Improve uses of get, fix temperature bug (#1225)

* Improve uses of get, fix temperature bug

* Mention temperature in comments/docs

Author: pamelafox

app/backend/approaches/approach.py

@@ -96,7 +96,7 @@ def __init__(
         self.openai_host = openai_host
 
     def build_filter(self, overrides: dict[str, Any], auth_claims: dict[str, Any]) -> Optional[str]:
-        exclude_category = overrides.get("exclude_category") or None
+        exclude_category = overrides.get("exclude_category")
         security_filter = self.auth_helper.build_security_filters(overrides, auth_claims)
         filters = []
         if exclude_category:

app/backend/approaches/chatreadretrieveread.py

@@ -129,7 +129,7 @@ async def run_until_final_call(
             messages=messages,  # type: ignore
             # Azure Open AI takes the deployment name as the model name
             model=self.chatgpt_deployment if self.chatgpt_deployment else self.chatgpt_model,
-            temperature=0.0,
+            temperature=0.0,  # Minimize creativity for search query generation
             max_tokens=100,  # Setting too low risks malformed JSON, setting too high may affect performance
             n=1,
             tools=tools,
@@ -196,7 +196,7 @@ async def run_until_final_call(
             # Azure Open AI takes the deployment name as the model name
             model=self.chatgpt_deployment if self.chatgpt_deployment else self.chatgpt_model,
             messages=messages,
-            temperature=overrides.get("temperature") or 0.7,
+            temperature=overrides.get("temperature", 0.7),
             max_tokens=response_token_limit,
             n=1,
             stream=should_stream,

app/backend/approaches/chatreadretrievereadvision.py

@@ -110,7 +110,7 @@ async def run_until_final_call(
         chat_completion: ChatCompletion = await self.openai_client.chat.completions.create(
             model=self.gpt4v_deployment if self.gpt4v_deployment else self.gpt4v_model,
             messages=messages,
-            temperature=overrides.get("temperature") or 0.0,
+            temperature=0.0,  # Minimize creativity for search query generation
             max_tokens=100,
             n=1,
         )
@@ -194,7 +194,7 @@ async def run_until_final_call(
         chat_coroutine = self.openai_client.chat.completions.create(
             model=self.gpt4v_deployment if self.gpt4v_deployment else self.gpt4v_model,
             messages=messages,
-            temperature=overrides.get("temperature") or 0.7,
+            temperature=overrides.get("temperature", 0.7),
             max_tokens=response_token_limit,
             n=1,
             stream=should_stream,

app/backend/approaches/retrievethenread.py

@@ -99,7 +99,7 @@ async def run(
 
         user_content = [q]
 
-        template = overrides.get("prompt_template") or self.system_chat_template
+        template = overrides.get("prompt_template", self.system_chat_template)
         model = self.chatgpt_model
         message_builder = MessageBuilder(template, model)
 
@@ -118,7 +118,7 @@ async def run(
                 # Azure Open AI takes the deployment name as the model name
                 model=self.chatgpt_deployment if self.chatgpt_deployment else self.chatgpt_model,
                 messages=message_builder.messages,
-                temperature=overrides.get("temperature") or 0.3,
+                temperature=overrides.get("temperature", 0.3),
                 max_tokens=1024,
                 n=1,
             )

app/backend/approaches/retrievethenreadvision.py

@@ -112,7 +112,7 @@ async def run(
         image_list: list[ChatCompletionContentPartImageParam] = []
         user_content: list[ChatCompletionContentPartParam] = [{"text": q, "type": "text"}]
 
-        template = overrides.get("prompt_template") or (self.system_chat_template_gpt4v)
+        template = overrides.get("prompt_template", self.system_chat_template_gpt4v)
         model = self.gpt4v_model
         message_builder = MessageBuilder(template, model)
 
@@ -137,7 +137,7 @@ async def run(
             await self.openai_client.chat.completions.create(
                 model=self.gpt4v_deployment if self.gpt4v_deployment else self.gpt4v_model,
                 messages=message_builder.messages,
-                temperature=overrides.get("temperature") or 0.3,
+                temperature=overrides.get("temperature", 0.3),
                 max_tokens=1024,
                 n=1,
             )

app/backend/core/authentication.py

@@ -106,7 +106,7 @@ def get_auth_setup_for_client(self) -> dict[str, Any]:
     @staticmethod
     def get_token_auth_header(headers: dict) -> str:
         # Obtains the Access Token from the Authorization Header
-        auth = headers.get("Authorization", None)
+        auth = headers.get("Authorization")
         if auth:
             parts = auth.split()
 
@@ -122,7 +122,7 @@ def get_token_auth_header(headers: dict) -> str:
 
