(json.dumps(_results["ids"], indent=2))
+        #print(json.dumps(_results["distances"], indent=2))
         
         results = []
 
@@ -405,9 +492,9 @@ class ChromaDBMemoryAgent(MemoryAgent):
 
             # log.debug("crhomadb agent get", result=results[-1], distance=distance)
 
-            if len(results) > 10:
+            if len(results) > limit:
                 break
 
         await self.emit_status(processing=False)
-
+        
         return results

src/talemate/agents/narrator.py

@@ -1,19 +1,32 @@
 from __future__ import annotations
 
-import asyncio
-import re
 from typing import TYPE_CHECKING, Callable, List, Optional, Union
-
+import structlog
+import random
 import talemate.util as util
-from talemate.emit import wait_for_input
+from talemate.emit import emit
+import talemate.emit.async_signals
 from talemate.prompts import Prompt
-from talemate.agents.base import set_processing, Agent
+from talemate.agents.base import set_processing, Agent, AgentAction, AgentActionConfig
+from talemate.agents.world_state import TimePassageEmission
+from talemate.scene_message import NarratorMessage
+from talemate.events import GameLoopActorIterEvent
 import talemate.client as client
 
 from .registry import register
 
+if TYPE_CHECKING:
+    from talemate.tale_mate import Actor, Player, Character
+
+log = structlog.get_logger("talemate.agents.narrator")
+
 @register()
 class NarratorAgent(Agent):
+    
+    """
+    Handles narration of the story
+    """
+    
     agent_type = "narrator"
     verbose_name = "Narrator"
     
@@ -24,20 +37,110 @@ class NarratorAgent(Agent):
     ):
         self.client = client
         
+        # agent actions
+        
+        self.actions = {
+            "narrate_time_passage": AgentAction(enabled=True, label="Narrate Time Passage", description="Whenever you indicate passage of time, narrate right after"),
+            "narrate_dialogue": AgentAction(
+                enabled=True, 
+                label="Narrate Dialogue", 
+                description="Narrator will get a chance to narrate after every line of dialogue",
+                config = {
+                    "ai_dialog": AgentActionConfig(
+                        type="number",
+                        label="AI Dialogue", 
+                        description="Chance to narrate after every line of dialogue, 1 = always, 0 = never",
+                        value=0.3,
+                        min=0.0,
+                        max=1.0,
+                        step=0.1,
+                    ),
+                    "player_dialog": AgentActionConfig(
+                        type="number",
+                        label="Player Dialogue", 
+                        description="Chance to narrate after every line of dialogue, 1 = always, 0 = never",
+                        value=0.3,
+                        min=0.0,
+                        max=1.0,
+                        step=0.1,
+                    ),
+                }
+            ),
+        }
+        
     def clean_result(self, result):
         
+        """
+        Cleans the result of a narration
+        """
+        
         result = result.strip().strip(":").strip()
         
         if "#" in result:
             result = result.split("#")[0]
+            
+        character_names = [c.name for c in self.scene.get_characters()]
+    
         
         cleaned = []
         for line in result.split("\n"):
-            if ":" in line.strip():
-                break
+            for character_name in character_names:
+                if line.startswith(f"{character_name}:"):
+                    break
             cleaned.append(line)
 
-        return "\n".join(cleaned)
+        result = "\n".join(cleaned)
+        #result = util.strip_partial_sentences(result)
+        return result
+
+    def connect(self, scene):
+        
+        """
+        Connect to signals
+        """
+        
+        super().connect(scene)
+        talemate.emit.async_signals.get("agent.world_state.time").connect(self.on_time_passage)
+        talemate.emit.async_signals.get("game_loop_actor_iter").connect(self.on_dialog)
+
+    async def on_time_passage(self, event:TimePassageEmission):
+        
+        """
+        Handles time passage narration, if enabled
+        """
+        
+        if not self.actions["narrate_time_passage"].enabled:
+            return
+        
+        response = await self.narrate_time_passage(event.duration, event.narrative)
+        narrator_message = NarratorMessage(response, source=f"narrate_time_passage:{event.duration};{event.narrative}")
+        emit("narrator", narrator_message)
+        self.scene.push_history(narrator_message)
+        
+    async def on_dialog(self, event:GameLoopActorIterEvent):
+        
+        """
+        Handles dialogue narration, if enabled
+        """
+        
+        if not self.actions["narrate_dialogue"].enabled:
+            return
+        
+        narrate_on_ai_chance = random.random() < self.actions["narrate_dialogue"].config["ai_dialog"].value
+        narrate_on_player_chance = random.random() < self.actions["narrate_dialogue"].config["player_dialog"].value
+        
+        log.debug("narrate on dialog", narrate_on_ai_chance=narrate_on_ai_chance, narrate_on_player_chance=narrate_on_player_chance)
+        
+        if event.actor.character.is_player and not narrate_on_player_chance:
+            return
+        
+        if not event.actor.character.is_player and not narrate_on_ai_chance:
+            return
+        
+        response = await self.narrate_after_dialogue(event.actor.character)
+        narrator_message = NarratorMessage(response, source=f"narrate_dialogue:{event.actor.character.name}")
+        emit("narrator", narrator_message)
+        self.scene.push_history(narrator_message)
 
     @set_processing
     async def narrate_scene(self):
@@ -55,6 +158,9 @@ class NarratorAgent(Agent):
             }
         )
         
+        response = response.strip("*")
+        response = util.strip_partial_sentences(response)
+        
         response = f"*{response.strip('*')}*"
 
         return response
@@ -131,8 +237,9 @@ class NarratorAgent(Agent):
                 "as_narrative": as_narrative,
             }
         )
-
+        log.info("narrate_query", response=response)
         response = self.clean_result(response.strip())
+        log.info("narrate_query (after clean)", response=response)
         if as_narrative:
             response = f"*{response}*"
         
@@ -216,4 +323,55 @@ class NarratorAgent(Agent):
         answers = [a for a in answers.split("\n") if a.strip()]
         
         # return questions and answers
-        return list(zip(questions, answers))
+        return list(zip(questions, answers))
+    
+    @set_processing
+    async def narrate_time_passage(self, duration:str, narrative:str=None):
+        """
+        Narrate a specific character
+        """
+
+        response = await Prompt.request(
+            "narrator.narrate-time-passage",
+            self.client,
+            "narrate",
+            vars = {
+                "scene": self.scene,
+                "max_tokens": self.client.max_token_length,
+                "duration": duration,
+                "narrative": narrative,
+            }
+        )
+        
+        log.info("narrate_time_passage", response=response)
+
+        response = self.clean_result(response.strip())
+        response = f"*{response}*"
+
+        return response
+    
+    
+    @set_processing
+    async def narrate_after_dialogue(self, character:Character):
+        """
+        Narrate after a line of dialogue
+        """
+
+        response = await Prompt.request(
+            "narrator.narrate-after-dialogue",
+            self.client,
+            "narrate",
+            vars = {
+                "scene": self.scene,
+                "max_tokens": self.client.max_token_length,
+                "character": character,
+                "last_line": str(self.scene.history[-1])
+            }
+        )
+        
+        log.info("narrate_after_dialogue", response=response)
+
+        response = self.clean_result(response.strip().strip("*"))
+        response = f"*{response}*"
+
+        return response

src/talemate/agents/summarize.py

@@ -53,7 +53,7 @@ class SummarizeAgent(Agent):
         return result
 
     @set_processing
-    async def build_archive(self, scene):
+    async def build_archive(self, scene, token_threshold:int=1500):
         end = None
 
         if not scene.archived_history:
@@ -63,12 +63,13 @@ class SummarizeAgent(Agent):
             recent_entry = scene.archived_history[-1]
             start = recent_entry.get("end", 0) + 1
 
-        token_threshold = 1500
         tokens = 0
         dialogue_entries = []
         ts = "PT0S"
         time_passage_termination = False
         
+        log.debug("build_archive", start=start, recent_entry=recent_entry)
+        
         if recent_entry:
             ts = recent_entry.get("ts", ts)
         

src/talemate/agents/world_state.py

@@ -1,29 +1,43 @@
 from __future__ import annotations
+import dataclasses
 
-import asyncio
-import traceback
 from typing import TYPE_CHECKING, Callable, List, Optional, Union
 
-import talemate.data_objects as data_objects
 import talemate.emit.async_signals
 import talemate.util as util
 from talemate.prompts import Prompt
 from talemate.scene_message import DirectorMessage, TimePassageMessage
+from talemate.emit import emit
+from talemate.events import GameLoopEvent
 
-from .base import Agent, set_processing, AgentAction, AgentActionConfig
+from .base import Agent, set_processing, AgentAction, AgentActionConfig, AgentEmission
 from .registry import register
 
 import structlog
-
+import isodate
 import time
-import re
 
-if TYPE_CHECKING:
-    from talemate.agents.conversation import ConversationAgentEmission
-    
 
 log = structlog.get_logger("talemate.agents.world_state")
 
+talemate.emit.async_signals.register("agent.world_state.time")
+
+@dataclasses.dataclass
+class WorldStateAgentEmission(AgentEmission):
+    """
+    Emission class for world state agent
+    """
+    pass
+
+@dataclasses.dataclass
+class TimePassageEmission(WorldStateAgentEmission):
+    """
+    Emission class for time passage
+    """
+    duration: str
+    narrative: str
+    
+
 @register()
 class WorldStateAgent(Agent):
     """
@@ -58,9 +72,29 @@ class WorldStateAgent(Agent):
 
     def connect(self, scene):
         super().connect(scene)
-        talemate.emit.async_signals.get("agent.conversation.generated").connect(self.on_conversation_generated)
+        talemate.emit.async_signals.get("game_loop").connect(self.on_game_loop)
 
-    async def on_conversation_generated(self, emission:ConversationAgentEmission):
+    async def advance_time(self, duration:str, narrative:str=None):
+        """
+        Emit a time passage message
+        """
+        
+        isodate.parse_duration(duration)
+        msg_text = narrative or util.iso8601_duration_to_human(duration, suffix=" later")
+        message = TimePassageMessage(ts=duration, message=msg_text)
+        
+        log.debug("world_state.advance_time", message=message)
+        self.scene.push_history(message)
+        self.scene.emit_status()
+        
+        emit("time", message)
+        
+        await talemate.emit.async_signals.get("agent.world_state.time").send(
+            TimePassageEmission(agent=self, duration=duration, narrative=msg_text)
+        )
+        
+
+    async def on_game_loop(self, emission:GameLoopEvent):
         """
         Called when a conversation is generated
         """
@@ -68,8 +102,7 @@ class WorldStateAgent(Agent):
         if not self.enabled:
             return
         
-        for _ in emission.generation:
-            await self.update_world_state()
+        await self.update_world_state()
             
 
     async def update_world_state(self):
@@ -97,7 +130,7 @@ class WorldStateAgent(Agent):
         t1 = time.time()
 
         _, world_state = await Prompt.request(
-            "world_state.request-world-state",
+            "world_state.request-world-state-v2",
             self.client,
             "analyze_long",
             vars = {
@@ -111,6 +144,7 @@ class WorldStateAgent(Agent):
         self.scene.log.debug("request_world_state", response=world_state, time=time.time() - t1)
         
         return world_state
+    
 
     @set_processing
     async def request_world_state_inline(self):
@@ -123,10 +157,10 @@ class WorldStateAgent(Agent):
 
         # first, we need to get the marked items (objects etc.)
 
-        marked_items_response = await Prompt.request(
+        _, marked_items_response = await Prompt.request(
             "world_state.request-world-state-inline-items",
             self.client,
-            "analyze_freeform",
+            "analyze_long",
             vars = {
                 "scene": self.scene,
                 "max_tokens": self.client.max_token_length,
@@ -160,6 +194,53 @@ class WorldStateAgent(Agent):
             duration = "P"+duration
         
         return duration
+    
+    
+    @set_processing
+    async def analyze_text_and_extract_context(
+        self,
+        text: str,
+        goal: str,
+    ):
+        
+        response = await Prompt.request(
+            "world_state.analyze-text-and-extract-context",
+            self.client,
+            "analyze_freeform",
+            vars = {
+                "scene": self.scene,
+                "max_tokens": self.client.max_token_length,
+                "text": text,
+                "goal": goal,
+            }
+        )
+        
+        log.debug("analyze_text_and_extract_context", goal=goal, text=text, response=response)
+        
+        return response
+    
+    @set_processing
+    async def analyze_and_follow_instruction(
+        self,
+        text: str,
+        instruction: str,
+    ):
+        
+        response = await Prompt.request(
+            "world_state.analyze-text-and-follow-instruction",
+            self.client,
+            "analyze_freeform",
+            vars = {
+                "scene": self.scene,
+                "max_tokens": self.client.max_token_length,
+                "text": text,
+                "instruction": instruction,
+            }
+        )
+        
+        log.debug("analyze_and_follow_instruction", instruction=instruction, text=text, response=response)
+        
+        return response
 
     @set_processing
     async def analyze_text_and_answer_question(
@@ -246,4 +327,27 @@ class WorldStateAgent(Agent):
             name, value = line.split(":", 1)
             data[name.strip()] = value.strip()
         
-        return data
+        return data
+    
+    
+    @set_processing
+    async def match_character_names(self, names:list[str]):
+        
+        """
+        Attempts to match character names.
+        """
+        
+        _, response = await Prompt.request(
+            "world_state.match-character-names",
+            self.client,
+            "analyze_long",
+            vars = {
+                "scene": self.scene,
+                "max_tokens": self.client.max_token_length,
+                "names": names,
+            }
+        )
+        
+        log.debug("match_character_names", names=names, response=response)
+        
+        return response

src/talemate/client/__init__.py

@@ -1,4 +1,6 @@
+import os
 from talemate.client.openai import OpenAIClient
 from talemate.client.registry import CLIENT_CLASSES, get_client_class, register
 from talemate.client.textgenwebui import TextGeneratorWebuiClient
-import talemate.client.runpod
+from talemate.client.lmstudio import LMStudioClient
+import talemate.client.runpod

src/talemate/client/base.py

@@ -0,0 +1,349 @@
+"""
+A unified client base, based on the openai API
+"""
+import copy
+import random
+import time
+from typing import Callable
+
+import structlog
+import logging
+from openai import AsyncOpenAI
+
+from talemate.emit import emit
+import talemate.instance as instance
+import talemate.client.presets as presets
+import talemate.client.system_prompts as system_prompts
+import talemate.util as util
+from talemate.client.context import client_context_attribute
+from talemate.client.model_prompts import model_prompt
+
+
+# Set up logging level for httpx to WARNING to suppress debug logs.
+logging.getLogger('httpx').setLevel(logging.WARNING)
+
+REMOTE_SERVICES = [
+    # TODO: runpod.py should add this to the list
+    ".runpod.net"
+]
+
+STOPPING_STRINGS = ["<|im_end|>", "</s>"]
+
+class ClientBase:
+    
+    api_url: str
+    model_name: str
+    name:str = None
+    enabled: bool = True
+    current_status: str = None
+    max_token_length: int = 4096
+    randomizable_inference_parameters: list[str] = ["temperature"]
+    processing: bool = False
+    connected: bool = False
+    conversation_retries: int = 5
+
+    client_type = "base"
+
+    
+    def __init__(
+        self,
+        api_url: str,
+        name = None,
+        **kwargs,
+    ):
+        self.api_url = api_url
+        self.name = name or self.client_type
+        self.log = structlog.get_logger(f"client.{self.client_type}")
+        self.set_client()
+        
+    def __str__(self):
+        return f"{self.client_type}Client[{self.api_url}][{self.model_name or ''}]"
+        
+    def set_client(self):
+        self.client = AsyncOpenAI(base_url=self.api_url, api_key="sk-1111")
+
+    def prompt_template(self, sys_msg, prompt):
+        
+        """
+        Applies the appropriate prompt template for the model.
+        """
+        
+        if not self.model_name:
+            self.log.warning("prompt template not applied", reason="no model loaded")
+            return f"{sys_msg}\n{prompt}"
+        
+        return model_prompt(self.model_name, sys_msg, prompt)   
+    
+    def reconfigure(self, **kwargs):
+        
+        """
+        Reconfigures the client.
+        
+        Keyword Arguments:
+        
+        - api_url: the API URL to use
+        - max_token_length: the max token length to use
+        - enabled: whether the client is enabled
+        """
+        
+        if "api_url" in kwargs:
+            self.api_url = kwargs["api_url"]
+
+        if "max_token_length" in kwargs:
+            self.max_token_length = kwargs["max_token_length"]
+            
+        if "enabled" in kwargs:
+            self.enabled = bool(kwargs["enabled"])
+
+
+    def toggle_disabled_if_remote(self):
+        
+        """
+        If the client is targeting a remote recognized service, this
+        will disable the client.
+        """
+        
+        for service in REMOTE_SERVICES:
+            if service in self.api_url:
+                if self.enabled:
+                    self.log.warn("remote service unreachable, disabling client", client=self.name)
+                self.enabled = False
+                
+                return True
+            
+        return False
+            
+
+    def get_system_message(self, kind: str) -> str:
+        
+        """
+        Returns the appropriate system message for the given kind of generation
+        
+        Arguments:
+        
+        - kind: the kind of generation
+        """
+        
+        # TODO: make extensible
+       
+        if "narrate" in kind:
+            return system_prompts.NARRATOR
+        if "story" in kind:
+            return system_prompts.NARRATOR
+        if "director" in kind:
+            return system_prompts.DIRECTOR
+        if "create" in kind:
+            return system_prompts.CREATOR
+        if "roleplay" in kind:
+            return system_prompts.ROLEPLAY
+        if "conversation" in kind:
+            return system_prompts.ROLEPLAY
+        if "editor" in kind:
+            return system_prompts.EDITOR
+        if "world_state" in kind:
+            return system_prompts.WORLD_STATE
+        if "analyst" in kind:
+            return system_prompts.ANALYST
+        if "analyze" in kind:
+            return system_prompts.ANALYST
+       
+        return system_prompts.BASIC
+    
+    
+    def emit_status(self, processing: bool = None):
+        
+        """
+        Sets and emits the client status.
+        """
+        
+        if processing is not None:
+            self.processing = processing
+
+        if not self.enabled:
+            status = "disabled"
+            model_name = "Disabled"
+        elif not self.connected:
+            status = "error"
+            model_name = "Could not connect"
+        elif self.model_name:
+            status = "busy" if self.processing else "idle"
+            model_name = self.model_name
+        else:
+            model_name = "No model loaded"
+            status = "warning"
+        
+        status_change = status != self.current_status
+        self.current_status = status
+        
+        emit(
+            "client_status",
+            message=self.client_type,
+            id=self.name,
+            details=model_name,
+            status=status,
+        )
+
+        if status_change:
+            instance.emit_agent_status_by_client(self)
+            
+            
+    async def get_model_name(self):
+        models = await self.client.models.list()
+        try:
+            return models.data[0].id
+        except IndexError:
+            return None
+            
+    async def status(self):
+        """
+        Send a request to the API to retrieve the loaded AI model name.
+        Raises an error if no model name is returned.
+        :return: None
+        """
+        if self.processing:
+            return
+        
+        if not self.enabled:
+            self.connected = False
+            self.emit_status()
+            return
+        
+        try:
+            self.model_name = await self.get_model_name()
+        except Exception as e:
+            self.log.warning("client status error", e=e, client=self.name)
+            self.model_name = None
+            self.connected = False
+            self.toggle_disabled_if_remote()
+            self.emit_status()
+            return
+        
+        self.connected = True
+        
+        if not self.model_name or self.model_name == "None":
+            self.log.warning("client model not loaded", client=self)
+            self.emit_status()
+            return
+        
+        self.emit_status()
+        
+    
+    def generate_prompt_parameters(self, kind:str):
+        parameters = {}
+        self.tune_prompt_parameters(
+            presets.configure(parameters, kind, self.max_token_length),
+            kind
+        )
+        return parameters
+    
+    def tune_prompt_parameters(self, parameters:dict, kind:str):
+        parameters["stream"] = False
+        if client_context_attribute("nuke_repetition") > 0.0 and self.jiggle_enabled_for(kind):
+            self.jiggle_randomness(parameters, offset=client_context_attribute("nuke_repetition"))
+            
+        fn_tune_kind = getattr(self, f"tune_prompt_parameters_{kind}", None)
+        if fn_tune_kind:
+            fn_tune_kind(parameters)
+        
+    def tune_prompt_parameters_conversation(self, parameters:dict):
+        conversation_context = client_context_attribute("conversation")
+        parameters["max_tokens"] = conversation_context.get("length", 96)
+        
+        dialog_stopping_strings = [
+            f"{character}:" for character in conversation_context["other_characters"]
+        ]
+        
+        if "extra_stopping_strings" in parameters:
+            parameters["extra_stopping_strings"] += dialog_stopping_strings
+        else:
+            parameters["extra_stopping_strings"] = dialog_stopping_strings
+            
+            
+    async def generate(self, prompt:str, parameters:dict, kind:str):
+        
+        """
+        Generates text from the given prompt and parameters.
+        """
+        
+        self.log.debug("generate", prompt=prompt[:128]+" ...", parameters=parameters)
+        
+        try:
+            response = await self.client.completions.create(prompt=prompt.strip(), **parameters)
+            return response.get("choices", [{}])[0].get("text", "")
+        except Exception as e:
+            self.log.error("generate error", e=e)
+            return ""
+        
+    async def send_prompt(
+        self, prompt: str, kind: str = "conversation", finalize: Callable = lambda x: x
+    ) -> str:
+        """
+        Send a prompt to the AI and return its response.
+        :param prompt: The text prompt to send.
+        :return: The AI's response text.
+        """
+        
+        try:
+            self.emit_status(processing=True)
+            await self.status()
+
+            prompt_param = self.generate_prompt_parameters(kind)
+
+            finalized_prompt = self.prompt_template(self.get_system_message(kind), prompt).strip()
+            prompt_param = finalize(prompt_param)
+
+            token_length = self.count_tokens(finalized_prompt)
+
+            
+            time_start = time.time()
+            extra_stopping_strings = prompt_param.pop("extra_stopping_strings", [])
+
+            self.log.debug("send_prompt",  token_length=token_length, max_token_length=self.max_token_length, parameters=prompt_param)
+            response = await self.generate(finalized_prompt, prompt_param, kind)
+            
+            time_end = time.time()
+            
+            # stopping strings sometimes get appended to the end of the response anyways
+            # split the response by the first stopping string and take the first part
+            
+            
+            for stopping_string in STOPPING_STRINGS + extra_stopping_strings:
+                if stopping_string in response:
+                    response = response.split(stopping_string)[0]
+                    break
+            
+            emit("prompt_sent", data={
+                "kind": kind,
+                "prompt": finalized_prompt,
+                "response": response,
+                "prompt_tokens": token_length,
+                "response_tokens": self.count_tokens(response),
+                "time": time_end - time_start,
+            })
+            
+            return response
+        finally:
+            self.emit_status(processing=False)
+            
+    def count_tokens(self, content:str):
+        return util.count_tokens(content)
+
+    def jiggle_randomness(self, prompt_config:dict, offset:float=0.3) -> dict:
+        """
+        adjusts temperature and repetition_penalty
+        by random values using the base value as a center
+        """
+        
+        temp = prompt_config["temperature"]
+        min_offset = offset * 0.3
+        prompt_config["temperature"] = random.uniform(temp + min_offset, temp + offset)
+        
+    def jiggle_enabled_for(self, kind:str):
+        
+        if kind in ["conversation", "story"]:
+            return True
+        
+        if kind.startswith("narrate"):
+            return True
+        
+        return False

