fix: LangChain breaking changes and deprecations

Author: lpm0073

Makefile

@@ -19,8 +19,8 @@ PINECONE_INDEX_NAME=rag\n\
 PINECONE_VECTORSTORE_TEXT_KEY=lc_id\n\
 PINECONE_METRIC=dotproduct\n\
 PINECONE_DIMENSIONS=1536\n\
-OPENAI_CHAT_MODEL_NAME=gpt-3.5-turbo\n\
-OPENAI_PROMPT_MODEL_NAME=gpt-3.5-turbo-instruct\n\
+OPENAI_CHAT_MODEL_NAME=gpt-4\n\
+OPENAI_PROMPT_MODEL_NAME=gpt-4\n\
 OPENAI_CHAT_TEMPERATURE=0.0\n\
 OPENAI_CHAT_MAX_RETRIES=3\n\
 DEBUG_MODE=True\n" >> .env)

README.md

@@ -134,9 +134,9 @@ Set these as environment variables on the command line, or in a .env file that s
 OPENAI_API_ORGANIZATION=PLEASE-ADD-ME
 OPENAI_API_KEY=PLEASE-ADD-ME
 OPENAI_CHAT_MAX_RETRIES=3
-OPENAI_CHAT_MODEL_NAME=gpt-3.5-turbo
+OPENAI_CHAT_MODEL_NAME=gpt-4
 OPENAI_CHAT_TEMPERATURE=0.0
-OPENAI_PROMPT_MODEL_NAME=gpt-3.5-turbo-instruct
+OPENAI_PROMPT_MODEL_NAME=gpt-4
 
 # Pinecone API
 PINECONE_API_KEY=PLEASE-ADD-ME

models/conf.py

@@ -76,7 +76,7 @@ class SettingsDefaults:
 
     PINECONE_API_KEY: SecretStr = SecretStr(None)
     PINECONE_ENVIRONMENT = "gcp-starter"
-    PINECONE_INDEX_NAME = "rag"
+    PINECONE_INDEX_NAME = "openai-embeddings"
     PINECONE_VECTORSTORE_TEXT_KEY = "lc_id"
     PINECONE_METRIC = "dotproduct"
     PINECONE_DIMENSIONS = 1536
@@ -86,8 +86,8 @@ class SettingsDefaults:
     OPENAI_ENDPOINT_IMAGE_N = 4
     OPENAI_ENDPOINT_IMAGE_SIZE = "1024x768"
     OPENAI_CHAT_CACHE = True
-    OPENAI_CHAT_MODEL_NAME = "gpt-3.5-turbo"
-    OPENAI_PROMPT_MODEL_NAME = "gpt-3.5-turbo-instruct"
+    OPENAI_CHAT_MODEL_NAME = "gpt-4"
+    OPENAI_PROMPT_MODEL_NAME = "gpt-4"
     OPENAI_CHAT_TEMPERATURE = 0.0
     OPENAI_CHAT_MAX_RETRIES = 3
 

models/hybrid_search_retreiver.py

@@ -36,7 +36,7 @@
 
 # from langchain_community.chat_models import ChatOpenAI
 # prompting and chat
-from langchain_openai import ChatOpenAI, OpenAI
+from langchain_openai import ChatOpenAI
 from pinecone_text.sparse import BM25Encoder  # pylint: disable=import-error
 
 # this project
@@ -114,17 +114,13 @@ def cached_chat_request(
         retval = self.chat.invoke(messages)
         return retval
 
+    # pylint: disable=unused-argument
     def prompt_with_template(
         self, prompt: PromptTemplate, concept: str, model: str = settings.openai_prompt_model_name
     ) -> str:
         """Prompt with template."""
-        llm = OpenAI(
-            model=model,
-            api_key=settings.openai_api_key.get_secret_value(),  # pylint: disable=no-member
-            organization=settings.openai_api_organization,
-        )
-        retval = llm(prompt.format(concept=concept))
-        return retval
+        retval = self.chat.invoke(prompt.format(concept=concept))
+        return retval.content if retval else "no response"
 
     def load(self, filepath: str):
         """Pdf loader."""

models/pinecone.py

@@ -106,6 +106,8 @@ def openai_embeddings(self) -> OpenAIEmbeddings:
     def pinecone(self):
         """Pinecone lazy read-only property."""
         if self._pinecone is None:
+            print("Initializing Pinecone...")
+            print(f"API Key: {settings.pinecone_api_key.get_secret_value()}")
             self._pinecone = Pinecone(api_key=settings.pinecone_api_key.get_secret_value())
         return self._pinecone
 
@@ -153,7 +155,7 @@ def create(self):
         print("Creating index. This may take a few minutes...")
         serverless_spec = ServerlessSpec(
             cloud="aws",
-            region="us-west-2",
+            region="us-east-1",
         )
         try:
             self.pinecone.create_index(

models/tests/mock_data/.env.test_01

@@ -9,7 +9,7 @@
     OPENAI_ENDPOINT_IMAGE_N = 1
     OPENAI_ENDPOINT_IMAGE_SIZE = "TEST_1024x768"
     OPENAI_CHAT_CACHE = False
-    OPENAI_CHAT_MODEL_NAME = "TEST_gpt-3.5-turbo"
-    OPENAI_PROMPT_MODEL_NAME = "TEST_gpt-3.5-turbo-instruct"
+    OPENAI_CHAT_MODEL_NAME = "TEST_gpt-4"
+    OPENAI_PROMPT_MODEL_NAME = "TEST_gpt-4"
     OPENAI_CHAT_TEMPERATURE = 1.0
     OPENAI_CHAT_MAX_RETRIES = 5

