draft - first draft

Author: fletchersarip93

README.md

@@ -49,14 +49,14 @@ try:
         model="gpt-5",
         messages=[{"role": "user", "content": "Hello world"}],
     )
-    print(chat.llm_response.choices[0].message.content)
+    print(chat.choices[0].message.content)
 
     # Or with the Responses API
     resp = client.responses.create(
         model="gpt-5",
         input="What are the main features of your premium plan?",
     )
-    print(resp.llm_response.output_text)
+    print(resp.output_text)
 except GuardrailTripwireTriggered as e:
     print(f"Guardrail triggered: {e}")
 ```

docs/index.md

@@ -35,7 +35,7 @@ response = await client.responses.create(
     input="Hello"
 )
 # Guardrails run automatically
-print(response.llm_response.output_text)
+print(response.output_text)
 ```
 
 ## Next Steps

docs/quickstart.md

@@ -70,16 +70,15 @@ async def main():
             input="Hello world"
         )
 
-        # Access OpenAI response via .llm_response
-        print(response.llm_response.output_text)
+        print(response.output_text)
 
     except GuardrailTripwireTriggered as exc:
         print(f"Guardrail triggered: {exc.guardrail_result.info}")
 
 asyncio.run(main())
 ```
 
-**That's it!** Your existing OpenAI code now includes automatic guardrail validation based on your pipeline configuration. Just use `response.llm_response` instead of `response`.
+**That's it!** Your existing OpenAI code now includes automatic guardrail validation based on your pipeline configuration.
 
 ## Multi-Turn Conversations
 
@@ -98,7 +97,7 @@ while True:
             model="gpt-4o"
         )
 
-        response_content = response.llm_response.choices[0].message.content
+        response_content = response.choices[0].message.content
         print(f"Assistant: {response_content}")
 
         # ✅ Only append AFTER guardrails pass

docs/ref/checks/hallucination_detection.md

@@ -76,7 +76,7 @@ response = await client.responses.create(
 )
 
 # Guardrails automatically validate against your reference documents
-print(response.llm_response.output_text)
+print(response.output_text)
 ```
 
 ### How It Works

docs/tripwires.md

@@ -25,7 +25,7 @@ try:
         model="gpt-5",
         input="Tell me a secret"
     )
-    print(response.llm_response.output_text)
+    print(response.output_text)
 
 except GuardrailTripwireTriggered as exc:
     print(f"Guardrail triggered: {exc.guardrail_result.info}")

examples/basic/azure_implementation.py

@@ -75,7 +75,7 @@ async def process_input(
         )
 
         # Extract the response content from the GuardrailsResponse
-        response_text = response.llm_response.choices[0].message.content
+        response_text = response.choices[0].message.content
 
         # Only show output if all guardrails pass
         print(f"\nAssistant: {response_text}")

examples/basic/hello_world.py

@@ -49,13 +49,13 @@ async def process_input(
             previous_response_id=response_id,
         )
 
-        console.print(f"\nAssistant output: {response.llm_response.output_text}", end="\n\n")
+        console.print(f"\nAssistant output: {response.output_text}", end="\n\n")
 
         # Show guardrail results if any were run
         if response.guardrail_results.all_results:
             console.print(f"[dim]Guardrails checked: {len(response.guardrail_results.all_results)}[/dim]")
 
-        return response.llm_response.id
+        return response.id
 
     except GuardrailTripwireTriggered:
         raise

examples/basic/local_model.py

@@ -48,7 +48,7 @@ async def process_input(
         )
 
         # Access response content using standard OpenAI API
-        response_content = response.llm_response.choices[0].message.content
+        response_content = response.choices[0].message.content
         console.print(f"\nAssistant output: {response_content}", end="\n\n")
 
         # Add to conversation history

examples/basic/multi_bundle.py

@@ -66,15 +66,15 @@ async def process_input(
     with Live(output_text, console=console, refresh_per_second=10) as live:
         try:
             async for chunk in stream:
-                # Access streaming response exactly like native OpenAI API through .llm_response
-                if hasattr(chunk.llm_response, "delta") and chunk.llm_response.delta:
-                    output_text += chunk.llm_response.delta
+                # Access streaming response exactly like native OpenAI API
+                if hasattr(chunk, "delta") and chunk.delta:
+                    output_text += chunk.delta
                     live.update(output_text)
 
             # Get the response ID from the final chunk
             response_id_to_return = None
-            if hasattr(chunk.llm_response, "response") and hasattr(chunk.llm_response.response, "id"):
-                response_id_to_return = chunk.llm_response.response.id
+            if hasattr(chunk, "response") and hasattr(chunk.response, "id"):
+                response_id_to_return = chunk.response.id
 
             return response_id_to_return
 

examples/basic/multiturn_chat_with_alignment.py

@@ -235,7 +235,7 @@ async def main(malicious: bool = False) -> None:
                     tools=tools,
                 )
                 print_guardrail_results("initial", resp)
-                choice = resp.llm_response.choices[0]
+                choice = resp.choices[0]
                 message = choice.message
                 tool_calls = getattr(message, "tool_calls", []) or []
 
@@ -327,7 +327,7 @@ async def main(malicious: bool = False) -> None:
                     )
 
                     print_guardrail_results("final", resp)
-                    final_message = resp.llm_response.choices[0].message
+                    final_message = resp.choices[0].message
                     console.print(
                         Panel(
                             final_message.content or "(no output)",