Feedback

README.md

@@ -1,3 +1,4 @@
+[![Review Assignment Due Date](https://classroom.github.com/assets/deadline-readme-button-22041afd0340ce965d47ae6ef1cefeee28c7c493a6346c4f15d667ab976d596c.svg)](https://classroom.github.com/a/tLTGCA4G)
 ---
 title: "Activity 1 - Hello, Azure AI"
 type: lab

REFLECTION.md

@@ -2,14 +2,26 @@
 
 Answer these questions after completing the activity (2-3 sentences each). Connect your answers to specific things you observed while coding and experimenting.
 
+
+Thanks for all you do to make this class possible. 
+
+![Hand Waving](https://github.com/Mayur-Pagote/README_Design_Kit/blob/main/public/Assets/Hand%20Waving.gif)
+
 ## 1. Service Surprises
 
 Which of the three Azure AI services (OpenAI, Content Safety, Language) surprised you the most? Connect this to something specific you observed during your experiments -- a response you didn't expect, a behavior that seemed too easy or too hard, or a result that made you rethink how the service works.
 
+_The content saftey suprosed me the most. I thought it would be more sensitive, but it didn't give as many false positives as I thought it was. But it did error out when I fed it antisaftey prompts._
+
 ## 2. Lazy Initialization
 
 How would you explain the lazy initialization pattern to a colleague? Why is it used instead of creating clients at the top of the file?
 
+_When you create them at the top you are using memory resources. Loading them when the item is actually called cane save on memory and make the program start quicker_
+
 ## 3. Content Safety in the Real World
 
 A resident files this complaint: *"A man was assaulted at this intersection because the street light has been out for months."* This text describes real violence but is a legitimate safety concern. Should the system block it, flag it for human review, or pass it through? What factors would you weigh in making that decision?
+
+_The system should allow it through. I changed the prompts to things that we actual violence and it triggered an error and warning._
+_It knew that it wasnt a threat of violence it was explaining something_

app/main.py

@@ -45,13 +45,12 @@ def _get_openai_client():
     global _openai_client
     if _openai_client is None:
         # TODO: Uncomment and configure
-        #   from openai import AzureOpenAI
-        #   _openai_client = AzureOpenAI(
-        #       azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
-        #       api_key=os.environ["AZURE_OPENAI_API_KEY"],
-        #       api_version="2024-10-21",
-        #   )
-        raise NotImplementedError("Configure the Azure OpenAI client")
+        from openai import AzureOpenAI
+        _openai_client = AzureOpenAI(
+            azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
+            api_key=os.environ["AZURE_OPENAI_API_KEY"],
+            api_version="2024-10-21",
+        )
     return _openai_client
 
 
@@ -62,90 +61,68 @@ def _get_content_safety_client():
         # NOTE: The Content Safety SDK handles API versioning internally --
         # no api_version parameter is needed (unlike the OpenAI SDK).
         # TODO: Uncomment and configure
-        #   from azure.ai.contentsafety import ContentSafetyClient
-        #   from azure.core.credentials import AzureKeyCredential
-        #   _content_safety_client = ContentSafetyClient(
-        #       endpoint=os.environ["AZURE_CONTENT_SAFETY_ENDPOINT"],
-        #       credential=AzureKeyCredential(os.environ["AZURE_CONTENT_SAFETY_KEY"]),
-        #   )
-        raise NotImplementedError("Configure the Content Safety client")
+        from azure.ai.contentsafety import ContentSafetyClient
+        from azure.core.credentials import AzureKeyCredential
+        _content_safety_client = ContentSafetyClient(
+            endpoint=os.environ["AZURE_CONTENT_SAFETY_ENDPOINT"],
+            credential=AzureKeyCredential(os.environ["AZURE_CONTENT_SAFETY_KEY"]),
+        )
     return _content_safety_client
 
 
 def _get_language_client():
     """Lazily initialize the Azure AI Language client."""
     global _language_client
     if _language_client is None:
-        # NOTE: The Language SDK handles API versioning internally --
-        # no api_version parameter is needed (unlike the OpenAI SDK).
-        # TODO: Uncomment and configure
-        #   from azure.ai.textanalytics import TextAnalyticsClient
-        #   from azure.core.credentials import AzureKeyCredential
-        #   _language_client = TextAnalyticsClient(
-        #       endpoint=os.environ["AZURE_AI_LANGUAGE_ENDPOINT"],
-        #       credential=AzureKeyCredential(os.environ["AZURE_AI_LANGUAGE_KEY"]),
-        #   )
-        raise NotImplementedError("Configure the AI Language client")
+        from azure.ai.textanalytics import TextAnalyticsClient
+        from azure.core.credentials import AzureKeyCredential
+        _language_client = TextAnalyticsClient(
+            endpoint=os.environ["AZURE_AI_LANGUAGE_ENDPOINT"],
+            credential=AzureKeyCredential(os.environ["AZURE_AI_LANGUAGE_KEY"]),
+        )
+        #raise NotImplementedError("Configure the AI Language client")
     return _language_client
 