models/tests/test_configuration.py

@@ -97,7 +97,7 @@ def test_env_overrides(self):
         assert mock_settings.openai_endpoint_image_n == 1
         assert mock_settings.openai_endpoint_image_size == "TEST_1024x768"
         assert mock_settings.openai_chat_cache is False
-        assert mock_settings.openai_chat_model_name == "TEST_gpt-3.5-turbo"
+        assert mock_settings.openai_chat_model_name == "TEST_gpt-4"
         assert mock_settings.openai_prompt_model_name == "TEST_text-davinci-003"
         assert mock_settings.openai_chat_temperature == 1.0
         assert mock_settings.openai_chat_max_retries == 5
@@ -139,7 +139,7 @@ def test_configure_with_class_constructor(self):
             openai_endpoint_image_n=1,
             openai_endpoint_image_size="TEST_1024x768",
             openai_chat_cache=False,
-            openai_chat_model_name="TEST_gpt-3.5-turbo",
+            openai_chat_model_name="TEST_gpt-4",
             openai_prompt_model_name="TEST_text-davinci-003",
             openai_chat_temperature=1.0,
             openai_chat_max_retries=5,
@@ -156,7 +156,7 @@ def test_configure_with_class_constructor(self):
         assert mock_settings.openai_endpoint_image_n == 1
         assert mock_settings.openai_endpoint_image_size == "TEST_1024x768"
         assert mock_settings.openai_chat_cache is False
-        assert mock_settings.openai_chat_model_name == "TEST_gpt-3.5-turbo"
+        assert mock_settings.openai_chat_model_name == "TEST_gpt-4"
         assert mock_settings.openai_prompt_model_name == "TEST_text-davinci-003"
         assert mock_settings.openai_chat_temperature == 1.0
         assert mock_settings.openai_chat_max_retries == 5
@@ -184,7 +184,7 @@ def test_readonly_settings(self):
         with pytest.raises(PydanticValidationError):
             mock_settings.openai_chat_cache = False
         with pytest.raises(PydanticValidationError):
-            mock_settings.openai_chat_model_name = "TEST_gpt-3.5-turbo"
+            mock_settings.openai_chat_model_name = "TEST_gpt-4"
         with pytest.raises(PydanticValidationError):
             mock_settings.openai_prompt_model_name = "TEST_text-davinci-003"
         with pytest.raises(PydanticValidationError):

models/yt.py

@@ -15,23 +15,24 @@
 
 # 5.) sequential chains
 # 4.) chains
-from langchain.chains import LLMChain, SimpleSequentialChain
-
-# 1.) wrappers
-from langchain.llms.openai import OpenAI
+from langchain.chains.llm import LLMChain
+from langchain.chains.sequential import SimpleSequentialChain
 
 # 3.) prompt templates
 from langchain.prompts import PromptTemplate
-from langchain.python import PythonREPL
 
 # 2.) models and messages
 from langchain.schema import HumanMessage, SystemMessage  # AIMessage (not used)
 
 # 6.) embeddings
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 
+# 1.) wrappers
+from langchain_community.llms.openai import OpenAI
+
 # 8.) LangChain agents
 from langchain_experimental.agents.agent_toolkits.python.base import create_python_agent
+from langchain_experimental.utilities.python import PythonREPL
 
 # from langchain_community.chat_models import ChatOpenAI
 from langchain_openai import ChatOpenAI, OpenAIEmbeddings
@@ -77,14 +78,14 @@ class LangChainDev:
     def test_01_basic(self):
         """Test a basic request"""
 
-        llm = OpenAI(model_name="gpt-3.5-turbo-instruct")
+        llm = OpenAI(model_name="gpt-4")
         retval = llm("explain large language models in one sentence")
         print(retval)
 
     # 2.) models and messages. minute 6:08
     def test_02_chat_model(self):
         """Test a chat model"""
-        chat = ChatOpenAI(model_name="gpt-3.5-turbo-instruct", temperature=0.3)
+        chat = ChatOpenAI(model_name="gpt-4", temperature=0.3)
         messages = [
             SystemMessage(content="You are an expert data scientist"),
             HumanMessage(content="Write a Python script that trains a neural network on simulated data"),
@@ -104,7 +105,7 @@ def get_prompt(self):
 
     def test_03_prompt_templates(self):
         """Test prompt templates"""
-        llm = OpenAI(model_name="gpt-3.5-turbo-instruct")
+        llm = OpenAI(model_name="gpt-4")
         prompt = self.get_prompt()
         retval = llm(prompt.format(concept="regularization"))
         print(retval)
@@ -117,7 +118,7 @@ def get_chain(self, llm, prompt):
 
     def test_04_chain(self):
         """Test a chain"""
-        llm = OpenAI(model_name="gpt-3.5-turbo-instruct")
+        llm = OpenAI(model_name="gpt-4")
         prompt = self.get_prompt()
         chain = self.get_chain(llm=llm, prompt=prompt)
         print(chain.run("autoencoder"))
@@ -139,7 +140,7 @@ def get_prompt_two(self):
 
     def get_explanation(self):
         """Get an explanation"""
-        llm = OpenAI(model_name="gpt-3.5-turbo-instruct")
+        llm = OpenAI(model_name="gpt-4")
         prompt = self.get_prompt()
         chain_one = self.get_chain(llm=llm, prompt=prompt)