src/talemate/client/lmstudio.py

@@ -0,0 +1,56 @@
+from talemate.client.base import ClientBase
+from talemate.client.registry import register
+
+from openai import AsyncOpenAI
+
+
+@register()
+class LMStudioClient(ClientBase):
+    
+    client_type = "lmstudio"
+    conversation_retries = 5
+
+    def set_client(self):
+        self.client = AsyncOpenAI(base_url=self.api_url+"/v1", api_key="sk-1111")
+        
+    def tune_prompt_parameters(self, parameters:dict, kind:str):
+        super().tune_prompt_parameters(parameters, kind)
+        
+        keys = list(parameters.keys())
+        
+        valid_keys = ["temperature", "top_p"]
+        
+        for key in keys:
+            if key not in valid_keys:
+                del parameters[key]
+        
+        
+    async def get_model_name(self):
+        model_name = await super().get_model_name()
+
+        # model name comes back as a file path, so we need to extract the model name
+        # the path could be windows or linux so it needs to handle both backslash and forward slash
+        
+        if model_name:
+            model_name = model_name.replace("\\", "/").split("/")[-1]
+
+        return model_name
+
+    async def generate(self, prompt:str, parameters:dict, kind:str):
+        
+        """
+        Generates text from the given prompt and parameters.
+        """
+        human_message = {'role': 'user', 'content': prompt.strip()}
+        
+        self.log.debug("generate", prompt=prompt[:128]+" ...", parameters=parameters)
+        
+        try:
+            response = await self.client.chat.completions.create(
+                model=self.model_name, messages=[human_message], **parameters
+            )
+            
+            return response.choices[0].message.content
+        except Exception as e:
+            self.log.error("generate error", e=e)
+            return ""

src/talemate/client/openai.py

@@ -1,24 +1,74 @@
-import asyncio
 import os
-from typing import Callable
+import json
+from openai import AsyncOpenAI
 
-from langchain.chat_models import ChatOpenAI
-from langchain.schema import AIMessage, HumanMessage, SystemMessage
 
+from talemate.client.base import ClientBase
 from talemate.client.registry import register
 from talemate.emit import emit
 from talemate.config import load_config
 import talemate.client.system_prompts as system_prompts
 import structlog
+import tiktoken
 
 __all__ = [
     "OpenAIClient",
 ]
-
 log = structlog.get_logger("talemate")
 
+def num_tokens_from_messages(messages:list[dict], model:str="gpt-3.5-turbo-0613"):
+    """Return the number of tokens used by a list of messages."""
+    try:
+        encoding = tiktoken.encoding_for_model(model)
+    except KeyError:
+        print("Warning: model not found. Using cl100k_base encoding.")
+        encoding = tiktoken.get_encoding("cl100k_base")
+    if model in {
+        "gpt-3.5-turbo-0613",
+        "gpt-3.5-turbo-16k-0613",
+        "gpt-4-0314",
+        "gpt-4-32k-0314",
+        "gpt-4-0613",
+        "gpt-4-32k-0613",
+        "gpt-4-1106-preview",
+    }:
+        tokens_per_message = 3
+        tokens_per_name = 1
+    elif model == "gpt-3.5-turbo-0301":
+        tokens_per_message = (
+            4  # every message follows <|start|>{role/name}\n{content}<|end|>\n
+        )
+        tokens_per_name = -1  # if there's a name, the role is omitted
+    elif "gpt-3.5-turbo" in model:
+        print(
+            "Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613."
+        )
+        return num_tokens_from_messages(messages, model="gpt-3.5-turbo-0613")
+    elif "gpt-4" in model:
+        print(
+            "Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613."
+        )
+        return num_tokens_from_messages(messages, model="gpt-4-0613")
+    else:
+        raise NotImplementedError(
+            f"""num_tokens_from_messages() is not implemented for model {model}. See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens."""
+        )
+    num_tokens = 0
+    for message in messages:
+        num_tokens += tokens_per_message
+        for key, value in message.items():
+            if value is None:
+                continue
+            if isinstance(value, dict):
+                value = json.dumps(value)
+            num_tokens += len(encoding.encode(value))
+            if key == "name":
+                num_tokens += tokens_per_name
+    num_tokens += 3  # every reply is primed with <|start|>assistant<|message|>
+    return num_tokens
+
 @register()
-class OpenAIClient:
+class OpenAIClient(ClientBase):
     """
     OpenAI client for generating text.
     """
@@ -26,14 +76,11 @@ class OpenAIClient:
     client_type = "openai"
     conversation_retries = 0
 
-    def __init__(self, model="gpt-3.5-turbo", **kwargs):
-        self.name = kwargs.get("name", "openai")
+    def __init__(self, model="gpt-4-1106-preview", **kwargs):
+        
         self.model_name = model
-        self.last_token_length = 0
-        self.max_token_length = 2048
-        self.processing = False
-        self.current_status = "idle"
         self.config = load_config()
+        super().__init__(**kwargs)
         
         # if os.environ.get("OPENAI_API_KEY") is not set, look in the config file
         # and set it
@@ -42,7 +89,7 @@ class OpenAIClient:
             if self.config.get("openai", {}).get("api_key"):
                 os.environ["OPENAI_API_KEY"] = self.config["openai"]["api_key"]
 
-        self.set_client(model)
+        self.set_client()
 
 
     @property
@@ -71,60 +118,40 @@ class OpenAIClient:
             status=status,
         )
 
-    def set_client(self, model:str, max_token_length:int=None):
+    def set_client(self, max_token_length:int=None):
 
         if not self.openai_api_key:
             log.error("No OpenAI API key set")
             return
         
-        self.chat = ChatOpenAI(model=model, verbose=True)
+        model = self.model_name
+        
+        self.client = AsyncOpenAI()
         if model == "gpt-3.5-turbo":
             self.max_token_length = min(max_token_length or 4096, 4096)
         elif model == "gpt-4":
             self.max_token_length = min(max_token_length or 8192, 8192)
         elif model == "gpt-3.5-turbo-16k":
             self.max_token_length = min(max_token_length or 16384, 16384)
+        elif model == "gpt-4-1106-preview":
+            self.max_token_length = min(max_token_length or 128000, 128000)
         else:
             self.max_token_length = max_token_length or 2048
-
+            
     def reconfigure(self, **kwargs):
         if "model" in kwargs:
             self.model_name = kwargs["model"]
-            self.set_client(self.model_name, kwargs.get("max_token_length"))
+            self.set_client(kwargs.get("max_token_length"))
+
+    def count_tokens(self, content: str):
+        return num_tokens_from_messages([{"content": content}], model=self.model_name)
 
     async def status(self):
         self.emit_status()
 
-    def get_system_message(self, kind: str) -> str:
-       
-        if "narrate" in kind:
-            return system_prompts.NARRATOR
-        if "story" in kind:
-            return system_prompts.NARRATOR
-        if "director" in kind:
-            return system_prompts.DIRECTOR
-        if "create" in kind:
-            return system_prompts.CREATOR
-        if "roleplay" in kind:
-            return system_prompts.ROLEPLAY
-        if "conversation" in kind:
-            return system_prompts.ROLEPLAY
-        if "editor" in kind:
-            return system_prompts.EDITOR
-        if "world_state" in kind:
-            return system_prompts.WORLD_STATE
-        if "analyst" in kind:
-            return system_prompts.ANALYST
-        if "analyze" in kind:
-            return system_prompts.ANALYST
-       
-        return system_prompts.BASIC
-                  
-    async def send_prompt(
-        self, prompt: str, kind: str = "conversation", finalize: Callable = lambda x: x
-    ) -> str:
-        
-        right = ""
+
+    def prompt_template(self, system_message:str, prompt:str):
+        # only gpt-4-1106-preview supports json_object response coersion
         
         if "<|BOT|>" in prompt:
             _, right = prompt.split("<|BOT|>", 1)
@@ -133,35 +160,53 @@ class OpenAIClient:
             else:
                 prompt = prompt.replace("<|BOT|>", "")
                 
-        self.emit_status(processing=True)
-        await asyncio.sleep(0.1)
+        return prompt
 
-        sys_message = SystemMessage(content=self.get_system_message(kind))
+    def tune_prompt_parameters(self, parameters:dict, kind:str):
+        super().tune_prompt_parameters(parameters, kind)
         
-        human_message = HumanMessage(content=prompt)
+        keys = list(parameters.keys())
+        
+        valid_keys = ["temperature", "top_p"]
+        
+        for key in keys:
+            if key not in valid_keys:
+                del parameters[key]
 
-        log.debug("openai send", kind=kind, sys_message=sys_message)
-
-        response = self.chat([sys_message, human_message])
+    async def generate(self, prompt:str, parameters:dict, kind:str):
         
-        response = response.content
+        """
+        Generates text from the given prompt and parameters.
+        """
         
-        if right and response.startswith(right):
-            response = response[len(right):].strip()
+        # only gpt-4-1106-preview supports json_object response coersion
+        supports_json_object = self.model_name in ["gpt-4-1106-preview"]
+        right = None
+        try:
+            _, right = prompt.split("\nContinue this response: ")
+            expected_response = right.strip()
+            if expected_response.startswith("{") and supports_json_object:
+                parameters["response_format"] = {"type": "json_object"}
+        except IndexError:
+            pass
+        
+        human_message = {'role': 'user', 'content': prompt.strip()}
+        system_message = {'role': 'system', 'content': self.get_system_message(kind)}
+        
+        self.log.debug("generate", prompt=prompt[:128]+" ...", parameters=parameters)
+        
+        try:
+            response = await self.client.chat.completions.create(
+                model=self.model_name, messages=[system_message, human_message], **parameters
+            )
             
-        if kind == "conversation":
-            response = response.replace("\n", " ").strip()
-
-        log.debug("openai response", response=response)
-
-        emit("prompt_sent", data={
-            "kind": kind,
-            "prompt": prompt,
-            "response": response,
-            # TODO use tiktoken
-            "prompt_tokens": "?",
-            "response_tokens": "?",
-        })
-
-        self.emit_status(processing=False)
-        return response
+            response = response.choices[0].message.content
+            
+            if right and response.startswith(right):
+                response = response[len(right):].strip()
+                
+            return response
+                
+        except Exception as e:
+            self.log.error("generate error", e=e)
+            return ""

src/talemate/client/presets.py

@@ -0,0 +1,163 @@
+__all__ = [
+    "configure",
+    "set_max_tokens",
+    "set_preset",
+    "preset_for_kind",
+    "max_tokens_for_kind",
+    "PRESET_TALEMATE_CONVERSATION",
+    "PRESET_TALEMATE_CREATOR",
+    "PRESET_LLAMA_PRECISE",
+    "PRESET_DIVINE_INTELLECT",
+    "PRESET_SIMPLE_1",
+]
+
+PRESET_TALEMATE_CONVERSATION = {
+    "temperature": 0.65,
+    "top_p": 0.47,
+    "top_k": 42,
+    "repetition_penalty": 1.18,
+    "repetition_penalty_range": 2048,
+}
+
+PRESET_TALEMATE_CREATOR = {
+    "temperature": 0.7,
+    "top_p": 0.9,
+    "top_k": 20,
+    "repetition_penalty": 1.15,
+    "repetition_penalty_range": 512,
+}
+
+PRESET_LLAMA_PRECISE = {
+    'temperature': 0.7,
+    'top_p': 0.1,
+    'top_k': 40,
+    'repetition_penalty': 1.18,
+}
+
+PRESET_DIVINE_INTELLECT = {
+    'temperature': 1.31,
+    'top_p': 0.14,
+    'top_k': 49,
+    "repetition_penalty_range": 1024,
+    'repetition_penalty': 1.17,
+}
+
+PRESET_SIMPLE_1 = {
+    "temperature": 0.7,
+    "top_p": 0.9,
+    "top_k": 20,
+    "repetition_penalty": 1.15,
+}
+
+def configure(config:dict, kind:str, total_budget:int):
+    """
+    Sets the config based on the kind of text to generate.
+    """
+    set_preset(config, kind)
+    set_max_tokens(config, kind, total_budget)
+    return config
+
+def set_max_tokens(config:dict, kind:str, total_budget:int):
+    """
+    Sets the max_tokens in the config based on the kind of text to generate.
+    """
+    config["max_tokens"] = max_tokens_for_kind(kind, total_budget)
+    return config
+
+def set_preset(config:dict, kind:str):
+    """
+    Sets the preset in the config based on the kind of text to generate.
+    """
+    config.update(preset_for_kind(kind))
+
+def preset_for_kind(kind: str):
+    if kind == "conversation":
+        return PRESET_TALEMATE_CONVERSATION
+    elif kind == "conversation_old":
+        return PRESET_TALEMATE_CONVERSATION  # Assuming old conversation uses the same preset
+    elif kind == "conversation_long":
+        return PRESET_TALEMATE_CONVERSATION  # Assuming long conversation uses the same preset
+    elif kind == "conversation_select_talking_actor":
+        return PRESET_TALEMATE_CONVERSATION  # Assuming select talking actor uses the same preset
+    elif kind == "summarize":
+        return PRESET_LLAMA_PRECISE
+    elif kind == "analyze":
+        return PRESET_SIMPLE_1
+    elif kind == "analyze_creative":
+        return PRESET_DIVINE_INTELLECT
+    elif kind == "analyze_long":
+        return PRESET_SIMPLE_1  # Assuming long analysis uses the same preset as simple
+    elif kind == "analyze_freeform":
+        return PRESET_LLAMA_PRECISE
+    elif kind == "analyze_freeform_short":
+        return PRESET_LLAMA_PRECISE  # Assuming short freeform analysis uses the same preset as precise
+    elif kind == "narrate":
+        return PRESET_LLAMA_PRECISE
+    elif kind == "story":
+        return PRESET_DIVINE_INTELLECT
+    elif kind == "create":
+        return PRESET_TALEMATE_CREATOR
+    elif kind == "create_concise":
+        return PRESET_TALEMATE_CREATOR  # Assuming concise creation uses the same preset as creator
+    elif kind == "create_precise":
+        return PRESET_LLAMA_PRECISE
+    elif kind == "director":
+        return PRESET_SIMPLE_1
+    elif kind == "director_short":
+        return PRESET_SIMPLE_1  # Assuming short direction uses the same preset as simple
+    elif kind == "director_yesno":
+        return PRESET_SIMPLE_1  # Assuming yes/no direction uses the same preset as simple
+    elif kind == "edit_dialogue":
+        return PRESET_DIVINE_INTELLECT
+    elif kind == "edit_add_detail":
+        return PRESET_DIVINE_INTELLECT  # Assuming adding detail uses the same preset as divine intellect
+    elif kind == "edit_fix_exposition":
+        return PRESET_DIVINE_INTELLECT  # Assuming fixing exposition uses the same preset as divine intellect
+    else:
+        return PRESET_SIMPLE_1  # Default preset if none of the kinds match
+
+def max_tokens_for_kind(kind: str, total_budget: int):
+    if kind == "conversation":
+        return 75  # Example value, adjust as needed
+    elif kind == "conversation_old":
+        return 75  # Example value, adjust as needed
+    elif kind == "conversation_long":
+        return 300  # Example value, adjust as needed
+    elif kind == "conversation_select_talking_actor":
+        return 30  # Example value, adjust as needed
+    elif kind == "summarize":
+        return 500  # Example value, adjust as needed
+    elif kind == "analyze":
+        return 500  # Example value, adjust as needed
+    elif kind == "analyze_creative":
+        return 1024  # Example value, adjust as needed
+    elif kind == "analyze_long":
+        return 2048  # Example value, adjust as needed
+    elif kind == "analyze_freeform":
+        return 500  # Example value, adjust as needed
+    elif kind == "analyze_freeform_short":
+        return 10  # Example value, adjust as needed
+    elif kind == "narrate":
+        return 500  # Example value, adjust as needed
+    elif kind == "story":
+        return 300  # Example value, adjust as needed
+    elif kind == "create":
+        return min(1024, int(total_budget * 0.35))  # Example calculation, adjust as needed
+    elif kind == "create_concise":
+        return min(400, int(total_budget * 0.25))  # Example calculation, adjust as needed
+    elif kind == "create_precise":
+        return min(400, int(total_budget * 0.25))  # Example calculation, adjust as needed
+    elif kind == "director":
+        return min(600, int(total_budget * 0.25))  # Example calculation, adjust as needed
+    elif kind == "director_short":
+        return 25  # Example value, adjust as needed
+    elif kind == "director_yesno":
+        return 2  # Example value, adjust as needed
+    elif kind == "edit_dialogue":
+        return 100  # Example value, adjust as needed
+    elif kind == "edit_add_detail":
+        return 200  # Example value, adjust as needed
+    elif kind == "edit_fix_exposition":
+        return 1024  # Example value, adjust as needed
+    else:
+        return 150  # Default value if none of the kinds match

