Create Langchain LLM converter to quickly supply it to Langchain's chain (#102)

* Create Langchain LLM converter to quickly supply it to Langchain's chain

* Clean up

knowledgehub/agents/langchain_based.py

@@ -53,9 +53,7 @@ class LangchainAgent(BaseAgent):
             # reinit Langchain AgentExecutor
             self.agent = initialize_agent(
                 langchain_plugins,
-                # TODO: could cause bugs for non-langchain llms
+                self.llm.to_langchain_format(),
-                # related to https://github.com/Cinnamon/kotaemon/issues/73
-                self.llm._obj,  # type: ignore
                 agent=self.AGENT_TYPE_MAP[self.agent_type],
                 handle_parsing_errors=True,
                 verbose=True,

knowledgehub/llms/__init__.py

@@ -1,16 +1,12 @@
-from typing import Union
-
 from kotaemon.base.schema import AIMessage, BaseMessage, HumanMessage, SystemMessage
 
+from .base import BaseLLM
 from .branching import GatedBranchingPipeline, SimpleBranchingPipeline
 from .chats import AzureChatOpenAI, ChatLLM
 from .completions import LLM, AzureOpenAI, OpenAI
 from .linear import GatedLinearPipeline, SimpleLinearPipeline
 from .prompts import BasePromptComponent, PromptTemplate
 
-BaseLLM = Union[ChatLLM, LLM]
-
-
 __all__ = [
     "BaseLLM",
     # chat-specific components

knowledgehub/llms/base.py

@@ -0,0 +1,8 @@
+from langchain_core.language_models.base import BaseLanguageModel
+
+from kotaemon.base import BaseComponent
+
+
+class BaseLLM(BaseComponent):
+    def to_langchain_format(self) -> BaseLanguageModel:
+        raise NotImplementedError

knowledgehub/llms/chats/base.py

@@ -3,11 +3,12 @@ from __future__ import annotations
 import logging
 
 from kotaemon.base import BaseComponent
+from kotaemon.llms.base import BaseLLM
 
 logger = logging.getLogger(__name__)
 
 
-class ChatLLM(BaseComponent):
+class ChatLLM(BaseLLM):
     def flow(self):
         if self.inflow is None:
             raise ValueError("No inflow provided.")

knowledgehub/llms/chats/langchain_based.py

@@ -68,6 +68,9 @@ class LCChatMixin:
             logits=[],
         )
 
+    def to_langchain_format(self):
+        return self._obj
+
     def __repr__(self):
         kwargs = []
         for key, value_obj in self._kwargs.items():

knowledgehub/llms/completions/base.py

@@ -1,5 +1,5 @@
-from kotaemon.base import BaseComponent
+from kotaemon.llms.base import BaseLLM
 
 
-class LLM(BaseComponent):
+class LLM(BaseLLM):
     pass

knowledgehub/llms/completions/langchain_based.py

@@ -45,6 +45,9 @@ class LCCompletionMixin:
             logits=[],
         )
 
+    def to_langchain_format(self):
+        return self._obj
+
     def __repr__(self):
         kwargs = []
         for key, value_obj in self._kwargs.items():

tests/test_agent.py

@@ -189,7 +189,7 @@ def test_react_agent_langchain(openai_completion, llm, mock_google_search):
     langchain_plugins = [tool.to_langchain_format() for tool in plugins]
     agent = initialize_agent(
         langchain_plugins,
-        llm._obj,
+        llm.to_langchain_format(),
         agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
         verbose=True,
     )

tests/test_llms_chat_models.py

@@ -48,7 +48,7 @@ def test_azureopenai_model(openai_completion):
         temperature=0,
     )
     assert isinstance(
-        model._obj, AzureChatOpenAILC
+        model.to_langchain_format(), AzureChatOpenAILC
     ), "Agent not wrapped in Langchain's AzureChatOpenAI"
 
     # test for str input - stream mode

tests/test_llms_completion_models.py

@@ -41,7 +41,7 @@ def test_azureopenai_model(openai_completion):
         request_timeout=60,
     )
     assert isinstance(
-        model._obj, AzureOpenAILC
+        model.to_langchain_format(), AzureOpenAILC
     ), "Agent not wrapped in Langchain's AzureOpenAI"
 
     output = model("hello world")
@@ -64,7 +64,7 @@ def test_openai_model(openai_completion):
         request_timeout=60,
     )
     assert isinstance(
-        model._obj, OpenAILC
+        model.to_langchain_format(), OpenAILC
     ), "Agent is not wrapped in Langchain's OpenAI"
 
     output = model("hello world")