Flip image horizontal instead of pass through

examples/flows/standard/describe-image/README.md

@@ -1,5 +1,5 @@
 # Describe image flow
-A flow that take image input and uses OpenAI GPT-4V tool to describe it. We also added a pass_through node from input to output to demo PromptFlow's ability to output image in both node output and flow output.
+A flow that take image input, flip it horizontally and uses OpenAI GPT-4V tool to describe it.
 
 Tools used in this flow：
 - `OpenAI GPT-4V` tool

examples/flows/standard/describe-image/flip_image.py

@@ -0,0 +1,14 @@
+import io
+from promptflow import tool
+from promptflow.contracts.multimedia import Image
+from PIL import Image as PIL_Image
+
+
+@tool
+def passthrough(input_image: Image) -> Image:
+    image_stream = io.BytesIO(input_image)
+    pil_image = PIL_Image.open(image_stream)
+    flipped_image = pil_image.transpose(PIL_Image.FLIP_LEFT_RIGHT)
+    buffer = io.BytesIO()
+    flipped_image.save(buffer, format="PNG")
+    return Image(buffer.getvalue(), mime_type="image/png")

examples/flows/standard/describe-image/flow.dag.yaml

@@ -11,7 +11,7 @@ outputs:
     reference: ${question_on_image.output}
   output_image:
     type: string
-    reference: ${passthrough.output}
+    reference: ${flip_image.output}
 nodes:
 - name: question_on_image
   type: custom_llm
@@ -22,13 +22,13 @@ nodes:
   inputs:
     connection: openai_connection
     question: ${inputs.question}
-    test_image: ${passthrough.output}
+    test_image: ${flip_image.output}
     model: gpt-4-vision-preview
     max_tokens: 200
-- name: passthrough
+- name: flip_image
   type: python
   source:
     type: code
-    path: passthrough.py
+    path: flip_image.py
   inputs:
     input_image: ${inputs.input_image}