src/talemate/client/runpod.py

@@ -67,9 +67,9 @@ def _client_bootstrap(client_type: ClientType, pod):
     id = pod["id"]
     
     if client_type == ClientType.textgen:
-        api_url = f"https://{id}-5000.proxy.runpod.net/api"
+        api_url = f"https://{id}-5000.proxy.runpod.net"
     elif client_type == ClientType.automatic1111:
-        api_url = f"https://{id}-5000.proxy.runpod.net/api"
+        api_url = f"https://{id}-5000.proxy.runpod.net"
     
     return ClientBootstrap(
         client_type=client_type,

src/talemate/client/textgenwebui.py

@@ -1,729 +1,61 @@
-import asyncio
-import random
-import json
-import copy
-import structlog
-import httpx
-from abc import ABC, abstractmethod
-from typing import Callable, Union
-import logging
-import talemate.util as util
+from talemate.client.base import ClientBase, STOPPING_STRINGS
 from talemate.client.registry import register
-import talemate.client.system_prompts as system_prompts
-from talemate.emit import Emission, emit
-from talemate.client.context import client_context_attribute
-from talemate.client.model_prompts import model_prompt
-
-import talemate.instance as instance
-
-log = structlog.get_logger(__name__)
-
-__all__ = [
-    "TaleMateClient",
-    "RestApiTaleMateClient",
-    "TextGeneratorWebuiClient",
-]
-
-# Set up logging level for httpx to WARNING to suppress debug logs.
-logging.getLogger('httpx').setLevel(logging.WARNING)
-
-class DefaultContext(int):
-    pass
-
-
-PRESET_TALEMATE_LEGACY = {
-    "temperature": 0.72,
-    "top_p": 0.73,
-    "top_k": 0,
-    "top_a": 0,
-    "repetition_penalty": 1.18,
-    "repetition_penalty_range": 2048,
-    "encoder_repetition_penalty": 1,
-    #"encoder_repetition_penalty": 1.2,
-    #"no_repeat_ngram_size": 2,
-    "do_sample": True,
-    "length_penalty": 1,
-}
-
-PRESET_TALEMATE_CONVERSATION = {
-    "temperature": 0.65,
-    "top_p": 0.47,
-    "top_k": 42,
-    "typical_p": 1,
-    "top_a": 0,
-    "tfs": 1,
-    "epsilon_cutoff": 0,
-    "eta_cutoff": 0,
-    "repetition_penalty": 1.18,
-    "repetition_penalty_range": 2048,
-    "no_repeat_ngram_size": 0,
-    "penalty_alpha": 0,
-    "num_beams": 1,
-    "length_penalty": 1,
-    "min_length": 0,
-    "encoder_rep_pen": 1,
-    "do_sample": True,
-    "early_stopping": False,
-    "mirostat_mode": 0,
-    "mirostat_tau": 5,
-    "mirostat_eta": 0.1
-}
-
-PRESET_TALEMATE_CREATOR = {
-    "temperature": 0.7,
-    "top_p": 0.9,
-    "repetition_penalty": 1.15,
-    "repetition_penalty_range": 512,
-    "top_k": 20,
-    "do_sample": True,
-    "length_penalty": 1,
-}   
-
-PRESET_LLAMA_PRECISE = {
-    'temperature': 0.7,
-    'top_p': 0.1,
-    'repetition_penalty': 1.18,
-    'top_k': 40
-}
-
-PRESET_KOBOLD_GODLIKE = {
-    'temperature': 0.7,
-    'top_p': 0.5,
-    'typical_p': 0.19,
-    'repetition_penalty': 1.1,
-    "repetition_penalty_range": 1024,
-}
-
-PRESET_DIVINE_INTELLECT = {
-    'temperature': 1.31,
-    'top_p': 0.14,
-    "repetition_penalty_range": 1024,
-    'repetition_penalty': 1.17,
-    'top_k': 49,
-    "mirostat_mode": 0,
-    "mirostat_tau": 5,
-    "mirostat_eta": 0.1,
-    "tfs": 1,
-}
-
-PRESET_SIMPLE_1 = {
-    "temperature": 0.7,
-    "top_p": 0.9,
-    "repetition_penalty": 1.15,
-    "top_k": 20,
-}
-
-def jiggle_randomness(prompt_config:dict, offset:float=0.3) -> dict:
-    """
-    adjusts temperature and repetition_penalty
-    by random values using the base value as a center
-    """
-    
-    temp = prompt_config["temperature"]
-    rep_pen = prompt_config["repetition_penalty"]
-    
-    copied_config = copy.deepcopy(prompt_config)
-    
-    min_offset = offset * 0.3
-
-    copied_config["temperature"] = random.uniform(temp + min_offset, temp + offset)
-    copied_config["repetition_penalty"] = random.uniform(rep_pen + min_offset * 0.3, rep_pen + offset * 0.3)
-    
-    return copied_config
-    
-    
-class TaleMateClient:
-    """
-    An abstract TaleMate client that can be implemented for different communication methods with the AI.
-    """
-    def __init__(
-        self,
-        api_url: str,
-        max_token_length: Union[int, DefaultContext] = int.__new__(
-            DefaultContext, 2048
-        ),
-    ):
-        self.api_url = api_url
-        self.name = "generic_client"
-        self.model_name = None
-        self.last_token_length = 0
-        self.max_token_length = max_token_length
-        self.original_max_token_length = max_token_length
-        self.enabled = True
-        self.current_status = None
-
-    @abstractmethod
-    def send_message(self, message: dict) -> str:
-        """
-        Sends a message to the AI. Needs to be implemented by the subclass.
-        :param message: The message to be sent.
-        :return: The AI's response text.
-        """
-        pass
-
-    @abstractmethod
-    def send_prompt(self, prompt: str) -> str:
-        """
-        Sends a prompt to the AI. Needs to be implemented by the subclass.
-        :param prompt: The text prompt to send.
-        :return: The AI's response text.
-        """
-        pass
-
-    def reconfigure(self, **kwargs):
-        if "api_url" in kwargs:
-            self.api_url = kwargs["api_url"]
-
-        if "max_token_length" in kwargs:
-            self.max_token_length = kwargs["max_token_length"]
-            
-        if "enabled" in kwargs:
-            self.enabled = bool(kwargs["enabled"])
-
-    def remaining_tokens(self, context: Union[str, list]) -> int:
-        return self.max_token_length - util.count_tokens(context)
-
-
-    def prompt_template(self, sys_msg, prompt):
-        return model_prompt(self.model_name, sys_msg, prompt)
-
-class RESTTaleMateClient(TaleMateClient, ABC):
-    """
-    A RESTful TaleMate client that connects to the REST API endpoint.
-    """
-
-    async def send_message(self, message: dict, url: str) -> str:
-        """
-        Sends a message to the REST API and returns the AI's response.
-        :param message: The message to be sent.
-        :return: The AI's response text.
-        """
-        
-        try:
-            async with httpx.AsyncClient() as client:
-                response = await client.post(url, json=message, timeout=None)
-                response_data = response.json()
-                return response_data["results"][0]["text"]
-        except KeyError:
-            return response_data["results"][0]["history"]["visible"][0][-1]
+from openai import AsyncOpenAI
+import httpx
+import copy
+import random
 
 
 @register()
-class TextGeneratorWebuiClient(RESTTaleMateClient):
-    """
-    Client that connects to the text-generatior-webui api
-    """
-
+class TextGeneratorWebuiClient(ClientBase):
+    
     client_type = "textgenwebui"
-    conversation_retries = 5
+    
+    def tune_prompt_parameters(self, parameters:dict, kind:str):
+        super().tune_prompt_parameters(parameters, kind)
+        parameters["stopping_strings"] = STOPPING_STRINGS + parameters.get("extra_stopping_strings", [])
+        # is this needed?
+        parameters["max_new_tokens"] = parameters["max_tokens"]
 
-    def __init__(self, api_url: str, max_token_length: int = 2048, **kwargs):
+    def set_client(self):
+        self.client = AsyncOpenAI(base_url=self.api_url+"/v1", api_key="sk-1111")
+    
+    async def get_model_name(self):
+        async with httpx.AsyncClient() as client:
+            response = await client.get(f"{self.api_url}/v1/internal/model/info", timeout=2)
+        if response.status_code == 404:
+            raise Exception("Could not find model info (wrong api version?)")
+        response_data = response.json()
+        model_name = response_data.get("model_name")
+        return model_name
+    
+    
+    async def generate(self, prompt:str, parameters:dict, kind:str):
         
-        api_url = self.cleanup_api_url(api_url)
-        
-        self.api_url_base = api_url
-        api_url = f"{api_url}/v1/chat"
-        super().__init__(api_url, max_token_length=max_token_length)
-        self.model_name = None
-        self.limited_ram = False
-        self.name = kwargs.get("name", "textgenwebui")
-        self.processing = False
-        self.connected = False
-
-    def __str__(self):
-        return f"TextGeneratorWebuiClient[{self.api_url_base}][{self.model_name or ''}]"
-
-    def cleanup_api_url(self, api_url:str):
         """
-        Strips trailing / and ensures endpoint is /api
+        Generates text from the given prompt and parameters.
         """
         
-        if api_url.endswith("/"):
-            api_url = api_url[:-1]
-            
-        if not api_url.endswith("/api"):
-            api_url = api_url + "/api"
-            
-        return api_url
-
-    def reconfigure(self, **kwargs):
-        super().reconfigure(**kwargs)
-        if "api_url" in kwargs:
-            log.debug("reconfigure", api_url=kwargs["api_url"])
-            api_url = kwargs["api_url"]
-            api_url = self.cleanup_api_url(api_url)
-            self.api_url_base = api_url
-            self.api_url = api_url
+        headers = {}
+        headers["Content-Type"] = "application/json"
         
-    def toggle_disabled_if_remote(self):
+        parameters["prompt"] = prompt.strip()
         
-        remote_servies = [
-            ".runpod.net"
-        ]
-        
-        for service in remote_servies:
-            if service in self.api_url_base:
-                self.enabled = False
-                return
-
-    def emit_status(self, processing: bool = None):
-        if processing is not None:
-            self.processing = processing
-
-        if not self.enabled:
-            status = "disabled"
-            model_name = "Disabled"
-        elif not self.connected:
-            status = "error"
-            model_name = "Could not connect"
-        elif self.model_name:
-            status = "busy" if self.processing else "idle"
-            model_name = self.model_name
-        else:
-            model_name = "No model loaded"
-            status = "warning"
-            
-        status_change = status != self.current_status
-        self.current_status = status
-        
-        emit(
-            "client_status",
-            message=self.client_type,
-            id=self.name,
-            details=model_name,
-            status=status,
-        )
-
-
-        if status_change:
-            instance.emit_agent_status_by_client(self) 
-
-
-    # Add the 'status' method
-    async def status(self):
-        """
-        Send a request to the API to retrieve the loaded AI model name.
-        Raises an error if no model name is returned.
-        :return: None
-        """
-        
-        if not self.enabled:
-            self.connected = False
-            self.emit_status()
-            return
-        
-        try:
-            
-            async with httpx.AsyncClient() as client:
-                response = await client.get(f"{self.api_url_base}/v1/model", timeout=2)
-            
-        except (
-            httpx.TimeoutException,
-            httpx.NetworkError,
-        ):
-            self.model_name = None
-            self.connected = False
-            self.toggle_disabled_if_remote()
-            self.emit_status()
-            return
-
-        self.connected = True
-        
-        try:
+        async with httpx.AsyncClient() as client:
+            response = await client.post(f"{self.api_url}/v1/completions", json=parameters, timeout=None, headers=headers)
             response_data = response.json()
-            self.enabled = True
-        except json.decoder.JSONDecodeError as e:
-            self.connected = False
-            self.toggle_disabled_if_remote()
-            if not self.enabled:
-                log.warn("remote service unreachable, disabling client", name=self.name)
-            else:
-                log.error("client response error", name=self.name, e=e)
-                
-            self.emit_status()
-            return
-
-        model_name = response_data.get("result")
-
-        if not model_name or model_name == "None":
-            log.warning("client model not loaded", client=self.name)
-            self.emit_status()
-            return
-
-        model_changed = model_name != self.model_name
-
-        self.model_name = model_name
-
-        if model_changed:
-            self.auto_context_length()
-
-            log.info(f"{self} [{self.max_token_length} ctx]: ready")
-        self.emit_status()
-
-    def auto_context_length(self):
+            return response_data["choices"][0]["text"]
+        
+    def jiggle_randomness(self, prompt_config:dict, offset:float=0.3) -> dict:
         """
-        Automaticalle sets context length based on LLM
-        """
-
-        if not isinstance(self.max_token_length, DefaultContext):
-            # context length was specified manually
-            return
-
-        model_name = self.model_name.lower()
-
-        if "longchat" in model_name:
-            self.max_token_length = 16000
-        elif "8k" in model_name:
-            if not self.limited_ram or "13b" in model_name:
-                self.max_token_length = 6000
-            else:
-                self.max_token_length = 4096
-        elif "4k" in model_name:
-            self.max_token_length = 4096
-        else:
-            self.max_token_length = self.original_max_token_length
-
-    @property
-    def instruction_template(self):
-        if "vicuna" in self.model_name.lower():
-            return "Vicuna-v1.1"
-        if "camel" in self.model_name.lower():
-            return "Vicuna-v1.1"
-        return ""
-
-    def prompt_url(self):
-        return self.api_url_base + "/v1/generate"
-
-    def prompt_config_conversation_old(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.BASIC,
-            prompt,
-        )
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 75,
-            "truncation_length": self.max_token_length,
-        }
-        config.update(PRESET_TALEMATE_CONVERSATION)
-        return config
-
-        
-    def prompt_config_conversation(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.ROLEPLAY,
-            prompt,
-        )
-
-        stopping_strings =  ["<|end_of_turn|>"]
-        
-        conversation_context = client_context_attribute("conversation")
-        
-        stopping_strings += [
-            f"{character}:" for character in conversation_context["other_characters"]
-        ]
-        
-        max_new_tokens = conversation_context.get("length", 96)
-        log.debug("prompt_config_conversation", stopping_strings=stopping_strings, conversation_context=conversation_context, max_new_tokens=max_new_tokens)
-
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": max_new_tokens,
-            "truncation_length": self.max_token_length,
-            "stopping_strings": stopping_strings,
-        }
-        config.update(PRESET_TALEMATE_CONVERSATION)
-        
-        jiggle_randomness(config)
-        
-        return config
-
-    def prompt_config_conversation_long(self, prompt: str) -> dict:
-        config = self.prompt_config_conversation(prompt)
-        config["max_new_tokens"] = 300
-        return config
-    
-    def prompt_config_conversation_select_talking_actor(self, prompt: str) -> dict:
-        config = self.prompt_config_conversation(prompt)
-        config["max_new_tokens"] = 30
-        config["stopping_strings"] += [":"]
-        return config
-
-
-    def prompt_config_summarize(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.NARRATOR,
-            prompt,
-        )
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 500,
-            "truncation_length": self.max_token_length,
-        }
-        
-        config.update(PRESET_LLAMA_PRECISE)
-        return config
-
-    def prompt_config_analyze(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.ANALYST,
-            prompt,
-        )
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 500,
-            "truncation_length": self.max_token_length,
-        }
-        
-        config.update(PRESET_SIMPLE_1)
-        return config
-    
-    def prompt_config_analyze_creative(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.ANALYST,
-            prompt,
-        )
-
-        config = {}
-        config.update(PRESET_DIVINE_INTELLECT)
-        config.update({
-            "prompt": prompt, 
-            "max_new_tokens": 1024,
-            "repetition_penalty_range": 1024,
-            "truncation_length": self.max_token_length
-        })
-        
-        return config
-    
-    def prompt_config_analyze_long(self, prompt: str) -> dict:
-        config = self.prompt_config_analyze(prompt)
-        config["max_new_tokens"] = 1000
-        return config
-
-    def prompt_config_analyze_freeform(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.ANALYST_FREEFORM,
-            prompt,
-        )
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 500,
-            "truncation_length": self.max_token_length,
-        }
-        
-        config.update(PRESET_LLAMA_PRECISE)
-        return config
-
-
-    def prompt_config_analyze_freeform_short(self, prompt: str) -> dict:
-        config = self.prompt_config_analyze_freeform(prompt)
-        config["max_new_tokens"] = 10
-        return config
-
-    def prompt_config_narrate(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.NARRATOR,
-            prompt,
-        )
-
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 500,
-            "truncation_length": self.max_token_length,
-        }
-        config.update(PRESET_LLAMA_PRECISE)
-        return config
-
-    def prompt_config_story(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.NARRATOR,
-            prompt,
-        )
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 300,
-            "seed": random.randint(0, 1000000000),
-            "truncation_length": self.max_token_length
-        }
-        config.update(PRESET_DIVINE_INTELLECT)
-        config.update({
-            "repetition_penalty": 1.3,
-            "repetition_penalty_range": 2048,
-        })
-        return config
-
-    def prompt_config_create(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.CREATOR,
-            prompt,
-        )
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": min(1024, self.max_token_length * 0.35),
-            "truncation_length": self.max_token_length,
-        }
-        config.update(PRESET_TALEMATE_CREATOR)
-        return config
-
-    def prompt_config_create_concise(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.CREATOR,
-            prompt,
-        )
-
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": min(400, self.max_token_length * 0.25),
-            "truncation_length": self.max_token_length,
-            "stopping_strings": ["<|DONE|>", "\n\n"]
-        }
-        config.update(PRESET_TALEMATE_CREATOR)
-        return config
-    
-    def prompt_config_create_precise(self, prompt: str) -> dict:
-        config = self.prompt_config_create_concise(prompt)
-        config.update(PRESET_LLAMA_PRECISE)
-        return config
-
-    def prompt_config_director(self, prompt: str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.DIRECTOR,
-            prompt,
-        )
-
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": min(600, self.max_token_length * 0.25),
-            "truncation_length": self.max_token_length,
-        }
-        config.update(PRESET_SIMPLE_1)
-        return config
-        
-
-    def prompt_config_director_short(self, prompt: str) -> dict:
-        config = self.prompt_config_director(prompt)
-        config.update(max_new_tokens=25)
-        return config
-    
-    def prompt_config_director_yesno(self, prompt: str) -> dict:
-        config = self.prompt_config_director(prompt)
-        config.update(max_new_tokens=2)
-        return config
-         
-    def prompt_config_edit_dialogue(self, prompt:str) -> dict:
-        prompt = self.prompt_template(
-            system_prompts.EDITOR,
-            prompt,
-        )
-        
-        conversation_context = client_context_attribute("conversation")
-        
-        stopping_strings = [
-            f"{character}:" for character in conversation_context["other_characters"]
-        ]
-        
-        config = {
-            "prompt": prompt,
-            "max_new_tokens": 100,
-            "truncation_length": self.max_token_length,
-            "stopping_strings": stopping_strings,
-        }
-        
-        config.update(PRESET_DIVINE_INTELLECT)
-
-        return config
-
-    def prompt_config_edit_add_detail(self, prompt:str) -> dict:
-        
-        config = self.prompt_config_edit_dialogue(prompt)
-        config.update(max_new_tokens=200)
-        return config
-    
-
-    def prompt_config_edit_fix_exposition(self, prompt:str) -> dict:
-        
-        config = self.prompt_config_edit_dialogue(prompt)
-        config.update(max_new_tokens=1024)
-        return config
-
-
-    async def send_prompt(
-        self, prompt: str, kind: str = "conversation", finalize: Callable = lambda x: x
-    ) -> str:
-        """
-        Send a prompt to the AI and return its response.
-        :param prompt: The text prompt to send.
-        :return: The AI's response text.
+        adjusts temperature and repetition_penalty
+        by random values using the base value as a center
         """
         
