Changed to Azure endpoint

app.py

@@ -28,16 +28,19 @@ from pathlib import Path
 
 logger = logging.getLogger(__name__)
 
-from utils import PromptTemplate, convert_gradio_to_openai, seed_openai_key
+from utils import (
+    PromptTemplate,
+    convert_gradio_to_openai,
+    initialize_client,
+    seed_azure_key
+)
 
 
 # %% Initialization
-
-CONFIG_DIR: Path = Path("./SPIA2024_localtesting")
-if os.environ.get(f"OPENAI_API_KEY", "DEFAULT") == "DEFAULT":
-    seed_openai_key()
-client = openai.OpenAI()
-
+CONFIG_DIR: Path = Path("./SPIA2024")
+if os.environ.get("AZURE_ENDPOINT") is None: # Set Azure credentials from local files
+    seed_azure_key()
+client = initialize_client()
 
 # %% (functions)
 def load_config(
@@ -55,19 +58,13 @@ def load_config(
 
 
 def initialize_interview(
-    system_message: str,
     initial_message: str,
-    model_args: dict[str, str | float]
 ) -> tuple[gr.Chatbot,
            gr.Textbox,
            gr.Button,
            gr.Button,
            gr.Button]:
     "Read system prompt and start interview. Change visibilities of elements."
-    if len(initial_message) == 0:  # If empty inital message, ask the LM to write it.
-        initial_message = call_openai(
-            [], system_message, client, model_args, stream=False
-        )
     chat_history = [
         [None, initial_message]
     ]  # First item is for user, in this case bot starts interaction.
@@ -119,39 +116,6 @@ def save_interview(
     logger.info("Uploading complete.")
 
 
-def call_openai(
-    messages: list[dict[str, str]],
-    system_message: str | None,
-    client: openai.Client,
-    model_args: dict,
-    stream: bool = False,
-):
-    "Utility function for calling OpenAI chat. Expects formatted messages."
-    if not messages:
-        messages = []
-    if system_message:
-        messages = [{"role": "system", "content": system_message}] + messages
-    try:
-        response = client.chat.completions.create(
-            messages=messages, **model_args, stream=stream
-        )
-        if stream:
-            for chunk in response:
-                yield chunk.choices[0].message.content
-        else:
-            content = response.choices[0].message.content
-            return content
-    except openai.APIConnectionError | openai.APIStatusError as e:
-        error_msg = (
-            "API unreachable.\n" f"STATUS_CODE: {e.status_code}" f"ERROR: {e.response}"
-        )
-        gr.Error(error_msg)
-        logger.error(error_msg)
-    except openai.RateLimitError:
-        warning_msg = "Hit rate limit. Wait a moment and retry."
-        gr.Warning(warning_msg)
-        logger.warning(warning_msg)
-
 
 def user_message(
     message: str, chat_history: list[list[str | None]]
@@ -206,7 +170,7 @@ def reset_interview() -> (
 # LAYOUT
 with gr.Blocks() as demo:
     gr.Markdown("# StewartLab LM Interviewer")
-    userDisplay = gr.Markdown("")
+    userDisplay = gr.Markdown("", visible=False)
 
     # Config values
     configDir = gr.State(value=CONFIG_DIR)
@@ -249,7 +213,7 @@ with gr.Blocks() as demo:
         outputs=[initialMessage, systemMessage, modelArgs, wandbArgs],
     ).then(
         initialize_interview,
-        inputs=[systemMessage, initialMessage, modelArgs],
+        inputs=[initialMessage],
         outputs=[
             chatDisplay,
             chatInput,
@@ -269,7 +233,7 @@ with gr.Blocks() as demo:
         outputs=[initialMessage, systemMessage, modelArgs, wandbArgs],
     ).then(
         initialize_interview,
-        inputs=[systemMessage, initialMessage, modelArgs],
+        inputs=[initialMessage],
         outputs=[
             chatDisplay,
             chatInput,

utils.py

@@ -94,46 +94,20 @@ def convert_openai_to_gradio(
     return chat_history
 
 
-def seed_openai_key(cfg: str = "~/.cfg/openai.cfg") -> None:
-    """
-    Reads OpenAI key from config file and adds it to environment.
-    Assumed config location is "~/.cfg/openai.cfg"
-    """
-    # Get OpenAI Key
+def seed_azure_key(cfg: str = "~/.cfg/openai.cfg") -> None:
     config = ConfigParser()
     try:
         config.read(Path(cfg).expanduser())
     except:
         raise ValueError(f"Could not using read file at: {cfg}.")
-    os.environ["OPENAI_API_KEY"] = config["API_KEY"]["secret"]
-
-
-
-def query_openai(
-    messages: list[dict[str, str]],
-    system_message: str | None,
-    client: openai.Client,
-    model_args: dict,
-    stream: bool = False,
-) -> None:
-    "Utility function for calling OpenAI chat. Expects formatted messages."
-    if not messages:
-        messages = []
-    if system_message:
-        messages = [{"role": "system", "content": system_message}] + messages
-    try:
-        response = client.chat.completions.create(
-            messages=messages, **model_args, stream=stream
-        )
-        if stream:
-            for chunk in response:
-                yield chunk.choices[0].message.content
-        else:
-            content = response.choices[0].message.content
-            return content
-    except openai.APIConnectionError | openai.APIStatusError as e:
-        error_msg = (
-            "API unreachable.\n" f"STATUS_CODE: {e.status_code}" f"ERROR: {e.response}"
-        )
-    except openai.RateLimitError:
-        warning_msg = "Hit rate limit. Wait a moment and retry."
+    os.environ["AZURE_ENDPOINT"] = config["AZURE"]["endpoint"]
+    os.environ["AZURE_SECRET"] = config["AZURE"]["key"]
+
+
+def initialize_client() -> openai.AsyncClient:
+    client = openai.AzureOpenAI(
+        azure_endpoint=os.environ["AZURE_ENDPOINT"],
+        api_key=os.environ["AZURE_SECRET"],
+        api_version="2023-05-15",
+    )
+    return client