Skip to content

Commit

Permalink
Upgrade OpenAI SDK to v1 (Azure-Samples#1017)
Browse files Browse the repository at this point in the history 
* Update version ranges for 1.3.5 openai lib

* Update the embeddings library in scripts to use OpenAI 1.3.5 remove some redundant methods

* Update the embedding response to use a typed model

* Rewrite test_prepdocs to patch out OpenAI v1 models and responses

* Update approaches to use new APIs

* Update backend service and read approaches to use new SDK

* Fix get_search_query. Update RRR approach tests

* Update search manager tests

* Change patching for app tests

* Use deployment ID only in the constructor of the Azure OpenAI Client object and remove it from the approach constructors (and all the logic that went with it)

* Explicitly include aiohttp in prepdocs requirements

* Use two clients because the new SDK doesn't support a deployment name in the chat or embeddings methods

* Ruff ruff

* Simplify typing constructor

* Update types for message history

* Convert RRR to dict before returning

* Bend the rules of physics to get mypy to pass

* Run black over scripts again

* Fix content filtering, update snapshot tests, implement pydantic models for streaming responses.

* Update the snapshots with the new required fields for chunked completions. Update the iterator to pass pydantic model validation

* Force keyword arguments as the list of arguments is long and complicated

* Refactor to have a single client object

* Drop argument

* Type the chat message builder with pydantic

* Rebuild requirements from merge conflicts

* Update formatting

* Fix issue with follow-up questions

* Simplify content check

* Don't use deployment field for non azure

* Update requirements.in

* Remove upper bound

* Remove dependabot constraint

* Merge the clients again

* Fix test_app client name

* Inline the ternary statement to pick either a model or deployment name for the OpenAI SDK calls

---------

Co-authored-by: Pamela Fox <pamela.fox@gmail.com>
Co-authored-by: Pamela Fox <pamelafox@microsoft.com>
  • Loading branch information
@tonybaloney @pamelafox
3 people authored Dec 4, 2023
1 parent 1f3c876 commit 94f71a2
Show file tree
Hide file tree
Showing 55 changed files with 877 additions and 402 deletions.
1 change: 0 additions & 1 deletion .github/dependabot.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -32,4 +32,3 @@ updates:
- "*"
ignore:
- dependency-name: azure-search-documents
- dependency-name: openai[datalib]
105 changes: 46 additions & 59 deletions app/backend/app.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,19 +3,18 @@
import logging
import mimetypes
import os
import time
from pathlib import Path
from typing import AsyncGenerator

import aiohttp
import openai
from azure.core.exceptions import ResourceNotFoundError
from azure.identity.aio import DefaultAzureCredential
from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
from azure.monitor.opentelemetry import configure_azure_monitor
from azure.search.documents.aio import SearchClient
from azure.storage.blob.aio import BlobServiceClient
from openai import APIError, AsyncAzureOpenAI, AsyncOpenAI
from opentelemetry.instrumentation.aiohttp_client import AioHttpClientInstrumentor
from opentelemetry.instrumentation.asgi import OpenTelemetryMiddleware
from opentelemetry.instrumentation.httpx import HTTPXClientInstrumentor
from quart import (
Blueprint,
Quart,
Expand All @@ -33,13 +32,12 @@
from approaches.retrievethenread import RetrieveThenReadApproach
from core.authentication import AuthenticationHelper

CONFIG_OPENAI_TOKEN = "openai_token"
CONFIG_CREDENTIAL = "azure_credential"
CONFIG_ASK_APPROACH = "ask_approach"
CONFIG_CHAT_APPROACH = "chat_approach"
CONFIG_BLOB_CONTAINER_CLIENT = "blob_container_client"
CONFIG_AUTH_CLIENT = "auth_client"
CONFIG_SEARCH_CLIENT = "search_client"
CONFIG_OPENAI_CLIENT = "openai_client"
ERROR_MESSAGE = """The app encountered an error processing your request.
If you are an administrator of the app, view the full error in the logs. See aka.ms/appservice-logs for more information.
Error type: {error_type}
Expand Down Expand Up @@ -102,14 +100,14 @@ async def content_file(path: str):