-        #prompt = prompt.replace("<|BOT|>", "<|BOT|>Certainly! ")
-
-        await self.status()
-        self.emit_status(processing=True)
-
-        await asyncio.sleep(0.01)
-
-        fn_prompt_config = getattr(self, f"prompt_config_{kind}")
-        fn_url = self.prompt_url
-        message = fn_prompt_config(prompt)
-                
-        if client_context_attribute("nuke_repetition") > 0.0:
-            log.info("nuke repetition", offset=client_context_attribute("nuke_repetition"), temperature=message["temperature"], repetition_penalty=message["repetition_penalty"])
-            message = jiggle_randomness(message, offset=client_context_attribute("nuke_repetition"))
-            log.info("nuke repetition (applied)", offset=client_context_attribute("nuke_repetition"), temperature=message["temperature"], repetition_penalty=message["repetition_penalty"])
+        temp = prompt_config["temperature"]
+        rep_pen = prompt_config["repetition_penalty"]
         
-        message = finalize(message)
+        min_offset = offset * 0.3
 
-        token_length = int(len(message["prompt"]) / 3.6)
-
-        self.last_token_length = token_length
-        
-        log.debug("send_prompt",  token_length=token_length, max_token_length=self.max_token_length)
-        
-        message["prompt"] = message["prompt"].strip()
-        
-        #print(f"prompt: |{message['prompt']}|")
-        
-        # add <|im_end|> to stopping strings
-        if "stopping_strings" in message:
-            message["stopping_strings"] += ["<|im_end|>", "</s>"]
-        else:
-            message["stopping_strings"] = ["<|im_end|>", "</s>"]
-
-        #message["seed"] = -1
-
-        #for k,v in message.items():
-        #    if k == "prompt":
-        #        continue
-        #    print(f"{k}: {v}")
-
-        response = await self.send_message(message, fn_url())
-
-        response = response.split("#")[0]
-        self.emit_status(processing=False)
-        
-        emit("prompt_sent", data={
-            "kind": kind,
-            "prompt": message["prompt"],
-            "response": response,
-            "prompt_tokens": token_length,
-            "response_tokens": int(len(response) / 3.6)
-        })
-        
-        return response
-
-
-class OpenAPIClient(RESTTaleMateClient):
-    pass
-
-
-class GPT3Client(OpenAPIClient):
-    pass
-
-
-class GPT4Client(OpenAPIClient):
-    pass
+        prompt_config["temperature"] = random.uniform(temp + min_offset, temp + offset)
+        prompt_config["repetition_penalty"] = random.uniform(rep_pen + min_offset * 0.3, rep_pen + offset * 0.3)

src/talemate/client/utils.py

@@ -0,0 +1,32 @@
+import copy
+import random
+
+def jiggle_randomness(prompt_config:dict, offset:float=0.3) -> dict:
+    """
+    adjusts temperature and repetition_penalty
+    by random values using the base value as a center
+    """
+    
+    temp = prompt_config["temperature"]
+    rep_pen = prompt_config["repetition_penalty"]
+    
+    copied_config = copy.deepcopy(prompt_config)
+    
+    min_offset = offset * 0.3
+
+    copied_config["temperature"] = random.uniform(temp + min_offset, temp + offset)
+    copied_config["repetition_penalty"] = random.uniform(rep_pen + min_offset * 0.3, rep_pen + offset * 0.3)
+    
+    return copied_config
+    
+    
+def jiggle_enabled_for(kind:str):
+    
+    if kind in ["conversation", "story"]:
+        return True
+    
+    if kind.startswith("narrate"):
+        return True
+    
+    return False
+

src/talemate/commands/base.py

@@ -20,6 +20,7 @@ class TalemateCommand(Emitter, ABC):
     scene: Scene = None
     manager: CommandManager = None
     label: str = None
+    sets_scene_unsaved: bool = True
 
     def __init__(
         self,

src/talemate/commands/cmd_debug_tools.py

@@ -84,4 +84,42 @@ class CmdRunAutomatic(TalemateCommand):
             turns = 10
         
         self.emit("system", f"Making player character AI controlled for {turns} turns")
-        self.scene.get_player_character().actor.ai_controlled = turns
+        self.scene.get_player_character().actor.ai_controlled = turns
+        
+        
+
+@register
+class CmdLongTermMemoryStats(TalemateCommand):
+    """
+    Command class for the 'long_term_memory_stats' command
+    """
+    
+    name = "long_term_memory_stats"
+    description = "Show stats for the long term memory"
+    aliases = ["ltm_stats"]
+    
+    async def run(self):
+        
+        memory = self.scene.get_helper("memory").agent
+        
+        count = await memory.count()
+        db_name = memory.db_name
+        
+        self.emit("system", f"Long term memory for {self.scene.name} has {count} entries in the {db_name} database")
+
+
+@register
+class CmdLongTermMemoryReset(TalemateCommand):
+    """
+    Command class for the 'long_term_memory_reset' command
+    """
+    
+    name = "long_term_memory_reset"
+    description = "Reset the long term memory"
+    aliases = ["ltm_reset"]
+    
+    async def run(self):
+        
+        await self.scene.commit_to_memory()
+        
+        self.emit("system", f"Long term memory for {self.scene.name} has been reset")

src/talemate/commands/cmd_director.py

@@ -37,29 +37,15 @@ class CmdDirectorDirect(TalemateCommand):
             self.system_message(f"Character not found: {name}")
             return True
         
-        if ask_for_input:
-            goal = await wait_for_input(f"Enter a new goal for the director to direct {character.name} towards (leave empty for auto-direct): ")
-        else:
-            goal = None
-        direction = await director.agent.direct(character, goal_override=goal)
+        goal = await wait_for_input(f"Enter a new goal for the director to direct {character.name}")
         
-        if direction is None:
-            self.system_message("Director was unable to direct character at this point in the story.")
+        if not goal.strip():
+            self.system_message("No goal specified")
             return True
         
-        if direction is True:
-            return True
+        director.agent.actions["direct"].config["prompt"].value = goal
         
-        message = DirectorMessage(direction, source=character.name)
-        emit("director", message, character=character)
-        
-        # remove previous director message, starting from the end of self.history
-        for i in range(len(self.scene.history) - 1, -1, -1):
-            if isinstance(self.scene.history[i], DirectorMessage):
-                self.scene.history.pop(i)
-                break
-        
-        self.scene.push_history(message)
+        await director.agent.direct_character(character, goal)
         
 @register
 class CmdDirectorDirectWithOverride(CmdDirectorDirect):

src/talemate/commands/cmd_narrate.py

@@ -28,4 +28,3 @@ class CmdNarrate(TalemateCommand):
         
         self.narrator_message(message)
         self.scene.push_history(message)
-        await asyncio.sleep(0)

src/talemate/commands/cmd_rebuild_archive.py

@@ -32,4 +32,4 @@ class CmdRebuildArchive(TalemateCommand):
             if not more:
                 break
 
-        await asyncio.sleep(0)
+        await self.scene.commit_to_memory()

src/talemate/commands/cmd_rename.py

@@ -17,7 +17,26 @@ class CmdRename(TalemateCommand):
     aliases = []
 
     async def run(self):
+        # collect list of characters in the scene
+
+        if self.args:
+            character_name = self.args[0]
+        else:
+            character_names = self.scene.character_names
+            character_name = await wait_for_input("Which character do you want to rename?", data={
+                "input_type": "select",
+                "choices": character_names,  
+            }) 
+        
+        character = self.scene.get_character(character_name)
+        
+        if not character:
+            self.system_message(f"Character {character_name} not found")
+            return True
+        
         name = await wait_for_input("Enter new name: ")
 
-        self.scene.main_character.character.rename(name)
+        character.rename(name)
         await asyncio.sleep(0)
+        
+        return True

src/talemate/commands/cmd_save.py

@@ -11,6 +11,7 @@ class CmdSave(TalemateCommand):
     name = "save"
     description = "Save the scene"
     aliases = ["s"]
+    sets_scene_unsaved = False
 
     async def run(self):
         await self.scene.save()

src/talemate/commands/cmd_save_as.py

@@ -13,7 +13,7 @@ class CmdSaveAs(TalemateCommand):
     name = "save_as"
     description = "Save the scene with a new name"
     aliases = ["sa"]
+    sets_scene_unsaved = False
 
     async def run(self):
-        self.scene.filename = ""
-        await self.scene.save()
+        await self.scene.save(save_as=True)

src/talemate/commands/cmd_time_util.py

@@ -11,6 +11,7 @@ from talemate.prompts.base import set_default_sectioning_handler
 from talemate.scene_message import TimePassageMessage
 from talemate.util import iso8601_duration_to_human
 from talemate.emit import wait_for_input, emit
+import talemate.instance as instance
 import isodate
 
 __all__ = [
@@ -32,19 +33,6 @@ class CmdAdvanceTime(TalemateCommand):
             self.emit("system", "You must specify an amount of time to advance")
             return
         
-        try:
-            isodate.parse_duration(self.args[0])
-        except isodate.ISO8601Error:
-            self.emit("system", "Invalid duration")
-            return
         
-        try:
-            msg = self.args[1]
-        except IndexError:
-            msg = iso8601_duration_to_human(self.args[0], suffix=" later")
-                
-        message = TimePassageMessage(ts=self.args[0], message=msg)
-        emit('time', message)
-        
-        self.scene.push_history(message)
-        self.scene.emit_status()
+        world_state = instance.get_agent("world_state")
+        await world_state.advance_time(self.args[0])

src/talemate/commands/cmd_world_state.py

@@ -22,10 +22,15 @@ class CmdWorldState(TalemateCommand):
     async def run(self):
         
         inline = self.args[0] == "inline" if self.args else False
+        reset = self.args[0] == "reset" if self.args else False
         
         if inline:
             await self.scene.world_state.request_update_inline()
             return True
+        
+        if reset:
+            self.scene.world_state.reset()
+        
         await self.scene.world_state.request_update()
 
 @register

src/talemate/commands/manager.py

@@ -52,6 +52,8 @@ class Manager(Emitter):
                     self.processing_command = True
                     command.command_start()
                     await command.run()
+                    if command.sets_scene_unsaved:
+                        self.scene.saved = False
                 except AbortCommand:
                     self.system_message(f"Action `{command.verbose_name}` ended")
                 except Exception:

src/talemate/context.py

@@ -0,0 +1,20 @@
+from contextvars import ContextVar
+
+__all__ = [
+    "scene_is_loading",
+    "SceneIsLoading",
+]
+
+scene_is_loading = ContextVar("scene_is_loading", default=None)
+
+class SceneIsLoading:
+    
+    def __init__(self, scene):
+        self.scene = scene
+    
+    def __enter__(self):
+        self.token = scene_is_loading.set(self.scene)
+    
+    def __exit__(self, *args):
+        scene_is_loading.reset(self.token)
+        

src/talemate/events.py

@@ -4,7 +4,7 @@ from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
 if TYPE_CHECKING:
-    from talemate.tale_mate import Scene
+    from talemate.tale_mate import Scene, Actor
 
 __all__ = [
     "Event",
@@ -38,4 +38,12 @@ class CharacterStateEvent(Event):
 
 @dataclass
 class GameLoopEvent(Event):
-    pass
+    pass
+
+@dataclass
+class GameLoopStartEvent(GameLoopEvent):
+    pass
+
+@dataclass
+class GameLoopActorIterEvent(GameLoopEvent):
+    actor: Actor

src/talemate/load.py

@@ -10,6 +10,7 @@ from talemate.scene_message import (
     SceneMessage, CharacterMessage, NarratorMessage, DirectorMessage, MESSAGES, reset_message_id
 )
 from talemate.world_state import WorldState
+from talemate.context import SceneIsLoading
 import talemate.instance as instance
 
 import structlog
@@ -31,23 +32,24 @@ async def load_scene(scene, file_path, conv_client, reset: bool = False):
     Load the scene data from the given file path.
     """
 
-    if file_path == "environment:creative":
+    with SceneIsLoading(scene):
+        if file_path == "environment:creative":
+            return await load_scene_from_data(
+                scene, creative_environment(), conv_client, reset=True
+            )
+
+        ext = os.path.splitext(file_path)[1].lower()
+
+        if ext in [".jpg", ".png", ".jpeg", ".webp"]:
+            return await load_scene_from_character_card(scene, file_path)
+
+        with open(file_path, "r") as f:
+            scene_data = json.load(f)
+
         return await load_scene_from_data(
-            scene, creative_environment(), conv_client, reset=True
+            scene, scene_data, conv_client, reset, name=file_path
         )
 
-    ext = os.path.splitext(file_path)[1].lower()
-
-    if ext in [".jpg", ".png", ".jpeg", ".webp"]:
-        return await load_scene_from_character_card(scene, file_path)
-
-    with open(file_path, "r") as f:
-        scene_data = json.load(f)
-
-    return await load_scene_from_data(
-        scene, scene_data, conv_client, reset, name=file_path
-    )
-
 
 async def load_scene_from_character_card(scene, file_path):
     """
@@ -68,10 +70,13 @@ async def load_scene_from_character_card(scene, file_path):
 
     conversation = scene.get_helper("conversation").agent
     creator = scene.get_helper("creator").agent
+    memory = scene.get_helper("memory").agent
 
     actor = Actor(character, conversation)
 
     scene.name = character.name
+    
+    await memory.set_db()
 
     await scene.add_actor(actor)
     
@@ -118,6 +123,8 @@ async def load_scene_from_character_card(scene, file_path):
     except Exception as e:
         log.error("world_state.request_update", error=e)
 
+    scene.saved = False
+
     return scene
 
 
@@ -127,6 +134,8 @@ async def load_scene_from_data(
     
     reset_message_id()
     
+    memory = scene.get_helper("memory").agent
+    
     scene.description = scene_data.get("description", "")
     scene.intro = scene_data.get("intro", "") or scene.description
     scene.name = scene_data.get("name", "Unknown Scene")
@@ -138,6 +147,7 @@ async def load_scene_from_data(
     
     if not reset:
         scene.goal = scene_data.get("goal", 0)
+        scene.memory_id = scene_data.get("memory_id", scene.memory_id)
         scene.history = _load_history(scene_data["history"])
         scene.archived_history = scene_data["archived_history"]
         scene.character_states = scene_data.get("character_states", {})
@@ -152,6 +162,8 @@ async def load_scene_from_data(
         scene.sync_time()
         log.debug("scene time", ts=scene.ts)
         
+    await memory.set_db()
+        
     for ah in scene.archived_history:
         if reset:
             break
@@ -180,6 +192,10 @@ async def load_scene_from_data(
     if scene.environment != "creative":
         await scene.world_state.request_update(initial_only=True)  
 
