feat(azure): quivr compatible with it (#3005)

# Description

Please include a summary of the changes and the related issue. Please
also include relevant motivation and context.

## Checklist before requesting a review

Please delete options that are not relevant.

- [ ] My code follows the style guidelines of this project
- [ ] I have performed a self-review of my code
- [ ] I have commented hard-to-understand areas
- [ ] I have ideally added tests that prove my fix is effective or that
my feature works
- [ ] New and existing unit tests pass locally with my changes
- [ ] Any dependent changes have been merged

## Screenshots (if appropriate):

backend/api/quivr_api/models/settings.py

@@ -1,10 +1,11 @@
 from typing import Optional
+from urllib.parse import urlparse
 from uuid import UUID
 
 from langchain.embeddings.base import Embeddings
 from langchain_community.embeddings.ollama import OllamaEmbeddings
 from langchain_community.vectorstores.supabase import SupabaseVectorStore
-from langchain_openai import OpenAIEmbeddings
+from langchain_openai import AzureOpenAIEmbeddings, OpenAIEmbeddings
 from posthog import Posthog
 from pydantic_settings import BaseSettings, SettingsConfigDict
 from sqlalchemy import Engine, create_engine
@@ -116,6 +117,7 @@ class PostHogSettings(BaseSettings):
 class BrainSettings(BaseSettings):
     model_config = SettingsConfigDict(validate_default=False)
     openai_api_key: str = ""
+    azure_openai_embeddings_url: str = ""
     supabase_url: str = ""
     supabase_service_key: str = ""
     resend_api_key: str = "null"
@@ -191,7 +193,22 @@ def get_embedding_client() -> Embeddings:
             base_url=settings.ollama_api_base_url,
         )  # pyright: ignore reportPrivateUsage=none
     else:
-        embeddings = OpenAIEmbeddings()  # pyright: ignore reportPrivateUsage=none
+        if settings.azure_openai_embeddings_url:
+            # https://quivr-test.openai.azure.com/openai/deployments/embedding/embeddings?api-version=2023-05-15
+            # parse the url to get the deployment name
+            deployment = settings.azure_openai_embeddings_url.split("/")[5]
+            netloc = "https://" + urlparse(settings.azure_openai_embeddings_url).netloc
+            api_version = settings.azure_openai_embeddings_url.split("=")[1]
+            logger.debug(f"Using Azure OpenAI embeddings: {deployment}")
+            logger.debug(f"Using Azure OpenAI embeddings: {netloc}")
+            logger.debug(f"Using Azure OpenAI embeddings: {api_version}")
+            embeddings = AzureOpenAIEmbeddings(
+                azure_deployment=deployment,
+                azure_endpoint=netloc,
+                api_version=api_version,
+            )
+        else:
+            embeddings = OpenAIEmbeddings()  # pyright: ignore reportPrivateUsage=none
     return embeddings
 
 

backend/api/quivr_api/modules/sync/utils/upload.py

@@ -3,6 +3,7 @@ from typing import Optional
 from uuid import UUID
 
 from fastapi import HTTPException, UploadFile
+
 from quivr_api.celery_worker import process_file_and_notify
 from quivr_api.logger import get_logger
 from quivr_api.modules.brain.entity.brain_entity import RoleEnum
@@ -74,13 +75,13 @@ async def upload_file(
                 status_code=403,
                 detail=f"File {upload_file.filename} already exists in storage.",
             )
-        
+
         else:
             notification_service.update_notification_by_id(
                 notification_id,
                 NotificationUpdatableProperties(
                     status=NotificationsStatusEnum.ERROR,
-                    description=f"There was an error uploading the file",
+                    description="There was an error uploading the file",
                 ),
             )
             raise HTTPException(

backend/core/quivr_core/llm/llm_endpoint.py

@@ -1,9 +1,15 @@
+import logging
+from urllib.parse import parse_qs, urlparse
+
 from langchain_core.language_models.chat_models import BaseChatModel
 from pydantic.v1 import SecretStr
+
 from quivr_core.brain.info import LLMInfo
 from quivr_core.config import LLMEndpointConfig
 from quivr_core.utils import model_supports_function_calling
 
+logger = logging.getLogger("quivr_core")
+
 
 class LLMEndpoint:
     def __init__(self, llm_config: LLMEndpointConfig, llm: BaseChatModel):
@@ -19,13 +25,30 @@ class LLMEndpoint:
     @classmethod
     def from_config(cls, config: LLMEndpointConfig = LLMEndpointConfig()):
         try:
-            from langchain_openai import ChatOpenAI
+            from langchain_openai import AzureChatOpenAI, ChatOpenAI
 
-            _llm = ChatOpenAI(
+            if config.model.startswith("azure/"):
-                model=config.model,
+                # Parse the URL
-                api_key=SecretStr(config.llm_api_key) if config.llm_api_key else None,
+                parsed_url = urlparse(config.llm_base_url)
-                base_url=config.llm_base_url,
+                deployment = parsed_url.path.split("/")[3]  # type: ignore
-            )
+                api_version = parse_qs(parsed_url.query).get("api-version", [None])[0]  # type: ignore
+                azure_endpoint = f"https://{parsed_url.netloc}"  # type: ignore
+                _llm = AzureChatOpenAI(
+                    azure_deployment=deployment,  # type: ignore
+                    api_version=api_version,
+                    api_key=SecretStr(config.llm_api_key)
+                    if config.llm_api_key
+                    else None,
+                    azure_endpoint=azure_endpoint,
+                )
+            else:
+                _llm = ChatOpenAI(
+                    model=config.model,
+                    api_key=SecretStr(config.llm_api_key)
+                    if config.llm_api_key
+                    else None,
+                    base_url=config.llm_base_url,
+                )
             return cls(llm=_llm, llm_config=config)
 
         except ImportError as e: