Bump up dependencies to the newest versions

allms/models/abstract.py

@@ -5,6 +5,7 @@
 import urllib
 from abc import ABC, abstractmethod
 from functools import partial
+from urllib.error import URLError
 
 import google
 import openai
@@ -77,8 +78,8 @@ def __init__(
 
         self._predict_example = create_base_retry_decorator(
             error_types=[
-                openai.error.RateLimitError, openai.error.APIError, openai.error.Timeout,
-                openai.error.APIConnectionError, openai.error.ServiceUnavailableError,
+                openai.RateLimitError, openai.APIError, openai.Timeout,
+                openai.APIConnectionError, openai.InternalServerError,
                 google.api_core.exceptions.ResourceExhausted, urllib.error.HTTPError
             ],
             max_retries=max_retries,
@@ -252,16 +253,17 @@ async def _predict_example(
                     model_response = await chain.arun({})
                 else:
                     model_response = await chain.arun(**input_data.input_mappings)
-        except openai.error.InvalidRequestError as invalid_request_error:
+        except openai.InternalServerError as invalid_request_error:
             logger.info(f"Error for id {input_data.id} has occurred. Message: {invalid_request_error} ")
-            if invalid_request_error.error.code == "content_filter":
+            if invalid_request_error.code == "content_filter":
                 model_response = None
                 error_message = f"{IODataConstants.CONTENT_FILTER_MESSAGE}: {invalid_request_error}"
             else:
                 model_response = None
                 error_message = f"{IODataConstants.ERROR_MESSAGE_STR}: {invalid_request_error}"
 
-        except (InvalidArgument, ValueError, TimeoutError, openai.error.Timeout, GCPInvalidRequestError) as other_error:
+        except (InvalidArgument, ValueError, TimeoutError, openai.APIError, GCPInvalidRequestError,
+                openai.APITimeoutError) as other_error:
             model_response = None
             logger.info(f"Error for id {input_data.id} has occurred. Message: {other_error} ")
             error_message = f"{type(other_error).__name__}: {other_error}"

allms/models/azure_openai.py

@@ -1,7 +1,7 @@
 from asyncio import AbstractEventLoop
 from typing import Optional
 
-from langchain.chat_models import AzureChatOpenAI
+from langchain_openai import AzureChatOpenAI
 
 from allms.defaults.azure_defaults import AzureGptTurboDefaults
 from allms.defaults.general_defaults import GeneralDefaults
@@ -38,7 +38,7 @@ def _create_llm(self) -> AzureChatOpenAI:
             deployment_name=self._config.deployment,
             api_version=self._config.api_version,
             model_name=self._config.model_name,
-            base_url=self._config.base_url,
+            azure_endpoint=self._config.base_url,
             api_key=self._config.api_key,
             azure_ad_token=self._config.azure_ad_token,
             temperature=self._temperature,

allms/models/vertexai_base.py

@@ -1,9 +1,9 @@
 from typing import List, Optional, Any, Dict
 
 from google.cloud.aiplatform.models import Prediction
-from langchain_google_vertexai import VertexAI, VertexAIModelGarden
 from langchain_core.callbacks import AsyncCallbackManagerForLLMRun
 from langchain_core.outputs import LLMResult, Generation
+from langchain_google_vertexai import VertexAI, VertexAIModelGarden
 from pydash import chain
 
 from allms.constants.vertex_ai import VertexModelConstants

allms/models/vertexai_gemini.py

@@ -47,7 +47,7 @@ def __init__(
             event_loop=event_loop
         )
 
-    def _create_llm(self) -> VertexAI:
+    def _create_llm(self) -> CustomVertexAI:
         return CustomVertexAI(
             model_name=self._config.gemini_model_name,
             max_output_tokens=self._max_output_tokens,

allms/utils/response_parsing_utils.py

@@ -3,6 +3,7 @@
 
 from langchain.output_parsers import PydanticOutputParser
 from langchain.schema import OutputParserException
+from pydantic import ValidationError
 
 from allms.domain.response import ResponseData, ResponseParsingOutput
 
@@ -45,6 +46,15 @@ def _parse_response(
                     The exception message: {output_parser_exception}
                     """
             )
+        except ValidationError as validation_error:
+            return ResponseParsingOutput(
+                response=None,
+                error_message=f"""
+                    A ValidationError has occurred for the model response: {model_response_data.response}
+                    The exception message: {validation_error}
+                    """
+            )
+
 
     def parse_model_output(
         self, 

pyproject.toml

@@ -7,19 +7,21 @@ readme = "README.md"
 packages = [{include = "allms"}]
 
 [tool.poetry.dependencies]
-python = ">=3.8.1,<4.0"
+python = ">=3.9.0,<4.0"
 fsspec = "^2023.6.0"
-google-cloud-aiplatform = ">=1.57.0"
+google-cloud-aiplatform = "1.70.0"
 pydash = "^7.0.6"
 transformers = "^4.34.1"
-pydantic = "1.10.13"
-langchain = "^0.1.8"
-langchain-google-vertexai = "1.0.4"
-aioresponses = "^0.7.6"
-tiktoken = "^0.6.0"
-openai = "^0.27.8"
+pydantic = "2.7.4"
+langchain = "0.3.6"
+tiktoken = "^0.7.0"
+openai = "1.52.0"
 pytest-mock = "^3.14.0"
+respx = "^0.21.1"
+langchain-community = "^0.3.5"
+langchain-google-vertexai = "^2.0.7"
 sentencepiece = "^0.2.0"
+langchain-openai = "^0.2.5"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^7.4.0"

tests/conftest.py

@@ -5,7 +5,6 @@
 from unittest.mock import patch
 
 import pytest
-from aioresponses import aioresponses
 from langchain_community.llms.fake import FakeListLLM
 
 from allms.domain.configuration import (
@@ -94,9 +93,3 @@ def models():
                     event_loop=event_loop
                 )
             }
-
-
-@pytest.fixture
-def mock_aioresponse():
-    with aioresponses() as http_mock:
-        yield http_mock