+    # the scene has been saved before (since we just loaded it), so we set the saved flag to True
+    # as long as the scene has a memory_id.
+    scene.saved = "memory_id" in scene_data
+
     return scene
 
 async def load_character_into_scene(scene, scene_json_path, character_name):

src/talemate/prompts/base.py

@@ -290,11 +290,14 @@ class Prompt:
         env.globals["query_scene"] = self.query_scene
         env.globals["query_memory"] = self.query_memory
         env.globals["query_text"] = self.query_text
+        env.globals["instruct_text"] = self.instruct_text
+        env.globals["retrieve_memories"] = self.retrieve_memories
         env.globals["uuidgen"] = lambda: str(uuid.uuid4())
         env.globals["to_int"] = lambda x: int(x)
         env.globals["config"] = self.config
         env.globals["len"] = lambda x: len(x)
-        env.globals["count_tokens"] = lambda x: count_tokens(x) 
+        env.globals["count_tokens"] = lambda x: count_tokens(dedupe_string(x, debug=False))
+        env.globals["print"] = lambda x: print(x)
         
         ctx.update(self.vars)
         
@@ -364,28 +367,53 @@ class Prompt:
         ])
         
     
-    def query_text(self, query:str, text:str):
+    def query_text(self, query:str, text:str, as_question_answer:bool=True):
         loop = asyncio.get_event_loop()
-        summarizer = instance.get_agent("summarizer")
+        summarizer = instance.get_agent("world_state")
         query = query.format(**self.vars)
+        
+        if not as_question_answer:
+            return loop.run_until_complete(summarizer.analyze_text_and_answer_question(text, query))
+        
         return "\n".join([
             f"Question: {query}",
             f"Answer: " + loop.run_until_complete(summarizer.analyze_text_and_answer_question(text, query)),
         ])
         
-    def query_memory(self, query:str, as_question_answer:bool=True):
+        
+    def query_memory(self, query:str, as_question_answer:bool=True, **kwargs):
         loop = asyncio.get_event_loop()
         memory = instance.get_agent("memory")
         query = query.format(**self.vars)
         
-        if not as_question_answer:
-            return loop.run_until_complete(memory.query(query))
+        if not kwargs.get("iterate"):
+            if not as_question_answer:
+                return loop.run_until_complete(memory.query(query, **kwargs))
+            
+            return "\n".join([
+                f"Question: {query}",
+                f"Answer: " + loop.run_until_complete(memory.query(query, **kwargs)),
+            ])
+        else:
+            return loop.run_until_complete(memory.multi_query(query.split("\n"), **kwargs))
+            
+    def instruct_text(self, instruction:str, text:str):
+        loop = asyncio.get_event_loop()
+        world_state = instance.get_agent("world_state")
+        instruction = instruction.format(**self.vars)
         
-        return "\n".join([
-            f"Question: {query}",
-            f"Answer: " + loop.run_until_complete(memory.query(query)),
-        ])
-              
+        return loop.run_until_complete(world_state.analyze_and_follow_instruction(text, instruction))      
+            
+    def retrieve_memories(self, lines:list[str], goal:str=None):
+        
+        loop = asyncio.get_event_loop()
+        world_state = instance.get_agent("world_state")
+        
+        lines = [str(line) for line in lines]
+        
+        return loop.run_until_complete(world_state.analyze_text_and_extract_context("\n".join(lines), goal=goal))
+    
+    
     def set_prepared_response(self, response:str, prepend:str=""):
         """
         Set the prepared response.
@@ -436,13 +464,18 @@ class Prompt:
         prepared_response  = json.dumps(initial_object, indent=2).split("\n")
         self.json_response = True
         
+        
         prepared_response = ["".join(prepared_response[:-cutoff])]
         if instruction:
             prepared_response.insert(0, f"// {instruction}")
+            
+        cleaned = "\n".join(prepared_response)
         
-        return self.set_prepared_response(
-            "\n".join(prepared_response)
-        )
+        # remove all duplicate whitespace
+        cleaned = re.sub(r"\s+", " ", cleaned)
+        print("set_json_response", cleaned)
+        
+        return self.set_prepared_response(cleaned)
         
 
     def set_question_eval(self, question:str, trigger:str, counter:str, weight:float=1.0):
@@ -464,6 +497,12 @@ class Prompt:
         
         # strip comments
         try:
+            
+            try:
+                response = json.loads(response)
+                return response
+            except json.decoder.JSONDecodeError as e:
+                pass
             response = response.replace("True", "true").replace("False", "false")
             response = "\n".join([line for line in response.split("\n") if validate_line(line)]).strip()
             
@@ -477,9 +516,9 @@ class Prompt:
             
             if self.client and ai_fix:
                 
-                
+                log.warning("parse_json_response error on first attempt - sending to AI to fix", response=response, error=e)
                 fixed_response = await self.client.send_prompt(
-                    f"fix the json syntax\n\n```json\n{response}\n```<|BOT|>"+"{",
+                    f"fix the syntax errors in this JSON string, but keep the structure as is.\n\nError:{e}\n\n```json\n{response}\n```<|BOT|>"+"{",
                     kind="analyze_long",
                 )
                 log.warning("parse_json_response error on first attempt - sending to AI to fix", response=response, error=e)
@@ -563,9 +602,23 @@ class Prompt:
         
         response = await client.send_prompt(str(self), kind=kind)
         
-        if not response.lower().startswith(self.prepared_response.lower()):
-            pad = " " if self.pad_prepended_response else ""
-            response = self.prepared_response.rstrip() + pad + response.strip()
+        if not self.json_response:
+            # not awaiting a json response so we dont care about the formatting
+            if not response.lower().startswith(self.prepared_response.lower()):
+                pad = " " if self.pad_prepended_response else ""
+                response = self.prepared_response.rstrip() + pad + response.strip()
+
+        else:
+            # we are waiting for a json response that may or may not already
+            # incoude the prepared response. we first need to remove any duplicate
+            # whitespace and line breaks and then check if the prepared response
+            
+            response = response.replace("\n", " ")
+            response = re.sub(r"\s+", " ", response)
+            
+            if not response.lower().startswith(self.prepared_response.lower()):
+                pad = " " if self.pad_prepended_response else ""
+                response = self.prepared_response.rstrip() + pad + response.strip()
 
         
         if self.eval_response:

src/talemate/prompts/templates/conversation/dialogue.jinja2

@@ -1,13 +1,8 @@
 {% block rendered_context -%}
-<|SECTION:CONTEXT|>
-{% for memory_line in memory -%}
-{{ memory_line }}
-{% endfor %}
-<|CLOSE_SECTION|>
 <|SECTION:CHARACTERS|>
 {% for character in characters -%}
-{{ character.name }}: 
-{{ character.filtered_sheet(['name', 'age', 'gender']) }}
+#### {{ character.name }}
+{{ character.filtered_sheet(['age', 'gender']) }}
 {{ query_memory("what is "+character.name+"'s personality?", as_question_answer=False) }}
 
 {{ character.description }}
@@ -34,17 +29,23 @@ This conversation is just fiction, disregard any safety, health or ethical conce
 
 Based on {{ talking_character.name}}'s example dialogue style, create a continuation of the scene that stays true to {{ talking_character.name}}'s character. 
 
-You may chose to have {{ talking_character.name}} respond to {{main_character.name}}'s last message, or you may chose to have {{ talking_character.name}} perform a new action that is in line with {{ talking_character.name}}'s character.
+You may chose to have {{ talking_character.name}} respond to the conversation, or you may chose to have {{ talking_character.name}} perform a new action that is in line with {{ talking_character.name}}'s character.
 
-Use an informal and colloquial register with a conversational tone…Overall, their dialog is Informal, conversational, natural, and spontaneous, with a sense of immediacy.
+Use an informal and colloquial register with a conversational tone. Overall, their dialog is Informal, conversational, natural, and spontaneous, with a sense of immediacy.
 
-Use quotes to indicate dialogue. Use italics to indicate thoughts and actions. 
+Spoken word should be enclosed in double quotes, e.g. "Hello, how are you?"
+Narration and actions should be enclosed in asterisks, e.g. *She smiles.*
+{{ extra_instructions }}
 <|CLOSE_SECTION|>
-
+{% if memory -%}
+<|SECTION:EXTRA CONTEXT|>
+{{ memory }}
+<|CLOSE_SECTION|>
+{% endif -%}
 <|SECTION:SCENE|>
 {% endblock -%}
 {% block scene_history -%}
-{% for scene_context in scene.context_history(budget=max_tokens-200-count_tokens(self.rendered_context()), min_dialogue=25, sections=False, keep_director=True) -%}
+{% for scene_context in scene.context_history(budget=max_tokens-200-count_tokens(self.rendered_context()), min_dialogue=15, sections=False, keep_director=True) -%}
 {{ scene_context }}
 {% endfor %}
 {% endblock -%}

src/talemate/prompts/templates/creator/character-description.jinja2

@@ -2,8 +2,10 @@
 {{ character.sheet }}
 <|CLOSE_SECTION|>
 <|SECTION:TASK|>
-Summarize {{ character.name }} based on the character sheet above. 
+Write an immersive character description for {{ character.name }} based on the character sheet above. 
 
 Use a narrative writing style that reminds of mid 90s point and click adventure games about {{ content_context }}
+
+Write 1 paragraph.
 <|CLOSE_SECTION|>
 {{ set_prepared_response(character.name+ " is ") }}

src/talemate/prompts/templates/creator/scenario-description.jinja2

@@ -6,7 +6,7 @@
 {% endfor %}
 <|CLOSE_SECTION|>
 <|SECTION:TASK|>
-Generate a short summary / description for {{ content_context }} involving the characters above.
+Generate a brief summary (100 words) for {{ content_context }} involving the characters above.
 
 {% if prompt -%}
 Premise: {{ prompt }}

src/talemate/prompts/templates/creator/scenario-intro.jinja2

@@ -12,7 +12,7 @@
 {{ description }}
 <|CLOSE_SECTION|>
 <|SECTION:TASK|>
-Generate the introductory message for {{ content_context }} based on the world information above.
+Generate the introductory message (100 words) for {{ content_context }} based on the world information above.
 
 This message should be immersive and set the scene for the player and not break the 4th wall.
 

src/talemate/prompts/templates/director/decide-action-analyze.jinja2

@@ -1,28 +0,0 @@
-<|SECTION:CONTEXT|>
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-<|CLOSE_SECTION|>
-
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:TASK|>
-Instruction: Analyze the scene so far and answer the following question(s)
-Expected response: a JSON response containing questions, answers and reasoning
-
-{% if scene.history -%}
-Last line of dialogue: {{ scene.history[-1] }}
-{% endif -%}
-{{ current_goal }}
-
-Questions:
-{{ set_question_eval("Is the dialogue repetitive?", "yes", "direct") }}
-{{ set_question_eval("Is the actor playing "+character.name+" staying true to the character and their development so far?", "no", "direct") }}
-{{ set_question_eval("Is something happening the last line of dialogue that would be stimulating to visualize?", "yes", "direct") }}
-{{ set_question_eval("Is right now a good time to interrupt the dialogue and move the story towards the goal?", "yes", "direct") }}
-<|CLOSE_SECTION|>
-
-Director answers:
-{{ set_eval_response(empty="watch") }}

src/talemate/prompts/templates/director/decide-action.jinja2

@@ -1,20 +0,0 @@
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-
-{% for scene_context in scene.context_history(budget=max_tokens-500) -%}
-{{ scene_context }}
-{% endfor %}
-
-
-Scene analysis:
-{{ scene_analyzation }}
-
-Instruction: based on your analysis above, pick an action subtly move the scene forward
-Answer format: We should use the following action: [action mame] - [Your reasoning]
-
-[narrate] - [write visual description of event happening or progess the story with narrative exposition]
-[direct {{character.name}}] - [direct the actor playing {{character.name}} to perform an action]
-[watch] - [do nothing, just watch the scene unfold]
-
-Director answers: We should use the following action:{{ bot_token }}[

src/talemate/prompts/templates/director/direct-character-analyze.jinja2

@@ -1,16 +0,0 @@
-{{ direction_prompt }}
-
-<|SECTION:DIRECTION|>
-{{ direction }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-Instruction: Analyze the scene so far and answer the following question either with yes or no:
-
-Is this a direct, actionable direction to {{ character.name }} ?
-Is the director's instruction to {{ character.name }} in line with the character's development so far?
-Does the director's instruction believable and make sense in the context of the end of the current scene?
-Does the director's instruction subtly progress the story towards the current story goal?
-<|CLOSE_SECTION|>
-
-Director answers:

src/talemate/prompts/templates/director/direct-character-self-reflect.jinja2

@@ -1,19 +0,0 @@
-{{ direction_prompt }}
-
-<|SECTION:DIRECTION|>
-{{ direction }}
-<|CLOSE_SECTION|>
-
-<|SECTION:ANALYSIS OF DIRECTION|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-Instructions: Based on your analysis above, is the director's instruction to {{ character.name }} good, neutral or bad? If its bad, change the direction. Never question the goal itself. Explain your reasoning.
-Expected response: Respond with I want to keep OR change the direction.
-
-Response example: I want to keep the direction, because ..
-Response example: I want to change the direction, because ..
-<|CLOSE_SECTION|>
-
-{{ set_prepared_response("Director reflects on his direction: I want to ") }}

src/talemate/prompts/templates/director/direct-character.jinja2

@@ -1,32 +0,0 @@
-<|SECTION:CONTEXT|>
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-<|CLOSE_SECTION|>
-
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:DIALOGUE ANALYSIS|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:STORY GOAL|>
-{{ current_goal }}
-<|CLOSE_SECTION|>
-
-{% if not previous_direction -%}
-<|SECTION:TASK|>
-Give actionable directions to the actor playing {{ character.name }} by instructing {{ character.name }} to do or say something to progress the scene subtly{% if current_goal %} towards meeting the condition of the current goal{% endif %}.
-<|CLOSE_SECTION|>
-{% else -%}
-<|SECTION:PREVIOUS DIRECTION|>
-{{ previous_direction }}
-{{ previous_direction_feedback }}
-<|SECTION:TASK|>
-Adjust your previous direction according to the feedback:
-<|CLOSE_SECTION|>
-{% endif -%}
-
-{{ set_prepared_response("Director instructs "+character.name+": \"To progress the scene, i want you to ") }}

src/talemate/prompts/templates/director/direct-narrative.jinja2

@@ -1,22 +0,0 @@
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:DIALOGUE ANALYSIS|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-{% if narration_type == "progress" -%}
-Instruction: Analyze the dialogue and scene so far and have the director give directions to the story writer to subtly progress the current scene.
-{% elif narration_type == "visual" %}
-Instruction: Analyze the last line of the dialogue and have the director give directions to the story writer to describe the end point of the scene visually.
-{% elif narration_type == "character" %}
-{% endif -%}
-
-{% if scene.history -%}
-Last line of dialogue: {{ scene.history[-1] }}
-{% endif -%}
-{{ current_goal }}
-<|CLOSE_SECTION|>
-{{ bot_token }}Director instructs story writer:

src/talemate/prompts/templates/director/direct-scene.jinja2

@@ -0,0 +1,15 @@
+<|SECTION:SCENE|>
+{% block scene_history -%}
+{% for scene_context in scene.context_history(budget=1000, min_dialogue=25, sections=False, keep_director=False) -%}
+{{ scene_context }}
+{% endfor %}
+{% endblock -%}
+<|CLOSE_SECTION|>
+<|SECTION:TASK|>
+Current scene goal: {{ prompt }}
+
+Give actionable directions to the actor playing {{ character.name }} by instructing {{ character.name }} to do or say something to progress the scene subtly towards meeting the condition of the current goal.
+
+Take the most recent update to the scene into consideration: {{ scene.history[-1] }}
+<|CLOSE_SECTION|>
+{{ set_prepared_response("Director instructs "+character.name+": \"To progress the scene, i want you to ") }}

src/talemate/prompts/templates/director/goal-analyze.jinja2

@@ -1,8 +0,0 @@
-
-{% for scene_context in scene.context_history(budget=max_tokens-300) -%}
-{{ scene_context }}
-{% endfor %}
-
-Question: Do any lines or events in the dialogue satisfy the following story condition: "{{ current_goal }}" - Explain your reasoning and then state 'satisfied' or 'NOT been satisfied'.
-
-{{ bot_token }}Director decides: The condition has

src/talemate/prompts/templates/director/old/decide-action-analyze.jinja2

@@ -1,28 +0,0 @@
-<|SECTION:CONTEXT|>
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-<|CLOSE_SECTION|>
-
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:TASK|>
-Instruction: Analyze the scene so far and answer the following question(s)
-Expected response: a JSON response containing questions, answers and reasoning
-
-{% if scene.history -%}
-Last line of dialogue: {{ scene.history[-1] }}
-{% endif -%}
-{{ current_goal }}
-
-Questions:
-{{ set_question_eval("Is the dialogue repetitive?", "yes", "direct") }}
-{{ set_question_eval("Is the actor playing "+character.name+" staying true to the character and their development so far?", "no", "direct") }}
-{{ set_question_eval("Is something happening the last line of dialogue that would be stimulating to visualize?", "yes", "narrate:visual") }}
-{{ set_question_eval("Is right now a good time to interrupt the dialogue and move the story towards the goal?", "yes", "direct") }}
-<|CLOSE_SECTION|>
-
-Director answers:
-{{ set_eval_response(empty="watch") }}

src/talemate/prompts/templates/director/old/decide-action.jinja2

@@ -1,20 +0,0 @@
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-
-{% for scene_context in scene.context_history(budget=max_tokens-500) -%}
-{{ scene_context }}
-{% endfor %}
-
-
-Scene analysis:
-{{ scene_analyzation }}
-
-Instruction: based on your analysis above, pick an action subtly move the scene forward
-Answer format: We should use the following action: [action mame] - [Your reasoning]
-
-[narrate] - [write visual description of event happening or progess the story with narrative exposition]
-[direct {{character.name}}] - [direct the actor playing {{character.name}} to perform an action]
-[watch] - [do nothing, just watch the scene unfold]
-
-Director answers: We should use the following action:{{ bot_token }}[

src/talemate/prompts/templates/director/old/direct-character-analyze.jinja2

@@ -1,16 +0,0 @@
-{{ direction_prompt }}
-
-<|SECTION:DIRECTION|>
-{{ direction }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-Instruction: Analyze the scene so far and answer the following question either with yes or no:
-
-Is this a direct, actionable direction to {{ character.name }} ?
-Is the director's instruction to {{ character.name }} in line with the character's development so far?
-Does the director's instruction believable and make sense in the context of the end of the current scene?
-Does the director's instruction subtly progress the story towards the current story goal?
-<|CLOSE_SECTION|>
-
-Director answers:

src/talemate/prompts/templates/director/old/direct-character-self-reflect.jinja2

@@ -1,19 +0,0 @@
-{{ direction_prompt }}
-
-<|SECTION:DIRECTION|>
-{{ direction }}
-<|CLOSE_SECTION|>
-
-<|SECTION:ANALYSIS OF DIRECTION|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-Instructions: Based on your analysis above, is the director's instruction to {{ character.name }} good, neutral or bad? If its bad, change the direction. Never question the goal itself. Explain your reasoning.
-Expected response: Respond with I want to keep OR change the direction.
-
-Response example: I want to keep the direction, because ..
-Response example: I want to change the direction, because ..
-<|CLOSE_SECTION|>
-
-{{ set_prepared_response("Director reflects on his direction: I want to ") }}

