fix: have proxy in a dedicated package (avoid dependency on server deps)

Author: samuelint

langchain_llamacpp_chat_model/__init__.py

@@ -1,4 +1,3 @@
 from .llama_chat_model import LlamaChatModel
-from .llama_proxy_chat_model import LlamaProxyChatModel
 
-__all__ = ["LlamaChatModel", "LlamaProxyChatModel"]
+__all__ = ["LlamaChatModel"]

langchain_llamacpp_chat_model/proxy/__init__.py

@@ -0,0 +1,3 @@
+from .llama_proxy_chat_model import LlamaProxyChatModel
+
+__all__ = ["LlamaProxyChatModel"]

langchain_llamacpp_chat_model/llama_proxy_chat_model.py langchain_llamacpp_chat_model/proxy/llama_proxy_chat_model.py

@@ -1,7 +1,7 @@
 import pytest
 from langchain_core.messages import AIMessage, HumanMessage
 
-from langchain_llamacpp_chat_model import LlamaProxyChatModel
+from langchain_llamacpp_chat_model.proxy import LlamaProxyChatModel
 
 from tests.test_functional.models_configuration import (
     create_llama_proxy,

tests/test_functional/test_llama_proxy.py

@@ -16,7 +16,7 @@ def llama(self, request) -> Llama:
 
     @pytest.fixture
     def instance(self, llama):
-        return LlamaChatModel(llama=llama)
+        return LlamaChatModel(llama=llama, temperature=0)
 
     def test_stream(self, instance: LlamaChatModel):
         stream = instance.stream("Say Hi!")
@@ -50,4 +50,4 @@ def test_conversation_memory(self, instance: LlamaChatModel):
             final_content += token.content
 
         assert len(final_content) > 0
-        assert "banana" in final_content
+        assert "banana" in final_content.lower()