def error_dict(error: Exception) -> dict:
if isinstance(error, openai.error.InvalidRequestError) and error.code == "content_filter":
if isinstance(error, APIError) and error.code == "content_filter":
return {"error": ERROR_MESSAGE_FILTER}
return {"error": ERROR_MESSAGE.format(error_type=type(error))}


def error_response(error: Exception, route: str, status_code: int = 500):
logging.exception("Exception in %s: %s", route, error)
if isinstance(error, openai.error.InvalidRequestError) and error.code == "content_filter":
if isinstance(error, APIError) and error.code == "content_filter":
status_code = 400
return jsonify(error_dict(error)), status_code

Expand All @@ -124,12 +122,9 @@ async def ask():
context["auth_claims"] = await auth_helper.get_auth_claims_if_enabled(request.headers)
try:
approach = current_app.config[CONFIG_ASK_APPROACH]
# Workaround for: https://github.com/openai/openai-python/issues/371
async with aiohttp.ClientSession() as s:
openai.aiosession.set(s)
r = await approach.run(
request_json["messages"], context=context, session_state=request_json.get("session_state")
)
r = await approach.run(
request_json["messages"], context=context, session_state=request_json.get("session_state")
)
return jsonify(r)
except Exception as error:
return error_response(error, "/ask")
Expand Down Expand Up @@ -178,19 +173,6 @@ def auth_setup():
return jsonify(auth_helper.get_auth_setup_for_client())


@bp.before_request
async def ensure_openai_token():
if openai.api_type != "azure_ad":
return
openai_token = current_app.config[CONFIG_OPENAI_TOKEN]
if openai_token.expires_on < time.time() + 60:
openai_token = await current_app.config[CONFIG_CREDENTIAL].get_token(
"https://cognitiveservices.azure.com/.default"
)
current_app.config[CONFIG_OPENAI_TOKEN] = openai_token
openai.api_key = openai_token.token


@bp.before_app_serving
async def setup_clients():
# Replace these with your own values, either in environment variables or directly here
Expand All @@ -204,8 +186,8 @@ async def setup_clients():
OPENAI_EMB_MODEL = os.getenv("AZURE_OPENAI_EMB_MODEL_NAME", "text-embedding-ada-002")
# Used with Azure OpenAI deployments
AZURE_OPENAI_SERVICE = os.getenv("AZURE_OPENAI_SERVICE")
AZURE_OPENAI_CHATGPT_DEPLOYMENT = os.getenv("AZURE_OPENAI_CHATGPT_DEPLOYMENT")
AZURE_OPENAI_EMB_DEPLOYMENT = os.getenv("AZURE_OPENAI_EMB_DEPLOYMENT")
AZURE_OPENAI_CHATGPT_DEPLOYMENT = os.getenv("AZURE_OPENAI_CHATGPT_DEPLOYMENT") if OPENAI_HOST == "azure" else None
AZURE_OPENAI_EMB_DEPLOYMENT = os.getenv("AZURE_OPENAI_EMB_DEPLOYMENT") if OPENAI_HOST == "azure" else None
# Used only with non-Azure OpenAI deployments
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
OPENAI_ORGANIZATION = os.getenv("OPENAI_ORGANIZATION")
Expand Down Expand Up @@ -250,50 +232,53 @@ async def setup_clients():
blob_container_client = blob_client.get_container_client(AZURE_STORAGE_CONTAINER)