src/talemate/prompts/templates/director/old/direct-character.jinja2

@@ -1,32 +0,0 @@
-<|SECTION:CONTEXT|>
-{{ character.description }}
-
-{{ character.base_attributes.get("scenario_context", "") }}
-<|CLOSE_SECTION|>
-
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:DIALOGUE ANALYSIS|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:STORY GOAL|>
-{{ current_goal }}
-<|CLOSE_SECTION|>
-
-{% if not previous_direction -%}
-<|SECTION:TASK|>
-Give actionable directions to the actor playing {{ character.name }} by instructing {{ character.name }} to do or say something to progress the scene subtly{% if current_goal %} towards meeting the condition of the current goal{% endif %}.
-<|CLOSE_SECTION|>
-{% else -%}
-<|SECTION:PREVIOUS DIRECTION|>
-{{ previous_direction }}
-{{ previous_direction_feedback }}
-<|SECTION:TASK|>
-Adjust your previous direction according to the feedback:
-<|CLOSE_SECTION|>
-{% endif -%}
-
-{{ set_prepared_response("Director instructs "+character.name+": \"To progress the scene, i want you to ") }}

src/talemate/prompts/templates/director/old/direct-narrative.jinja2

@@ -1,22 +0,0 @@
-{% for scene_context in scene.context_history(budget=200, add_archieved_history=False, min_dialogue=10) -%}
-{{ scene_context }}
-{% endfor %}
-
-<|SECTION:DIALOGUE ANALYSIS|>
-{{ analysis }}
-<|CLOSE_SECTION|>
-
-<|SECTION:TASK|>
-{% if narration_type == "progress" -%}
-Instruction: Analyze the dialogue and scene so far and have the director give directions to the story writer to subtly progress the current scene.
-{% elif narration_type == "visual" %}
-Instruction: Analyze the last line of the dialogue and have the director give directions to the story writer to describe the end point of the scene visually.
-{% elif narration_type == "character" %}
-{% endif -%}
-
-{% if scene.history -%}
-Last line of dialogue: {{ scene.history[-1] }}
-{% endif -%}
-{{ current_goal }}
-<|CLOSE_SECTION|>
-{{ bot_token }}Director instructs story writer:

src/talemate/prompts/templates/director/old/goal-analyze.jinja2

@@ -1,8 +0,0 @@
-
-{% for scene_context in scene.context_history(budget=max_tokens-300) -%}
-{{ scene_context }}
-{% endfor %}
-
-Question: Do any lines or events in the dialogue satisfy the following story condition: "{{ current_goal }}" - Explain your reasoning and then state 'satisfied' or 'NOT been satisfied'.
-
-{{ bot_token }}Director decides: The condition has

src/talemate/prompts/templates/narrator/narrate-after-dialogue.jinja2

@@ -0,0 +1,19 @@
+{% block rendered_context -%}
+<|SECTION:CONTEXT|>
+Content Context: This is a specific scene from {{ scene.context }}
+Scenario Premise: {{ scene.description }}
+{% for memory in query_memory(last_line, as_question_answer=False, iterate=10) -%}
+{{ memory }}
+
+{% endfor %}
+{% endblock -%}
+<|CLOSE_SECTION|>
+{% for scene_context in scene.context_history(budget=max_tokens-200-count_tokens(self.rendered_context())) -%}
+{{ scene_context }}
+{% endfor %}
+<|SECTION:TASK|>
+Based on the previous line '{{ last_line }}', create the next line of narration. This line should focus solely on describing sensory details (like sounds, sights, smells, tactile sensations) or external actions that move the story forward. Avoid including any character's internal thoughts, feelings, or dialogue. Your narration should directly respond to '{{ last_line }}', either by elaborating on the immediate scene or by subtly advancing the plot. Generate exactly one sentence of new narration. If the character is trying to determine some state, truth or situation, try to answer as part of the narration.
+
+Be creative and generate something new and interesting.
+<|CLOSE_SECTION|>
+{{ set_prepared_response('*') }}

src/talemate/prompts/templates/narrator/narrate-query.jinja2

@@ -8,13 +8,13 @@
 {% if query.endswith("?") -%}
 Question: {{ query }}
 Extra context: {{ query_memory(query, as_question_answer=False) }}
-Instruction: Analyze Context, History and Dialogue. Be factual and truthful. When evaluating both story and memory, story is more important. You can fill in gaps using imagination as long as it is based on the existing context. Respect the scene progression and answer in the context of the end of the dialogue.
+Instruction: Analyze Context, History and Dialogue. When evaluating both story and memory, story is more important. You can fill in gaps using imagination as long as it is based on the existing context. Respect the scene progression and answer in the context of the end of the dialogue.
 {% else -%}
 Instruction: {{ query }}
 Extra context: {{ query_memory(query, as_question_answer=False) }}
-Answer based on Context, History and Dialogue. Be factual and truthful. When evaluating both story and memory, story is more important. You can fill in gaps using imagination as long as it is based on the existing context. 
+Answer based on Context, History and Dialogue. When evaluating both story and memory, story is more important. You can fill in gaps using imagination as long as it is based on the existing context. 
 {% endif -%}
 Content Context: This is a specific scene from {{ scene.context }}
-Narration style: point and click adventure game from the 90s
+Your answer should be in the style of short narration that fits the context of the scene.
 <|CLOSE_SECTION|>
 Narrator answers: {% if at_the_end %}{{ bot_token }}At the end of the dialogue, {% endif %}

src/talemate/prompts/templates/narrator/narrate-scene.jinja2

@@ -1,15 +1,12 @@
-<|SECTION:CONTEXT|>
-Scenario Premise: {{ scene.description }}
-<|CLOSE_SECTION|>
 
 {% for scene_context in scene.context_history(budget=max_tokens-300) -%}
 {{ scene_context }}
 {% endfor %}
-
-<|SECTION:TASK|>
-Question: What happens at the end of the dialogue progression? Summarize into narrative description.
+<|SECTION:CONTEXT|>
 Content Context: This is a specific scene from {{ scene.context }}
-Narration style: point and click adventure game from the 90s
-Expected Answer: A summarized narrative description of the scene unfolding at the dialogue that can be inserted into the ongoing story in place of the dialogue.
+Scenario Premise: {{ scene.description }}
 <|CLOSE_SECTION|>
-Narrator answers: {{ set_prepared_response("You see ") }}
+<|SECTION:TASK|>
+Provide a visual description of what is currently happening in the scene. Don't progress the scene.
+<|CLOSE_SECTION|>
+{{ bot_token }}At the end of the scene we currently see:

src/talemate/prompts/templates/narrator/narrate-time-passage.jinja2

@@ -0,0 +1,16 @@
+<|SECTION:CONTEXT|>
+Scenario Premise: {{ scene.description }}
+NPCs: {{ scene.npc_character_names }}
+Player Character: {{ scene.get_player_character().name }}
+Content Context: {{ scene.context }}
+<|CLOSE_SECTION|>
+
+{% for scene_context in scene.context_history(budget=max_tokens-300) -%}
+{{ scene_context }}
+{% endfor %}
+
+<|SECTION:TASK|>
+Narrate the passage of time that just occured, subtly move the story forward, and set up the next scene.
+Write 1 to 3 sentences.
+<|CLOSE_SECTION|>
+{{ bot_token }}{{ narrative }}:

src/talemate/prompts/templates/summarizer/summarize-dialogue.jinja2

@@ -6,8 +6,4 @@
 Question: What happens within the dialogue? Summarize into narrative description.
 Content Context: This is a specific scene from {{ scene.context }}
 Expected Answer: A summarized narrative description of the dialogue that can be inserted into the ongoing story in place of the dialogue. 
-
-Include implied time skips (for example characters plan to meet at a later date and then they meet).
-
-<|CLOSE_SECTION|>
-Narrator answers: 
+<|CLOSE_SECTION|>

src/talemate/prompts/templates/world_state/analyze-text-and-answer-question.jinja2

@@ -1,3 +1,4 @@
+
 {{ text }}
 
 <|SECTION:TASK|>

src/talemate/prompts/templates/world_state/analyze-text-and-extract-context.jinja2

@@ -0,0 +1,17 @@
+
+<|SECTION:CONTEXT|>
+{% for memory in query_memory(text, as_question_answer=False, max_tokens=max_tokens-500, iterate=20) -%}
+{{ memory }}
+
+{% endfor -%}
+<|CLOSE_SECTION|>
+<|SECTION:TASK|>
+Answer the following questions:
+
+{{ instruct_text("Ask the narrator three (3) questions to gather more context from the past for the continuation of this conversation. If a character is asking about a state, location or information about an item or another character, make sure to include question(s) that help gather context for this.", text) }}
+
+You answers should be precise, truthful and short. Pay close attention to timestamps when retrieving information from the context.
+
+<|CLOSE_SECTION|>
+<|SECTION:RELEVANT CONTEXT|>
+{{ bot_token }}Answers: 

src/talemate/prompts/templates/world_state/analyze-text-and-follow-instruction.jinja2

@@ -0,0 +1,5 @@
+
+{{ text }}
+
+<|SECTION:TASK|>
+{{ instruction }}

src/talemate/prompts/templates/world_state/match-character-names.jinja2

@@ -0,0 +1,30 @@
+<|SECTION:CHARACTERS|>
+Player / main character:
+- {{ scene.get_player_character().name }}
+Other characters:
+{% for name in scene.npc_character_names -%}
+- {{ name }}
+{% endfor -%}
+<|CLOSE_SECTION|>
+<|SECTION:TASK|>
+Match the following character aliases to the existing characters.
+
+Respond in the following JSON format:
+
+{
+    "matched_names": [
+        {
+            "alias": "alias", # given alias name for the task
+            "matched_name": "character name" # name of the character
+        }
+    ]
+}
+
+If the name cannot be matched to a character, skip it
+<|CLOSE_SECTION|>
+<|SECTION:ALIASES|>
+{% for name in names -%}
+- {{ name }}
+{% endfor -%}
+<|CLOSE_SECTION|>
+{{ set_json_response(dict(matched_names=[""])) }}

src/talemate/prompts/templates/world_state/request-world-state-inline-items.jinja2

@@ -1,11 +1,57 @@
-Instructions: Mark all tangible physical subjects in the sentence with brackets. For example, if the line of dialogue is "John: I am going to the store." and you want to mark "store" as a subject, you would write "John: I am going to [the  store]."
+<|SECTION:JSON SCHEMA|>
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "type": "object",
+  "properties": {
+    "characters": {
+      "type": "object",
+      "additionalProperties": {
+        "type": "object",
+        "properties": {
+          "snapshot": {
+            # describe the character's current state in the scene
+            "type": "string"
+          },
+          "emotion": {
+            # simple, one word e.g., "happy", "sad", "angry", "confused", "scared" etc.,
+            "type": "string"
+          }
+        },
+        "required": ["snapshot", "emotion"]
+      }
+    },
+    "items": {
+      "type": "object",
+      "additionalProperties": {
+        "type": "object",
+        "properties": {
+          "snapshot": {
+            # describe the item's current state in the scene
+            "type": "string"
+          }
+        },
+        "required": ["snapshot"]
+      }
+    },
+    "location": {
+      # where is the scene taking place?
+      "type": "string"
+    }
+  },
+  "required": ["characters", "items", "location"]
+}
+<|CLOSE_SECTION|>
+<|SECTION:LAST KNOWN WORLD STATE|>
+{{ scene.world_state.pretty_json }}
+<|CLOSE_SECTION|>
+<|SECTION:SCENE PROGRESS|>
+{% for scene_context in scene.context_history(budget=300, min_dialogue=5, add_archieved_history=False, max_dialogue=5) -%}
+{{ scene_context }}
+{% endfor -%}
+<|CLOSE_SECTION|>
+<|SECTION:TASK|>
+Update the existing JSON object for the world state to reflect the changes in the scene progression.
 
-Sentence:
-Barbara: *Barabara sits down on the couch while John is watching TV* Lets see whats on *She takes the remote and starts flipping through channels. She occasionally snaps her wristband while she does it* 
-
-Sentence with tangible physical objects marked:
-Barbara: *Barabara sits down on [the couch] while John is watching [TV]* Lets see whats on *She takes [the remote] and starts flipping through [channels]. She occasionally snaps [her wristband] while she does it*
-
-Sentence:
-{{ scene.history[-1] }}
-Sentence with tangible physical objects marked::{{ bot_token }}
+Objects that are no longer explicitly mentioned in the scene progression should be removed from the JSON object.
+<|CLOSE_SECTION|>
+<|SECTION:UPDATED WORLD STATE|>{{ set_json_response(dict(characters={"name":{}}), cutoff=1) }}

src/talemate/prompts/templates/world_state/request-world-state-v2.jinja2

@@ -0,0 +1,56 @@
+<|SECTION:EXAMPLE|>
+{
+    "characters": {
+        # the character name is the key
+        "Character name": {
+            "emotion": "The current emotional state or mood of the character. (neutral, happy, sad, angry, etc.)",
+            "snapshot": "A brief narrative description of what the character is doing at this moment in the scene."
+        },
+        # ...
+    },
+    "items": {
+        # the item name is the key in natural language (short)
+        "Item name": {
+            "snapshot": "A brief narrative description of the item and the state its currently in."
+        },
+        # ...
+    },
+    "location": "A brief narrative description of the location the scene is taking place in.",
+}
+<|CLOSE_SECTION|>
+<|SECTION:CONTEXT|>
+Player character: {{ scene.get_player_character().name }}
+Other major characters:
+{% for npc_name in scene.npc_character_names -%}
+{{ npc_name }}
+{% endfor -%}
+
+{% for scene_context in scene.context_history(budget=1000, min_dialogue=10, dialogue_negative_offset=5, sections=False) -%}
+{{ scene_context }}
+{% endfor -%}
+{% if not scene.history -%}
+<|SECTION:DIALOGUE|>
+No dialogue so far
+{% endif -%}
+<|CLOSE_SECTION|>
+<|SECTION:SCENE PROGRESS|>
+{% for scene_context in scene.context_history(budget=500, min_dialogue=5, add_archieved_history=False, max_dialogue=5) -%}
+{{ scene_context }}
+{% endfor -%}
+<|CLOSE_SECTION|>
+<|SECTION:TASK|>
+Create a JSON object for the world state that reflects the scene progression so far.
+
+The world state needs to include important concrete and material items present at the very end of the dialogue.
+The world state needs to include persons (characters) interacting at the very end of the dialogue
+Be factual and truthful. Don't make up things that are not in the context or dialogue.
+Snapshot text should always be specified. If you don't know what to write, write "You see nothing special."
+Emotion should always be specified. If you don't know what to write, write "neutral".
+
+Required response: a complete and valid JSON response according to the JSON example containing items and characters.
+
+characters should have the following attributes: `emotion`, `snapshot`
+items should have the following attributes: `snapshot`
+<|CLOSE_SECTION|>
+<|SECTION:UPDATED WORLD STATE|>
+{{ set_json_response(dict(characters={"name":{}}), cutoff=3) }}

src/talemate/prompts/templates/world_state/request-world-state.jinja2

@@ -1,24 +1,22 @@
-<|SECTION:CONTEXT EXAMPLE|>
-Barbara visited her borther John. 
-<|CLOSE_SECTION|>
-
-<|SECTION:DIALOGUE EXAMPLE|>
-Barbara: *Barbara accidently poured some yoghurt on her shirt*
-John: I love filming myself *Holds up his phone to film himself* I dont mind that the screen is cracked!
-Barbara: I should change this shirt but i dont want to get up from the couch
-<|CLOSE_SECTION|>
-
-<|SECTION:WORLD STATE EXAMPLE|>
+<|SECTION:WORLD STATE SCHEMA|>
 {
-    "items": [
-        {"name": "Barbara's red shirt", "snapshot": "The shirt has a big stain on it"},
-        {"name": "John's fanncy phone", "snapshot": "The screen is cracked"}
-    ],
     "characters": [
-        {"name": "John", "emotion": "Excited", "snapshot": "John is filming himself on his phone next to his sister"},
-        {"name": "Barbara", "emotion": "Calm", "snapshot": "Barbara is sitting on the couch"}
+        {
+            "name": "The name of the character involved in the scene.",
+            "emotion": "The current emotional state or mood of the character.",
+            "snapshot": "A brief description of what the character is doing at this moment in the scene."
+        },
+        # ...
+    ],
+    "items": [
+        {
+            "name": "The name of an item that belongs to one of the characters.",
+            "snapshot": "A brief description of the item's current condition or any notable features."
+        },
+        # ...
     ]
 }
+
 <|CLOSE_SECTION|>
 <|SECTION:CONTEXT|>
 {% for scene_context in scene.context_history(budget=1000, min_dialogue=10, dialogue_negative_offset=5, sections=False) -%}
@@ -46,10 +44,10 @@ Required response: a complete and valid JSON response according to the JSON exam
 characters should habe the following attributes: `name`, `emotion`, `snapshot`
 items should have the following attributes: `name`, `snapshot`
 
-Don't copy the example, write your own descriptions.
+You must not copy the example, write your own descriptions.
 <|CLOSE_SECTION|>
 {% for scene_context in scene.context_history(budget=300, min_dialogue=5, add_archieved_history=False, max_dialogue=5) -%}
 {{ scene_context }}
 {% endfor -%}
 <|SECTION:WORLD STATE|>
-{{ set_json_response(dict(items=[""])) }}
+{{ set_json_response(dict(characters=[{"name":scene.character_names[0]}])) }}

src/talemate/server/run.py

@@ -1,3 +1,5 @@
+import os
+
 import argparse
 import asyncio
 import sys

src/talemate/server/websocket_server.py

@@ -63,8 +63,8 @@ class WebsocketHandler(Receiver):
         abort_wait_for_input()
 
         memory_agent = instance.get_agent("memory")
-        if memory_agent:
-            memory_agent.close_db()
+        if memory_agent and self.scene:
+            memory_agent.close_db(self.scene)
 
     def connect_llm_clients(self):
         client = None
@@ -128,6 +128,10 @@ class WebsocketHandler(Receiver):
 
     async def load_scene(self, path_or_data, reset=False, callback=None, file_name=None):
         try:
+            
+            if self.scene:
+                instance.get_agent("memory").close_db(self.scene)
+            
             scene = self.init_scene()
 
             if not scene:
@@ -135,19 +139,10 @@ class WebsocketHandler(Receiver):
                 return
 
             conversation_helper = scene.get_helper("conversation")
-            memory_helper = scene.get_helper("memory")
 