         # App services built-in authentication passes the access token directly as a header
         # To learn more, please visit https://learn.microsoft.com/azure/app-service/configure-authentication-oauth-tokens
-        token = headers.get("x-ms-token-aad-access-token", None)
+        token = headers.get("x-ms-token-aad-access-token")
         if token:
             return token
 
@@ -141,10 +141,10 @@ def build_security_filters(self, overrides: dict[str, Any], auth_claims: dict[st
             )
 
         oid_security_filter = (
-            "oids/any(g:search.in(g, '{}'))".format(auth_claims.get("oid") or "") if use_oid_security_filter else None
+            "oids/any(g:search.in(g, '{}'))".format(auth_claims.get("oid", "")) if use_oid_security_filter else None
         )
         groups_security_filter = (
-            "groups/any(g:search.in(g, '{}'))".format(", ".join(auth_claims.get("groups") or []))
+            "groups/any(g:search.in(g, '{}'))".format(", ".join(auth_claims.get("groups", [])))
             if use_groups_security_filter
             else None
         )
@@ -212,7 +212,7 @@ async def get_auth_claims_if_enabled(self, headers: dict) -> dict[str, Any]:
             # Read the claims from the response. The oid and groups claims are used for security filtering
             # https://learn.microsoft.com/azure/active-directory/develop/id-token-claims-reference
             id_token_claims = graph_resource_access_token["id_token_claims"]
-            auth_claims = {"oid": id_token_claims["oid"], "groups": id_token_claims.get("groups") or []}
+            auth_claims = {"oid": id_token_claims["oid"], "groups": id_token_claims.get("groups", [])}
 
             # A groups claim may have been omitted either because it was not added in the application manifest for the API application,
             # or a groups overage claim may have been emitted.

app/backend/core/modelhelper.py

@@ -56,4 +56,4 @@ def get_oai_chatmodel_tiktok(aoaimodel: str) -> str:
         raise ValueError(message)
     if aoaimodel not in AOAI_2_OAI and aoaimodel not in MODELS_2_TOKEN_LIMITS:
         raise ValueError(message)
-    return AOAI_2_OAI.get(aoaimodel) or aoaimodel
+    return AOAI_2_OAI.get(aoaimodel, aoaimodel)

docs/customization.md

@@ -37,9 +37,9 @@ Typically, the primary backend code you'll want to customize is the `app/backend
 
 The chat tab uses the approach programmed in [chatreadretrieveread.py](https://github.com/Azure-Samples/azure-search-openai-demo/blob/main/app/backend/approaches/chatreadretrieveread.py).
 
-1. It uses the OpenAI ChatCompletion API to turn the user question into a good search query.
+1. It calls the OpenAI ChatCompletion API (with a temperature of 0) to turn the user question into a good search query.
 2. It queries Azure AI Search for search results for that query (optionally using the vector embeddings for that query).
-3. It then combines the search results and original user question, and asks OpenAI ChatCompletion API to answer the question based on the sources. It includes the last 4K of message history as well (or however many tokens are allowed by the deployed model).
+3. It then combines the search results and original user question, and calls the OpenAI ChatCompletion API (with a temperature of 0.7) to answer the question based on the sources. It includes the last 4K of message history as well (or however many tokens are allowed by the deployed model).
 
 The `system_message_chat_conversation` variable is currently tailored to the sample data since it starts with "Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook." Change that to match your data.
 
@@ -56,7 +56,7 @@ If you followed the instructions in [docs/gpt4v.md](docs/gpt4v.md) to enable the
 The ask tab uses the approach programmed in [retrievethenread.py](https://github.com/Azure-Samples/azure-search-openai-demo/blob/main/app/backend/approaches/retrievethenread.py).
 
 1. It queries Azure AI Search for search results for the user question (optionally using the vector embeddings for that question).
-2. It then combines the search results and user question, and asks OpenAI ChatCompletion API to answer the question based on the sources.
+2. It then combines the search results and user question, and calls the OpenAI ChatCompletion API (with a temperature of 0.3) to answer the question based on the sources.
 
 The `system_chat_template` variable is currently tailored to the sample data since it starts with "You are an intelligent assistant helping Contoso Inc employees with their healthcare plan questions and employee handbook questions." Change that to match your data.
 

scripts/prepdocslib/filestrategy.py

@@ -59,8 +59,8 @@ async def run(self, search_info: SearchInfo):
             async for file in files:
                 try:
                     key = file.file_extension()
-                    processor = self.file_processors[key]
-                    if not processor:
+                    processor = self.file_processors.get(key)
+                    if processor is None:
                         # skip file if no parser is found
                         if search_info.verbose:
                             print(f"Skipping '{file.filename()}'.")