Fixed splitted lines in tutorials, reworked system prompt handling af…

…ter initialization

chatsky/llm/llm_api.py

@@ -8,9 +8,7 @@
 from pydantic import BaseModel
 import logging
 from chatsky.core.message import Message
-from chatsky.core.context import Context
 from chatsky.llm.methods import BaseMethod
-from chatsky.llm.utils import message_to_langchain
 from chatsky.llm._langchain_imports import StrOutputParser, BaseChatModel, BaseMessage, check_langchain_available
 
 
@@ -32,7 +30,7 @@ def __init__(
         check_langchain_available()
         self.model: BaseChatModel = model
         self.parser = StrOutputParser()
-        self.system_prompt = system_prompt
+        self.system_prompt = Message(text=system_prompt)
 
     async def respond(
         self,

chatsky/llm/utils.py

@@ -28,6 +28,7 @@ async def message_to_langchain(
     check_langchain_available()
     if isinstance(message, str):
         message = Message(text=message)
+
     if message.text is None:
         content = []
     elif len(message.text) > max_size:

chatsky/responses/llm.py

@@ -12,7 +12,7 @@
 from chatsky.core.message import Message
 from chatsky.core.context import Context
 from chatsky.llm.utils import message_to_langchain, context_to_history
-from chatsky.llm._langchain_imports import SystemMessage, check_langchain_available
+from chatsky.llm._langchain_imports import check_langchain_available
 from chatsky.llm.filters import BaseHistoryFilter, DefaultFilter
 from chatsky.core.script_function import BaseResponse, AnyResponse
 
@@ -51,10 +51,10 @@ class LLMResponse(BaseResponse):
     async def call(self, ctx: Context) -> Message:
         check_langchain_available()
         model = ctx.pipeline.models[self.model_name]
-        if model.system_prompt == "":
+        if model.system_prompt.text == "":
             history_messages = []
         else:
-            history_messages = [SystemMessage(model.system_prompt)]
+            history_messages = [await message_to_langchain(model.system_prompt, ctx=ctx, source="system")]
         current_node = ctx.current_node
         current_misc = current_node.misc
         if current_misc is not None:

tutorials/llm/1_basics.py

@@ -39,7 +39,6 @@
 therefore it will store all dialogue history.
 This is not advised if you are short on tokens or
 if you do not need to store all dialogue history.
-Alternatively you can instantiate model object inside
 
 Also note, that langchain reads environment variables for the models
 automatically and you do not necessarily need to set them explicitly.
@@ -58,7 +57,6 @@
 
 As you can see in this script, you can pass an additional prompt to the LLM.
 We will cover that thoroughly in the Prompt usage tutorial.
-of RESPONSE field in the nodes you need.
 """
 
 # %%

tutorials/llm/2_prompt_usage.py

@@ -77,7 +77,7 @@
 
 # %% [markdown]
 """
-Chatsky enables you to use more complex prompts then a simple string in need be.
+Chatsky enables you to use more complex prompts then a simple string if need be.
 In this example we create a VacantPlaces class, that can dynamically retrieve
 some external data and put them into the prompt.