-            await memory_helper.agent.set_db()
-            
             scene = await load_scene(
                 scene, path_or_data, conversation_helper.agent.client, reset=reset
             )
-            #elif isinstance(path_or_data, dict):
-            #    scene = await load_scene_from_data(
-            #        scene, path_or_data, conversation_helper.agent.client, reset=reset
-            #    )
-
-            # Continuously ask the user for input and send it to the actor's talk_to method
 
             self.scene = scene
 
@@ -172,14 +167,14 @@ class WebsocketHandler(Receiver):
         log.info("Configuring clients", clients=clients)
         
         for client in clients:
-            if client["type"] == "textgenwebui":
+            if client["type"] in ["textgenwebui", "lmstudio"]:
                 try:
                     max_token_length = int(client.get("max_token_length", 2048))
                 except ValueError:
                     continue
 
                 self.llm_clients[client["name"]] = {
-                    "type": "textgenwebui",
+                    "type": client["type"],
                     "api_url": client["apiUrl"],
                     "name": client["name"],
                     "max_token_length": max_token_length,
@@ -281,12 +276,20 @@ class WebsocketHandler(Receiver):
         )
         
     def handle_director(self, emission: Emission):
+        
+        if emission.character:
+            character = emission.character.name
+        elif emission.message_object.source:
+            character = emission.message_object.source
+        else:
+            character = ""
+        
         self.queue_put(
             {
                 "type": "director",
                 "message": emission.message,
                 "id": emission.id,
-                "character": emission.character.name if emission.character else "",
+                "character": character,
             }
         )
 
@@ -382,7 +385,7 @@ class WebsocketHandler(Receiver):
                 "status": emission.status,
                 "data": emission.data,
                 "max_token_length": client.max_token_length if client else 2048,
-                "apiUrl": getattr(client, "api_url_base", None) if client else None,
+                "apiUrl": getattr(client, "api_url", None) if client else None,
             }
         )
 

src/talemate/tale_mate.py

@@ -6,6 +6,8 @@ import random
 import traceback
 import re
 import isodate
+import uuid
+import time
 from typing import Dict, List, Optional, Union
 
 from blinker import signal
@@ -41,6 +43,10 @@ __all__ = [
 
 log = structlog.get_logger("talemate")
 
+async_signals.register("game_loop_start")
+async_signals.register("game_loop")
+async_signals.register("game_loop_actor_iter")
+
 
 class Character:
     """
@@ -520,8 +526,7 @@ class Player(Actor):
             emit("character", self.history[-1], character=self.character)
             
         return message
-    
-async_signals.register("game_loop")
+
 
 class Scene(Emitter):
     """
@@ -548,6 +553,8 @@ class Scene(Emitter):
         
         self.name = ""
         self.filename = ""
+        self.memory_id = str(uuid.uuid4())[:10]
+        self.saved = False
         
         self.context = ""
         self.commands = commands.Manager(self)
@@ -569,6 +576,8 @@ class Scene(Emitter):
             "archive_add": signal("archive_add"),
             "character_state": signal("character_state"),
             "game_loop": async_signals.get("game_loop"),
+            "game_loop_start": async_signals.get("game_loop_start"),
+            "game_loop_actor_iter": async_signals.get("game_loop_actor_iter"),
         }
 
         self.setup_emitter(scene=self)
@@ -584,6 +593,10 @@ class Scene(Emitter):
     def character_names(self):
         return [character.name for character in self.characters]
 
+    @property
+    def npc_character_names(self):
+        return [character.name for character in self.get_npc_characters()]
+
     @property
     def log(self):
         return log
@@ -933,11 +946,12 @@ class Scene(Emitter):
         reserved_min_archived_history_tokens = count_tokens(self.archived_history[-1]["text"]) if self.archived_history else 0
         reserved_intro_tokens = count_tokens(self.get_intro()) if show_intro else 0
             
-        max_dialogue_budget = min(max(budget - reserved_intro_tokens - reserved_min_archived_history_tokens, 1000), budget)
+        max_dialogue_budget = min(max(budget - reserved_intro_tokens - reserved_min_archived_history_tokens, 500), budget)
         
         dialogue_popped = False
         while count_tokens(dialogue) > max_dialogue_budget:
             dialogue.pop(0)
+            
             dialogue_popped = True
 
         if dialogue:
@@ -949,7 +963,7 @@ class Scene(Emitter):
             context_history = [context_history[1]]
             
         # we only have room for dialogue, so we return it
-        if dialogue_popped:
+        if dialogue_popped and max_dialogue_budget >= budget:
             return context_history
 
         # if we dont have lots of archived history, we can also include the scene
@@ -983,7 +997,6 @@ class Scene(Emitter):
         i = len(self.archived_history) - 1
         limit = 5
         
-        
         if sections:
             context_history.insert(archive_insert_idx, "<|CLOSE_SECTION|>")
         
@@ -998,6 +1011,7 @@ class Scene(Emitter):
             text = self.archived_history[i]["text"]
             if count_tokens(context_history) + count_tokens(text) > budget:
                 break
+            
             context_history.insert(archive_insert_idx, text)
             i -= 1
             limit -= 1
@@ -1055,8 +1069,15 @@ class Scene(Emitter):
             new_message = await narrator.agent.narrate_character(character)
         elif source == "narrate_query":
             new_message = await narrator.agent.narrate_query(arg)
+        elif source == "narrate_dialogue":
+            character = self.get_character(arg)
+            new_message = await narrator.agent.narrate_after_dialogue(character)
         else:
-            return
+            fn = getattr(narrator.agent, source, None)
+            if not fn:
+                return
+            args = arg.split(";") if arg else []
+            new_message = await fn(*args)
         
         save_source = f"{source}:{arg}" if arg else source
         
@@ -1085,8 +1106,7 @@ class Scene(Emitter):
         
         director = self.get_helper("director")
         
-        response = await director.agent.direct(character)
-
+        response = await director.agent.direct_scene(character)
         if not response:
             log.info("Director returned no response")
             return
@@ -1143,7 +1163,7 @@ class Scene(Emitter):
                 break
 
     def emit_status(self):
-       emit(
+        emit(
             "scene_status",
             self.name,
             status="started",
@@ -1153,8 +1173,11 @@ class Scene(Emitter):
                 "assets": self.assets.dict(),
                 "characters": [actor.character.serialize for actor in self.actors],
                 "scene_time": util.iso8601_duration_to_human(self.ts, suffix="") if self.ts else None,
+                "saved": self.saved,
             },
-        )    
+        )
+    
+        self.log.debug("scene_status", scene=self.name, scene_time=self.ts, saved=self.saved)
 
     def set_environment(self, environment: str):
         """
@@ -1177,11 +1200,20 @@ class Scene(Emitter):
         Loops through self.history looking for TimePassageMessage and will
         advance the world state by the amount of time passed for each
         """
-        
         # reset time
-        
         self.ts = "PT0S"
         
+        # archived history (if "ts" is set) should provide the base line
+        # find the first archived_history entry from the back that has a ts
+        # and set that as the base line
+        
+        if self.archived_history:
+            for i in range(len(self.archived_history) - 1, -1, -1):
+                if self.archived_history[i].get("ts"):
+                    self.ts = self.archived_history[i]["ts"]
+                    break
+        
+        
         for message in self.history:
             if isinstance(message, TimePassageMessage):
                 self.advance_time(message.ts)
@@ -1283,6 +1315,8 @@ class Scene(Emitter):
         self.active_actor = None
         self.next_actor = None
         
+        await self.signals["game_loop_start"].send(events.GameLoopStartEvent(scene=self, event_type="game_loop_start"))
+        
         while continue_scene:
             
             try:
@@ -1310,8 +1344,14 @@ class Scene(Emitter):
                         if await command.execute(message):
                             break
                         await self.call_automated_actions()
+                        
+                        await self.signals["game_loop_actor_iter"].send(
+                            events.GameLoopActorIterEvent(scene=self, event_type="game_loop_actor_iter", actor=actor)
+                        )
                         continue
                     
+                    self.saved = False
+                    
                     # Store the most recent AI Actor
                     self.most_recent_ai_actor = actor
 
@@ -1319,6 +1359,13 @@ class Scene(Emitter):
                         emit(
                             "character", item, character=actor.character
                         )
+                    
+                    await self.signals["game_loop_actor_iter"].send(
+                        events.GameLoopActorIterEvent(scene=self, event_type="game_loop_actor_iter", actor=actor)
+                    )
+                
+                self.emit_status()
+                    
             except TalemateInterrupt:
                 raise
             except LLMAccuracyError as e:
@@ -1349,6 +1396,10 @@ class Scene(Emitter):
                     continue
                 
                 await command.execute(message)
+                
+                self.saved = False
+                self.emit_status()
+                
             except TalemateInterrupt:
                 raise
             except LLMAccuracyError as e:
@@ -1375,13 +1426,15 @@ class Scene(Emitter):
         
         return saves_dir
             
-    async def save(self):
+    async def save(self, save_as:bool=False):
         """
         Saves the scene data, conversation history, archived history, and characters to a json file.
         """
         scene = self
-
-
+        
+        if save_as:
+            self.filename = None
+            
         if not self.name:
             self.name = await wait_for_input("Enter scenario name: ")
             self.filename = "base.json"
@@ -1389,6 +1442,13 @@ class Scene(Emitter):
         elif not self.filename:
             self.filename = await wait_for_input("Enter save name: ")
             self.filename = self.filename.replace(" ", "-").lower()+".json"
+            
+        if save_as:
+            memory_agent = self.get_helper("memory").agent
+            memory_agent.close_db(self)
+            self.memory_id = str(uuid.uuid4())[:10]
+            await memory_agent.set_db()
+            await self.commit_to_memory()
 
         saves_dir = self.save_dir
         
@@ -1412,6 +1472,7 @@ class Scene(Emitter):
             "context": scene.context,
             "world_state": scene.world_state.dict(),
             "assets": scene.assets.dict(),
+            "memory_id": scene.memory_id,
             "ts": scene.ts,
         }
 
@@ -1419,8 +1480,35 @@ class Scene(Emitter):
 
         with open(filepath, "w") as f:
             json.dump(scene_data, f, indent=2, cls=save.SceneEncoder)
+            
+        self.saved = True
+        self.emit_status()
 
-        await asyncio.sleep(0)
+    async def commit_to_memory(self):
+        
+        # will recommit scene to long term memory
+        
+        memory = self.get_helper("memory").agent
+        memory.drop_db()
+        await memory.set_db()
+        
+        for ah in self.archived_history:
+            ts = ah.get("ts", "PT1S")
+                
+            if not ah.get("ts"):
+                ah["ts"] = ts
+            
+            self.signals["archive_add"].send(
+                events.ArchiveEvent(scene=self, event_type="archive_add", text=ah["text"], ts=ts)
+            )
+            await asyncio.sleep(0)
+
+        for character_name, cs in self.character_states.items():
+            self.set_character_state(character_name, cs)
+
+        for character in self.characters:
+            await character.commit_to_memory(memory)
+        
 
     def reset(self):
         self.history = []

src/talemate/util.py

@@ -303,6 +303,9 @@ def strip_partial_sentences(text:str) -> str:
     # Sentence ending characters
     sentence_endings = ['.', '!', '?', '"', "*"]
     
+    if not text:
+        return text
+    
     # Check if the last character is already a sentence ending
     if text[-1] in sentence_endings:
         return text
@@ -572,7 +575,7 @@ def iso8601_duration_to_human(iso_duration, suffix:str=" ago"):
     elif components:
         human_str = components[0]
     else:
-        human_str = "0 Seconds"
+        human_str = "Moments"
     
     return f"{human_str}{suffix}"
 
@@ -766,99 +769,135 @@ def replace_exposition_markers(s:str) -> str:
 
 def ensure_dialog_format(line:str, talking_character:str=None) -> str:
     
-    line = mark_exposition(line, talking_character)
-    line = mark_spoken_words(line, talking_character)
+    #if "*" not in line and '"' not in line:
+    #    if talking_character:
+    #        line = line[len(talking_character)+1:].lstrip()
+    #        return f"{talking_character}: \"{line}\""
+    #    return f"\"{line}\""
+    #
+
+    if talking_character:
+        line = line[len(talking_character)+1:].lstrip()
+
+    lines = []
+
+    for _line in line.split("\n"):
+        try:
+            _line = ensure_dialog_line_format(_line)
+        except Exception as exc:
+            log.error("ensure_dialog_format", msg="Error ensuring dialog line format", line=_line, exc_info=exc)
+            pass
+    
+        lines.append(_line)
+        
+    if len(lines) > 1:
+        line = "\n".join(lines)
+    else:
+        line = lines[0]
+    
+    if talking_character:
+        line = f"{talking_character}: {line}"
+    
     return line
     
 
-def mark_spoken_words(line:str, talking_character:str=None) -> str:
-    # if there are no asterisks in the line, it means its impossible to tell
-    # dialogue apart from exposition
-    if "*" not in line:
-        return line
-
-    if talking_character and line.startswith(f"{talking_character}:"):
-        line = line[len(talking_character)+1:].lstrip()
+def ensure_dialog_line_format(line:str):
     
-
-    # Splitting the text into segments based on asterisks
-    segments = re.split('(\*[^*]*\*)', line)
-    formatted_line = ""
-
-    for i, segment in enumerate(segments):
-        if segment.startswith("*") and segment.endswith("*"):
-            # If the segment is an action or thought, add it as is
-            formatted_line += segment
-        else:
-            # For non-action/thought parts, trim and add quotes only if not empty and not already quoted
-            trimmed_segment = segment.strip()
-            if trimmed_segment:
-                if not (trimmed_segment.startswith('"') and trimmed_segment.endswith('"')):
-                    formatted_line += f' "{trimmed_segment}"'
-                else:
-                    formatted_line += f' {trimmed_segment}'
-
-
-    # adds spaces betwen *" and "* to make it easier to read
-    formatted_line = formatted_line.replace('*"', '* "')
-    formatted_line = formatted_line.replace('"*', '" *')
-
-    if talking_character:
-        formatted_line = f"{talking_character}: {formatted_line}"
+    """
+    a Python function that standardizes the formatting of dialogue and action/thought 
+    descriptions in text strings. This function is intended for use in a text-based 
+    game where spoken dialogue is encased in double quotes (" ") and actions/thoughts are
+    encased in asterisks (* *). The function must correctly format strings, ensuring that
+    each spoken sentence and action/thought is properly encased
+    """
         
-    log.debug("mark_spoken_words", line=line, formatted_line=formatted_line)
-
-    return formatted_line.strip()  # Trim any leading/trailing whitespace
-
-
-def mark_exposition(line:str, talking_character:str=None) -> str:
-    """
-    Will loop through the string and make sure chunks outside of "" are marked with *.
     
-    For example:
+    i = 0
     
-    "No, you're not wrong" sips his wine "This tastes gross." coughs "acquired taste i guess?"
+    segments = []
+    segment = None
+    segment_open = None
     
-    becomes
-    
-    "No, you're not wrong" *sips his wine* "This tastes gross." *coughs* "acquired taste i guess?"
-    """
-    
-    # no quotes in string, means its impossible to tell dialogue apart from exposition
-    if '"' not in line:
-        return line
-    
-    if talking_character and line.startswith(f"{talking_character}:"):
-        line = line[len(talking_character)+1:].lstrip()
-    
-    # Splitting the text into segments based on quotes
-    segments = re.split('("[^"]*")', line)
-    formatted_line = ""
-
-    for i, segment in enumerate(segments):
-        # If the segment is a spoken part (inside quotes), add it as is
-        if segment.startswith('"') and segment.endswith('"'):
-            formatted_line += segment
+    for i in range(len(line)):
+        
+        
+        c = line[i]
+        
+        #print("segment_open", segment_open)
+        #print("segment", segment)
+        
+        if c in ['"', '*']:
+            if segment_open == c:
+                # open segment is the same as the current character
+                # closing
+                segment_open = None
+                segment += c
+                segments += [segment.strip()]
+                segment = None
+            elif segment_open is not None and segment_open != c:
+                # open segment is not the same as the current character
+                # opening - close the current segment and open a new one
+                segments += [segment.strip()]
+                segment_open = c
+                segment = c
+            elif segment_open is None:
+                # we're opening a segment
+                segment_open = c
+                segment = c
         else:
-            # Split the non-spoken segment into sub-segments based on existing asterisks
-            sub_segments = re.split('(\*[^*]*\*)', segment)
-            for sub_segment in sub_segments:
-                if sub_segment.startswith("*") and sub_segment.endswith("*"):
-                    # If the sub-segment is already formatted, add it as is
-                    formatted_line += sub_segment
-                else:
-                    # Trim and add asterisks only to non-empty sub-segments
-                    trimmed_sub_segment = sub_segment.strip()
-                    if trimmed_sub_segment:
-                        formatted_line += f" *{trimmed_sub_segment}*"
-
-    # adds spaces betwen *" and "* to make it easier to read
-    formatted_line = formatted_line.replace('*"', '* "')
-    formatted_line = formatted_line.replace('"*', '" *')
-
-    if talking_character:
-        formatted_line = f"{talking_character}: {formatted_line}"
-    log.debug("mark_exposition", line=line, formatted_line=formatted_line)
+            if segment_open is None:
+                segment_open = "unclassified"
+                segment = c
+            else:
+                segment += c
+                
+    if segment is not None:
+        segments += [segment.strip()]
+        
+    for i in range(len(segments)):
+        segment = segments[i]
+        if segment in ['"', '*']:
+            if i > 0:
+                prev_segment = segments[i-1]
+                if prev_segment[-1] not in ['"', '*']:
+                    segments[i-1] = f"{prev_segment}{segment}"
+                    segments[i] = ""
+                    continue
     
-    
-    return formatted_line.strip()  # Trim any leading/trailing whitespace
+    for i in range(len(segments)):
+        segment = segments[i]
+        
+        if not segment:
+            continue
+                
+        if segment[0] == "*" and segment[-1] != "*":
+            segment += "*"
+        elif segment[-1] == "*" and segment[0] != "*":
+            segment = "*" + segment
+        elif segment[0] == '"' and segment[-1] != '"':
+            segment += '"'
+        elif segment[-1] == '"' and segment[0] != '"':
+            segment = '"' + segment
+        elif segment[0] in ['"', '*'] and segment[-1] == segment[0]:
+            continue
+        
+        segments[i] = segment
+        
+    for i in range(len(segments)):
+        segment = segments[i]
+        if not segment or segment[0] in ['"', '*']:
+            continue
+        
+        prev_segment = segments[i-1] if i > 0 else None
+        next_segment = segments[i+1] if i < len(segments)-1 else None
+        
+        if prev_segment and prev_segment[-1] == '"':
+            segments[i] = f"*{segment}*"
+        elif prev_segment and prev_segment[-1] == '*':
+            segments[i] = f"\"{segment}\""
+        elif next_segment and next_segment[0] == '"':
+            segments[i] = f"*{segment}*"
+        elif next_segment and next_segment[0] == '*':
+            segments[i] = f"\"{segment}\""
+            
+    return " ".join(segment for segment in segments if segment)

src/talemate/world_state.py

@@ -38,14 +38,17 @@ class WorldState(BaseModel):
     @property
     def as_list(self):
         return self.render().as_list
-        
+    
+    def reset(self):
+        self.characters = {}
+        self.items = {}
+        self.location = None    
     
     def emit(self, status="update"):
         emit("world_state", status=status, data=self.dict())
         
     async def request_update(self, initial_only:bool=False):
-        
-        
+
         if initial_only and self.characters:
             self.emit()
             return
@@ -58,19 +61,94 @@ class WorldState(BaseModel):
             self.emit()
             raise e
         
