diff --git a/open_notebook/graphs/chat.py b/open_notebook/graphs/chat.py
index 0c0e517..e403ace 100644
--- a/open_notebook/graphs/chat.py
+++ b/open_notebook/graphs/chat.py
@@ -11,7 +11,7 @@ from typing_extensions import TypedDict
 
 from open_notebook.config import LANGGRAPH_CHECKPOINT_FILE
 from open_notebook.domain.notebook import Notebook
-from open_notebook.graphs.utils import provision_model
+from open_notebook.graphs.utils import provision_langchain_model
 from open_notebook.prompter import Prompter
 
 
@@ -25,7 +25,7 @@ class ThreadState(TypedDict):
 def call_model_with_messages(state: ThreadState, config: RunnableConfig) -> dict:
     system_prompt = Prompter(prompt_template="chat").render(data=state)
     payload = [system_prompt] + state.get("messages", [])
-    model = provision_model(str(payload), config, "chat")
+    model = provision_langchain_model(str(payload), config, "chat")
     ai_message = model.invoke(payload)
     return {"messages": ai_message}
 
diff --git a/open_notebook/graphs/utils.py b/open_notebook/graphs/utils.py
index 3f84f3b..a3874e8 100644
--- a/open_notebook/graphs/utils.py
+++ b/open_notebook/graphs/utils.py
@@ -1,13 +1,14 @@
-from langchain.output_parsers import OutputFixingParser
-from langchain_core.messages import AIMessage
+from langchain_core.language_models.chat_models import BaseChatModel
+from langchain_core.messages import BaseMessage
 from loguru import logger
 
 from open_notebook.domain.models import model_manager
+from open_notebook.models.llms import LanguageModel
 from open_notebook.prompter import Prompter
 from open_notebook.utils import token_count
 
 
-def provision_model(content, config, default_type):
+def provision_langchain_model(content, config, default_type) -> BaseChatModel:
     """
     Returns the best model to use based on the context size and on whether there is a specific model being requested in Config.
     If context > 105_000, returns the large_context_model
@@ -20,13 +21,14 @@ def provision_model(content, config, default_type):
         logger.debug(
             f"Using large context model because the content has {tokens} tokens"
         )
-        return model_manager.get_default_model("large_context").to_langchain()
+        model = model_manager.get_default_model("large_context")
     elif config.get("configurable", {}).get("model_id"):
-        return model_manager.get_model(
-            config.get("configurable", {}).get("model_id")
-        ).to_langchain()
+        model = model_manager.get_model(config.get("configurable", {}).get("model_id"))
     else:
-        return model_manager.get_default_model(default_type).to_langchain()
+        model = model_manager.get_default_model(default_type)
+
+    assert isinstance(model, LanguageModel), f"Model is not a LanguageModel: {model}"
+    return model.to_langchain()
 
 
 # todo: turn into a graph
@@ -36,23 +38,12 @@ def run_pattern(
     messages=[],
     state: dict = {},
     parser=None,
-    output_fixing_model_id=None,
-) -> AIMessage:
+) -> BaseMessage:
     system_prompt = Prompter(prompt_template=pattern_name, parser=parser).render(
         data=state
     )
     payload = [system_prompt] + messages
-    chain = provision_model(str(payload), config, "transformation")
-
-    if parser:
-        chain = chain | parser
-
-    if output_fixing_model_id and parser:
-        output_fix_model = model_manager.get_model(output_fixing_model_id)
-        chain = chain | OutputFixingParser.from_llm(
-            parser=parser,
-            llm=output_fix_model,
-        )
+    chain = provision_langchain_model(str(payload), config, "transformation")
 
     response = chain.invoke(payload)
 
diff --git a/open_notebook/utils.py b/open_notebook/utils.py
index 340762e..86479e2 100644
--- a/open_notebook/utils.py
+++ b/open_notebook/utils.py
@@ -28,7 +28,7 @@ def split_text(txt: str, chunk=1000, overlap=0, separator=" "):
     return text_splitter.split_text(txt)
 
 
-def token_count(input_string):
+def token_count(input_string) -> int:
     """
     Count the number of tokens in the input string using the 'o200k_base' encoding.
 
@@ -46,7 +46,7 @@ def token_count(input_string):
     return token_count
 
 
-def token_cost(token_count, cost_per_million=0.150):
+def token_cost(token_count, cost_per_million=0.150) -> float:
     """
     Calculate the cost of tokens based on the token count and cost per million tokens.
 
@@ -60,11 +60,11 @@ def token_cost(token_count, cost_per_million=0.150):
     return cost_per_million * (token_count / 1_000_000)
 
 
-def remove_non_ascii(text):
+def remove_non_ascii(text) -> str:
     return re.sub(r"[^\x00-\x7F]+", "", text)
 
 
-def remove_non_printable(text):
+def remove_non_printable(text) -> str:
     # Remove control characters, except newlines and tabs
     text = "".join(
         char for char in text if unicodedata.category(char)[0] != "C" or char in "\n\t"
@@ -74,7 +74,7 @@ def remove_non_printable(text):
     return re.sub(r"[^\w\s.,!?\-\n\t]", "", text, flags=re.UNICODE)
 
 
-def surreal_clean(text):
+def surreal_clean(text) -> str:
     """
     Clean the input text by removing non-ASCII and non-printable characters,
     and adjusting colon placement for SurrealDB compatibility.