 
 # ---------------------------------------------------------------------------
 # TODO: Step 1 - Classify a 311 request with Azure OpenAI
 # ---------------------------------------------------------------------------
 def classify_311_request(request_text: str) -> dict:
-    """Send a Memphis 311 service request to Azure OpenAI for classification.
-
-    Args:
-        request_text: The citizen's complaint text.
-
-    Returns:
-        dict with keys: category, confidence, reasoning
-    """
-    # TODO: Step 1.1 - Get the OpenAI client
-    # TODO: Step 1.2 - Call client.chat.completions.create() with:
-    #   model=os.environ.get("AZURE_OPENAI_DEPLOYMENT", "gpt-4o")
-    #   A system message that classifies into: Pothole, Noise Complaint,
-    #   Trash/Litter, Street Light, Water/Sewer, Other
-    #   response_format={"type": "json_object"}, temperature=0
-    # TODO: Step 1.3 - Parse the JSON response with json.loads()
-    raise NotImplementedError("Implement classify_311_request in Step 1")
+    response = _get_openai_client().chat.completions.create( 
+    model=os.environ.get("AZURE_OPENAI_DEPLOYMENT", "gpt-4o"),
+    messages=[
+        {"role": "system", "content": "Classify this Memphis 311 request into one of: Pothole, Noise Complaint, Trash/Litter, Street Light, Water/Sewer, Other. Return JSON with keys: category, confidence, reasoning."},
+        {"role": "user", "content": request_text},
+    ],
+    response_format={"type": "json_object"},
+    temperature=0
+    )
+    result = json.loads(response.choices[0].message.content)
+    return result
 
 
 # ---------------------------------------------------------------------------
 # TODO: Step 2 - Check content safety
 # ---------------------------------------------------------------------------
 def check_content_safety(text: str) -> dict:
-    """Check text for harmful content using Azure Content Safety.
-
-    Args:
-        text: Text to analyze.
-
-    Returns:
-        dict with keys: safe (bool), categories (dict of category: severity)
-    """
-    # TODO: Step 2.1 - Get the Content Safety client
-    # TODO: Step 2.2 - Call client.analyze_text() with AnalyzeTextOptions
-    # TODO: Step 2.3 - Return safety results
-    raise NotImplementedError("Implement check_content_safety in Step 2")
+    from azure.ai.contentsafety.models import AnalyzeTextOptions
+    result = _get_content_safety_client().analyze_text(AnalyzeTextOptions(text=text))
+    categories = {
+        cat.category: cat.severity
+        for cat in result.categories_analysis
+    }
+    safe = all(severity == 0 for severity in categories.values())
+    return {"safe": True, "categories": categories} 
 
 
 # ---------------------------------------------------------------------------
 # TODO: Step 3 - Extract key phrases
 # ---------------------------------------------------------------------------
 def extract_key_phrases(text: str) -> list[str]:
-    """Extract key phrases from text using Azure AI Language.
-
-    Args:
-        text: Text to analyze.
-
-    Returns:
-        List of key phrase strings.
-    """
-    # TODO: Step 3.1 - Get the Language client
-    # TODO: Step 3.2 - Call client.extract_key_phrases([text])
-    # TODO: Step 3.3 - Return the list of key phrases
-    raise NotImplementedError("Implement extract_key_phrases in Step 3")
-
+    response = _get_language_client().extract_key_phrases([text])
+    if response[0].is_error:
+        raise ValueError(f"Error extracting key phrases: {response[0].error}")
+    return list(response[0].key_phrases)
 
 def main():
     """Main function -- call all three Azure AI services."""
@@ -272,4 +249,4 @@ def _print_dispatch_ticket(complaint_text: str, result: dict):
 
 
 if __name__ == "__main__":
-    main()
+    main()

reference/sdk-vs-api.md

@@ -152,4 +152,4 @@ Notice the differences: the auth header name changes (`api-key` for OpenAI vs `O
 
 ## Key Takeaway
 
-Use the SDK unless you have a specific reason not to. The SDK gives you retries, typed responses, and proper error handling -- exactly the things that are easy to forget when writing raw HTTP calls at 2 AM before a deadline.
+Use the SDK unless you have a specific reason not to. The SDK gives you retries, typed responses, and proper error handling -- exactly the things that are easy to forget when writing raw HTTP calls at 2 AM before a deadline.