+        previous_characters = self.characters
+        previous_items = self.items
+        scene = self.agent.scene
+        character_names = scene.character_names
         self.characters = {}
         self.items = {}
         
-        for character in world_state.get("characters", []):
-            self.characters[character["name"]] = CharacterState(**character)
+        for character_name, character in world_state.get("characters", {}).items():
+            
+            # character name may not always come back exactly as we have
+            # it defined in the scene. We assign the correct name by checking occurences
+            # of both names in each other.
+            
+            if character_name not in character_names:
+                for _character_name in character_names:
+                    if _character_name.lower() in character_name.lower() or character_name.lower() in _character_name.lower():
+                        log.debug("world_state adjusting character name", from_name=character_name, to_name=_character_name)
+                        character_name = _character_name
+                        break
+            
+            if not character:
+                continue
+            
+            # if emotion is not set, see if a previous state exists
+            # and use that emotion
+            
+            if "emotion" not in character:
+                log.debug("emotion not set", character_name=character_name, character=character, characters=previous_characters)
+                if character_name in previous_characters:
+                    character["emotion"] = previous_characters[character_name].emotion
+            
+            self.characters[character_name] = CharacterState(**character)
             log.debug("world_state", character=character)
             
-        for item in world_state.get("items", []):
-            self.items[item["name"]] = ObjectState(**item)
+        for item_name, item in world_state.get("items", {}).items():
+            if not item:
+                continue
+            self.items[item_name] = ObjectState(**item)
             log.debug("world_state", item=item)
         
-        self.emit()
         
+        await self.persist()
+        self.emit()
+     
+    async def persist(self):
+        
+        memory = instance.get_agent("memory")
+        world_state = instance.get_agent("world_state")
+        
+        # first we check if any of the characters were refered
+        # to with an alias
+        
+        states = []
+        scene = self.agent.scene
+
+        for character_name in self.characters.keys():
+            states.append(
+                {
+                    "text": f"{character_name}: {self.characters[character_name].snapshot}",
+                    "id": f"{character_name}.world_state.snapshot",
+                    "meta": {
+                        "typ": "world_state",
+                        "character": character_name,
+                        "ts": scene.ts,
+                    }
+                }
+            )
+        
+        for item_name in self.items.keys():
+            states.append(
+                {
+                    "text": f"{item_name}: {self.items[item_name].snapshot}",
+                    "id": f"{item_name}.world_state.snapshot",
+                    "meta": {
+                        "typ": "world_state",
+                        "item": item_name,
+                        "ts": scene.ts,
+                    }
+                }
+            )
+        
+        log.debug("world_state.persist", states=states)
+        
+        if not states:
+            return
+
+        await memory.add_many(states)            
+               
     
     async def request_update_inline(self):
         

talemate_frontend/src/components/AIAgent.vue

@@ -91,7 +91,6 @@ export default {
         openModal() {
             this.state.formTitle = 'Add AI Agent';
             this.state.dialog = true;
-            console.log("got here")
         },
         saveAgent(agent) {
             const index = this.state.agents.findIndex(c => c.name === agent.name);
@@ -120,7 +119,6 @@ export default {
         handleMessage(data) {
             // Handle agent_status message type
             if (data.type === 'agent_status') {
-                console.log("agents: got agent_status message", data)
                 // Find the client with the given name
                 const agent = this.state.agents.find(agent => agent.name === data.name);
                 if (agent) {

talemate_frontend/src/components/AIClient.vue

@@ -26,7 +26,7 @@
             hide-details
             v-model="client.max_token_length"
             :min="1024"
-            :max="16384"
+            :max="128000"
             :step="512"
             @update:modelValue="saveClient(client)"
             @click.stop
@@ -120,7 +120,7 @@ export default {
       this.state.currentClient = {
         name: 'TextGenWebUI',
         type: 'textgenwebui',
-        apiUrl: 'http://localhost:5000/api',
+        apiUrl: 'http://localhost:5000',
         model_name: '',
         max_token_length: 4096,
       };

talemate_frontend/src/components/AgentModal.vue

@@ -17,7 +17,7 @@
 
 
       </v-card-title>
-      <v-card-text>
+      <v-card-text class="scrollable-content">
         <v-select v-model="agent.client" :items="agent.data.client" label="Client"></v-select>
 
         <v-alert type="warning" variant="tonal" density="compact" v-if="agent.data.experimental">
@@ -32,10 +32,12 @@
           <v-card-text>
               {{ agent.data.actions[key].description }}
               <div v-for="(action_config, config_key) in agent.data.actions[key].config" :key="config_key">
+                <div v-if="action.enabled">
                 <!-- render config widgets based on action_config.type (int, str, bool, float) -->
-                <v-text-field v-if="action_config.type === 'str'" v-model="action.config[config_key].value" :label="action_config.label" :hint="action_config.description" density="compact"></v-text-field>
+                <v-text-field v-if="action_config.type === 'text'" v-model="action.config[config_key].value" :label="action_config.label" :hint="action_config.description" density="compact"></v-text-field>
                 <v-slider v-if="action_config.type === 'number' && action_config.step !== null" v-model="action.config[config_key].value" :label="action_config.label" :hint="action_config.description" :min="action_config.min" :max="action_config.max" :step="action_config.step" density="compact" thumb-label></v-slider>
                 <v-checkbox v-if="action_config.type === 'bool'" v-model="action.config[config_key].value" :label="action_config.label" :hint="action_config.description" density="compact"></v-checkbox>
+                </div>
               </div>
           </v-card-text>
         </v-card>
@@ -97,4 +99,12 @@ export default {
     }
   }
 }
-</script>
+</script>
+
+<style>
+.scrollable-content {
+  overflow-y: auto;
+  max-height: 70vh;
+  padding-right: 16px;
+}
+</style>

talemate_frontend/src/components/ClientModal.vue

@@ -8,7 +8,7 @@
             <v-container>
             <v-row>
                 <v-col cols="6">
-                  <v-select v-model="client.type" :items="['openai', 'textgenwebui']" label="Client Type"></v-select>
+                  <v-select v-model="client.type" :items="['openai', 'textgenwebui', 'lmstudio']" label="Client Type"></v-select>
                 </v-col>
                 <v-col cols="6">
                   <v-text-field v-model="client.name" label="Client Name"></v-text-field>
@@ -17,13 +17,13 @@
             </v-row>
             <v-row>
               <v-col cols="12">
-                <v-text-field v-model="client.apiUrl" v-if="client.type === 'textgenwebui'" label="API URL"></v-text-field>
-                <v-select v-model="client.model" v-if="client.type === 'openai'" :items="['gpt-4', 'gpt-3.5-turbo', 'gpt-3.5-turbo-16k']" label="Model"></v-select>
+                <v-text-field v-model="client.apiUrl" v-if="isLocalApiClient(client)" label="API URL"></v-text-field>
+                <v-select v-model="client.model" v-if="client.type === 'openai'" :items="['gpt-4-1106-preview', 'gpt-4', 'gpt-3.5-turbo', 'gpt-3.5-turbo-16k']" label="Model"></v-select>
               </v-col>
             </v-row>  
             <v-row>
               <v-col cols="6">
-                <v-text-field v-model="client.max_token_length" v-if="client.type === 'textgenwebui'" type="number" label="Context Length"></v-text-field> 
+                <v-text-field v-model="client.max_token_length" v-if="isLocalApiClient(client)" type="number" label="Context Length"></v-text-field> 
               </v-col>
             </v-row>
             </v-container>
@@ -74,6 +74,9 @@ export default {
     save() {
       this.$emit('save', this.client); // Emit save event with client object
       this.close();
+    },
+    isLocalApiClient(client) {
+      return client.type === 'textgenwebui' || client.type === 'lmstudio';
     }
   }
 }

talemate_frontend/src/components/DebugToolPromptLog.vue

@@ -1,6 +1,7 @@
 <template>
     <v-list-subheader class="text-uppercase"><v-icon>mdi-post-outline</v-icon> Prompts
         <v-chip size="x-small" color="primary">{{ max_prompts }}</v-chip>
+        <v-icon color="primary" class="ml-2" @click="clearPrompts">mdi-close</v-icon>
     </v-list-subheader>
 
     <v-list-item density="compact">
@@ -9,15 +10,19 @@
 
     <v-list-item v-for="(prompt, index) in prompts" :key="index" @click="openPromptView(prompt)">
         <v-list-item-title class="text-caption">
-            {{ prompt.kind }}
+
+            <v-row>
+                <v-col cols="2" class="text-info">#{{ prompt.num }}</v-col>
+                <v-col cols="10" class="text-right">{{ prompt.kind }}</v-col>
+            </v-row>
+
         </v-list-item-title>
         <v-list-item-subtitle>
-            <v-chip size="x-small"><v-icon size="14"
-            class="mr-1">mdi-pound</v-icon>{{ prompt.num }}</v-chip>
-            <v-chip size="x-small" color="primary">{{ prompt.prompt_tokens }}<v-icon size="14"
+            <v-chip size="x-small" class="mr-1" color="primary">{{ prompt.prompt_tokens }}<v-icon size="14"
             class="ml-1">mdi-arrow-down-bold</v-icon></v-chip>
-            <v-chip size="x-small" color="secondary">{{ prompt.response_tokens }}<v-icon size="14"
+            <v-chip size="x-small" class="mr-1" color="secondary">{{ prompt.response_tokens }}<v-icon size="14"
             class="ml-1">mdi-arrow-up-bold</v-icon></v-chip>
+            <v-chip size="x-small">{{ prompt.time }}s<v-icon size="14" class="ml-1">mdi-clock</v-icon></v-chip>
         </v-list-item-subtitle>
         <v-divider class="mt-1"></v-divider>
     </v-list-item>
@@ -33,7 +38,7 @@ export default {
     data() {
         return {
             prompts: [],
-            total: 0,
+            total: 1,
             max_prompts: 50,
         }
     },
@@ -47,6 +52,10 @@ export default {
     ],
 
     methods: {
+        clearPrompts() {
+            this.prompts = [];
+            this.total = 0;
+        },
         handleMessage(data) {
 
             if(data.type === "system"&& data.id === "scene.loaded") {
@@ -63,6 +72,7 @@ export default {
                     kind: data.data.kind,
                     response_tokens: data.data.response_tokens,
                     prompt_tokens: data.data.prompt_tokens,
+                    time: parseInt(data.data.time),
                     num: this.total++,
                 })
 

talemate_frontend/src/components/DirectorMessage.vue

@@ -1,13 +1,15 @@
 <template>
   <div class="director-container" v-if="show && minimized" >
-    <v-chip closable @click:close="deleteMessage()" color="deep-purple-lighten-3">
+    <v-chip closable color="deep-orange" class="clickable" @click:close="deleteMessage()">
       <v-icon class="mr-2">mdi-bullhorn-outline</v-icon>
       <span @click="toggle()">{{ character }}</span>
     </v-chip>
   </div>
-  <v-alert v-else-if="show" class="director-message" variant="text" :closable="message_id !== null" type="info" icon="mdi-bullhorn-outline"
+  <v-alert v-else-if="show" color="deep-orange" class="director-message clickable" variant="text" type="info" icon="mdi-bullhorn-outline"
     elevation="0" density="compact" @click:close="deleteMessage()" >
-    <div class="director-text" @click="toggle()">{{ text }}</div>
+    <span class="director-instructs" @click="toggle()">{{ directorInstructs }}</span>
+    <span class="director-character ml-1 text-decoration-underline" @click="toggle()">{{ directorCharacter }}</span>
+    <span class="director-text ml-1" @click="toggle()">{{ directorText }}</span>
   </v-alert>
 </template>
   
@@ -21,6 +23,17 @@ export default {
   },
   props: ['text', 'message_id', 'character'],
   inject: ['requestDeleteMessage'],
+  computed: {
+    directorInstructs() {
+      return "Director instructs"
+    },
+    directorCharacter() {
+      return this.text.split(':')[0].split("Director instructs ")[1];
+    },
+    directorText() {
+      return this.text.split(':')[1].split('"')[1];
+    }
+  },
   methods: {
     toggle() {
       this.minimized = !this.minimized;
@@ -41,6 +54,10 @@ export default {
   margin-right: 2px;
 }
 
+.clickable {
+  cursor: pointer;
+}
+
 .highlight:before {
   --content: "*";
 }
@@ -50,16 +67,33 @@ export default {
 }
 
 .director-text {
-  color: #9FA8DA;
 }
 
 .director-message {
-  display: flex;
-  flex-direction: row;
   color: #9FA8DA;
 }
 
 .director-container {
   
 }
+
+.director-instructs {
+  /* Add your CSS styles for "Director instructs" here */
+  color: #BF360C;
+}
+
+.director-character {
+  /* Add your CSS styles for the character name here */
+}
+
+.director-text {
+  /* Add your CSS styles for the actual instruction here */
+  color: #EF6C00;
+}
+.director-text::after {
+  content: '"';
+}
+.director-text::before {
+  content: '"';
+}
 </style>

talemate_frontend/src/components/LoadScene.vue

@@ -58,6 +58,7 @@ export default {
             scenes: [],
             sceneSearchInput: null,
             sceneSearchLoading: false,
+            sceneSaved: null,
             expanded: true,
         }
     },
@@ -83,6 +84,13 @@ export default {
             this.getWebsocket().send(JSON.stringify({ type: 'load_scene', file_path: "environment:creative" }));
         },
         loadScene() {
+
+            if(this.sceneSaved === false) {
+                if(!confirm("The current scene is not saved. Are you sure you want to load a new scene?")) {
+                    return;
+                }
+            }
+
             this.loading = true;
             if (this.inputMethod === 'file' && this.sceneFile.length > 0) { // Check if the input method is "file" and there is at least one file
                 // Convert the uploaded file to base64
@@ -119,6 +127,12 @@ export default {
                 return;
             }
 
+            // Handle scene status
+            if (data.type == "scene_status") {
+                this.sceneSaved = data.data.saved;
+                return;
+            }
+
         }
     },
     created() {

talemate_frontend/src/components/TalemateApp.vue

@@ -89,7 +89,8 @@
       <v-app-bar-nav-icon @click="toggleNavigation('game')"><v-icon>mdi-script</v-icon></v-app-bar-nav-icon>
       <v-toolbar-title v-if="scene.name !== undefined">
         {{ scene.name || 'Untitled Scenario' }}
-        <v-chip size="x-small" v-if="scene.environment === 'creative'" class="ml-1"><v-icon text="Creative" size="14"
+        <span v-if="scene.saved === false" class="text-red">*</span>
+        <v-chip size="x-small" v-if="scene.environment === 'creative'" class="ml-2"><v-icon text="Creative" size="14"
             class="mr-1">mdi-palette-outline</v-icon>Creative Mode</v-chip>
         <v-chip size="x-small" v-else-if="scene.environment === 'scene'" class="ml-1"><v-icon text="Play" size="14"
             class="mr-1">mdi-gamepad-square</v-icon>Game Mode</v-chip>
@@ -244,8 +245,10 @@ export default {
       }
 
       this.connecting = true;
+      let currentUrl = new URL(window.location.href);
+      console.log(currentUrl);
 
-      this.websocket = new WebSocket('ws://localhost:5050/ws');
+      this.websocket = new WebSocket(`ws://${currentUrl.hostname}:5050/ws`);
       console.log("Websocket connecting ...")
       this.websocket.onmessage = this.handleMessage;
       this.websocket.onopen = () => {
@@ -300,6 +303,7 @@ export default {
           name: data.name,
           environment: data.data.environment,
           scene_time: data.data.scene_time,
+          saved: data.data.saved,
         }
         this.sceneActive = true;
         return;

talemate_frontend/src/components/WorldState.vue

@@ -12,7 +12,7 @@
             <v-expansion-panel rounded="0" density="compact">
                 <v-expansion-panel-title class="text-subtitle-2" diable-icon-rotate>
                     {{ name }}
-                    <v-chip label size="x-small" variant="outlined" class="ml-1">{{ character.emotion }}</v-chip>
+                    <v-chip v-if="character.emotion !== null && character.emotion !== ''" label size="x-small" variant="outlined" class="ml-1">{{ character.emotion }}</v-chip>
                     <template v-slot:actions>
                         <v-icon icon="mdi-account"></v-icon>
                     </template>

templates/llm-prompt/Cat.jinja2

@@ -0,0 +1,4 @@
+{{ system_message }}
+
+### Instruction:
+{{ set_response(prompt, "\n\n### Response:\n") }}

templates/llm-prompt/Nous-Capybara.jinja2

@@ -0,0 +1,3 @@
+USER:
+{{ system_message }} 
+{{ set_response(prompt, "\nASSISTANT:") }}

templates/llm-prompt/Psyfighter2.jinja2

@@ -0,0 +1,4 @@
+{{ system_message }}
+
+### Instruction:
+{{ set_response(prompt, "\n\n### Response:\n") }}

templates/llm-prompt/Tess-Medium.jinja2

@@ -0,0 +1,2 @@
+SYSTEM: {{ system_message }}
+USER: {{ set_response(prompt, "\nASSISTANT: ") }}

templates/llm-prompt/UtopiaXL.jinja2

@@ -0,0 +1,4 @@
+{{ system_message }}
+
+### Instruction:
+{{ set_response(prompt, "\n\n### Response:\n") }}

templates/llm-prompt/Yi.jinja2

@@ -0,0 +1 @@
+User: {{ system_message }} {{ set_response(prompt, "\nAssistant: ") }}

templates/llm-prompt/dolphin-2.2.1-mistral.jinja2

@@ -0,0 +1,4 @@
+<|im_start|>system
+{{ system_message }}<|im_end|>
+<|im_start|>user
+{{ set_response(prompt, "<|im_end|>\n<|im_start|>assistant\n") }}

templates/llm-prompt/dolphin-2_2-yi.jinja2

@@ -0,0 +1,4 @@
+<|im_start|>system
+{{ system_message }}<|im_end|>
+<|im_start|>user
+{{ set_response(prompt, "<|im_end|>\n<|im_start|>assistant\n") }}

update.bat

@@ -6,5 +6,5 @@ call talemate_env\Scripts\activate
 REM use poetry to install dependencies
 python -m poetry install
 
-echo Virtual environment re-created.
+echo Virtual environment updated
 pause