# Used by the OpenAI SDK
openai_client: AsyncOpenAI

if OPENAI_HOST == "azure":
openai.api_type = "azure_ad"
openai.api_base = f"https://{AZURE_OPENAI_SERVICE}.openai.azure.com"
openai.api_version = "2023-07-01-preview"
openai_token = await azure_credential.get_token("https://cognitiveservices.azure.com/.default")
openai.api_key = openai_token.token
token_provider = get_bearer_token_provider(azure_credential, "https://cognitiveservices.azure.com/.default")
# Store on app.config for later use inside requests
current_app.config[CONFIG_OPENAI_TOKEN] = openai_token
openai_client = AsyncAzureOpenAI(
api_version="2023-07-01-preview",
azure_endpoint=f"https://{AZURE_OPENAI_SERVICE}.openai.azure.com",
azure_ad_token_provider=token_provider,
)
else:
openai.api_type = "openai"
openai.api_key = OPENAI_API_KEY
openai.organization = OPENAI_ORGANIZATION
openai_client = AsyncOpenAI(
api_key=OPENAI_API_KEY,
organization=OPENAI_ORGANIZATION,
)

current_app.config[CONFIG_CREDENTIAL] = azure_credential
current_app.config[CONFIG_OPENAI_CLIENT] = openai_client
current_app.config[CONFIG_SEARCH_CLIENT] = search_client
current_app.config[CONFIG_BLOB_CONTAINER_CLIENT] = blob_container_client
current_app.config[CONFIG_AUTH_CLIENT] = auth_helper

# Various approaches to integrate GPT and external knowledge, most applications will use a single one of these patterns
# or some derivative, here we include several for exploration purposes
current_app.config[CONFIG_ASK_APPROACH] = RetrieveThenReadApproach(
search_client,
OPENAI_HOST,
AZURE_OPENAI_CHATGPT_DEPLOYMENT,
OPENAI_CHATGPT_MODEL,
AZURE_OPENAI_EMB_DEPLOYMENT,
OPENAI_EMB_MODEL,
KB_FIELDS_SOURCEPAGE,
KB_FIELDS_CONTENT,
AZURE_SEARCH_QUERY_LANGUAGE,
AZURE_SEARCH_QUERY_SPELLER,
search_client=search_client,
openai_client=openai_client,
chatgpt_model=OPENAI_CHATGPT_MODEL,
chatgpt_deployment=AZURE_OPENAI_CHATGPT_DEPLOYMENT,
embedding_model=OPENAI_EMB_MODEL,
embedding_deployment=AZURE_OPENAI_EMB_DEPLOYMENT,
sourcepage_field=KB_FIELDS_SOURCEPAGE,
content_field=KB_FIELDS_CONTENT,
query_language=AZURE_SEARCH_QUERY_LANGUAGE,
query_speller=AZURE_SEARCH_QUERY_SPELLER,
)

current_app.config[CONFIG_CHAT_APPROACH] = ChatReadRetrieveReadApproach(
search_client,
OPENAI_HOST,
AZURE_OPENAI_CHATGPT_DEPLOYMENT,
OPENAI_CHATGPT_MODEL,
AZURE_OPENAI_EMB_DEPLOYMENT,
OPENAI_EMB_MODEL,
KB_FIELDS_SOURCEPAGE,
KB_FIELDS_CONTENT,
AZURE_SEARCH_QUERY_LANGUAGE,
AZURE_SEARCH_QUERY_SPELLER,
search_client=search_client,
openai_client=openai_client,
chatgpt_model=OPENAI_CHATGPT_MODEL,
chatgpt_deployment=AZURE_OPENAI_CHATGPT_DEPLOYMENT,
embedding_model=OPENAI_EMB_MODEL,
embedding_deployment=AZURE_OPENAI_EMB_DEPLOYMENT,
sourcepage_field=KB_FIELDS_SOURCEPAGE,
content_field=KB_FIELDS_CONTENT,
query_language=AZURE_SEARCH_QUERY_LANGUAGE,
query_speller=AZURE_SEARCH_QUERY_SPELLER,
)


Expand All @@ -305,6 +290,8 @@ def create_app():
configure_azure_monitor()
# This tracks HTTP requests made by aiohttp:
AioHttpClientInstrumentor().instrument()
# This tracks HTTP requests made by httpx/openai:
HTTPXClientInstrumentor().instrument()
# This middleware tracks app route requests:
app.asgi_app = OpenTelemetryMiddleware(app.asgi_app) # type: ignore[method-assign]

Expand Down
Loading

0 comments on commit 94f71a2

Please sign in to comment.