support multi output

app.py

@@ -3,18 +3,20 @@ import gradio as gr
 import numpy as np
 import os
 
-from langchain.chat_models import ChatOpenAI
+from langchain_community.chat_models import ChatOpenAI
 from langchain.schema import HumanMessage, SystemMessage, AIMessage
 from langchain_anthropic import ChatAnthropic, ChatAnthropicMessages
 from langchain_groq import ChatGroq
 import openai
 
 
-openai.api_key = os.environ["OPENAI_API_KEY"]
-# chat = ChatOpenAI(model="gpt-4-turbo-preview")
-# chat = ChatOpenAI(model="gpt-3.5-turbo")
+feature_text = "Brand: Duckly. \nProduct name: Duck runner pro. \nKey properties: t-shirt, for running, sweat wicking, for marathon, 100% cotton."
+
+garment_type = "all"
+
+reference_text = ""
 
-structure = \
+structure_text = \
 """# Headline {{ headline | inspiring, bold, action-oriented, max 8 words }}
 
 ## Introduction
@@ -22,7 +24,7 @@ structure = \
 
 ## Features and Benefits
 {% for feature in features %}
-### {{ loop.index }}: {{ feature.name | dynamic, direct, 5-6 words }}
+### Feature {{ loop.index }}: {{ feature.name | dynamic, direct, 5-6 words }}
 {{ feature.details | energetic, clear, 3-4 sentences }}
 {% endfor %}
 
@@ -30,11 +32,15 @@ structure = \
 {{ technical_specs | informative, to the point, concise list format }}
 """
 
-features = "Brand: Duckly. \nProduct name: Duck runner pro. \nKey properties: t-shirt, for running, sweat wicking, for marathon, 100% cotton."
-
-reference = " "
+structure_text_1 = """[type: UK website, style=true, language=English]
+{{ introduction_paragraph | motivational, passionate, 1-2 sentences }}
+{% for feature in features as bulleted list %}
+{{ feature.description | dynamic, direct, 3-6 words }}
+{% endfor %}
+{{ technical_specs | informative, to the point, concise list format }}"""
 
-garment_type = "all"
+structure_text_2 = """[type: Japanese newsletter, style=true, language=Japanese]
+{{ introduction_paragraph | motivational, passionate, 3-6 sentences }}"""
 
 languages = ["American English",
              "British English",
@@ -54,6 +60,9 @@ models = ["gpt-4-turbo",
           "claude-3-opus-20240229", 
           "llama3-70b-8192"]
 
+openai.api_key = os.environ["OPENAI_API_KEY"]
+
+
 import base64
 import requests
 
@@ -79,7 +88,7 @@ def get_json(text: str):
     return json.loads(text)
 
 
-def detect_features(image_paths, garment_type, language):
+def detect_features(image_paths, garment_type, language="English"):
     # Path to your image
     # image_path = "path_to_your_image.jpg"
 
@@ -95,7 +104,7 @@ def detect_features(image_paths, garment_type, language):
         }
 
         payload = {
-        "model": "gpt-4-vision-preview",
+        "model": "gpt-4-vision",
         "messages": [
             {
             "role": "user",
@@ -136,11 +145,17 @@ def detect_features(image_paths, garment_type, language):
         print(e.__class__, e)
         traceback.print_exc()
         return "", []
+    
+
+def generate(*data):
+    global visible
 
+    feature, image, garment_type, model, temperature = data[:5]
+    struct_ref = data[5:]
 
-def generate(features, image, garment_type, structure, reference, model, language, output_types, temperature):
-    print(f"features {features},\n image {image},\n structure{structure},\n model{model},\n language{language},\n temperature {temperature},\n reference {reference}")
-    image_features, base64_images = detect_features(image, garment_type, language)
+    print(f"features {feature},\n image {image},\n garment_type {garment_type},\n model {model},\n temperature {temperature},\n struct_ref {struct_ref}")
+
+    image_features, base64_images = detect_features(image, garment_type)
     detected_features = ""
     intended_use = ""
     alt_texts = []
@@ -155,93 +170,82 @@ def generate(features, image, garment_type, structure, reference, model, languag
         chat = ChatAnthropic(model_name=model, anthropic_api_key=os.environ["ANTHROPIC_API_KEY"])
     else:
         chat = ChatGroq(model_name=model, api_key=os.environ["GROQ_API_KEY"])
-    desc_messages = [
-        SystemMessage(content=f"""You are a helpful assistant that writes product descriptions for ecommerce websites. You write in {language} language."""),
-        HumanMessage(content=f"""Write a product description of about 200 words for a product with the following key properties. 
-                             Make sure that the description follows the structure of the reference structure below. 
-                             Make sure to use markdown format for the output.
-                             Make sure that the entire output is written entirely in {language} language. 
-                             Output the product description only, do not include any preceeding text like "Here is your product description".
-                             Use creative language that is suitable for e-commerce websites. Use a consistent tone of voice throughout the text. 
-                             If the reference text is not empty, produce the product description in the tone of voice and structure of the reference text. 
-                             \n\n {features + detected_features} \n{intended_use} \nReference structure: {structure}\n Reference text: {reference}""")]
-    news_messages = [
-        SystemMessage(content=f"""You are a helpful assistant that writes news letters for ecommerce websites. You write in {language} language."""),
-        HumanMessage(content=f"""Write a news letter of about 200 words for a product with the following key properties. 
-                             Make sure that the description follows the structure of the reference structure below. 
-                             Make sure to use markdown format for the output.
-                             Make sure that the entire output is written entirely in {language} language. 
-                             Output the product description only, do not include any preceeding text like "Here is your product description".
-                             Use creative language that is suitable for e-commerce websites. Use a consistent tone of voice throughout the text. 
-                             If the reference text is not empty, produce the product description in the tone of voice and structure of the reference text. 
-                             \n\n {features + detected_features} \n{intended_use} \nReference structure: {structure}\n Reference text: {reference}""")]
-    all_messages = [
-        SystemMessage(content=f"""You are a helpful assistant that writes news letters for ecommerce websites. You write in {language} language."""),
-        HumanMessage(content=f"""Write a news letter and a product description of about 200 words for a product with the following key properties. 
-                             Make sure that the description follows the structure of the reference structure below. 
-                             Make sure to use markdown format for the output.
-                             Make sure that the entire output is written entirely in {language} language. 
-                             Output the product description only, do not include any preceeding text like "Here is your product description".
-                             Use creative language that is suitable for e-commerce websites. Use a consistent tone of voice throughout the text. 
-                             If the reference text is not empty, produce the news letter and product description in the tone of voice and structure of the reference text. 
-                             Return the result in the following JSON format: {{"description": Product description, "news_letter": News letter}}.
-                             \n\n {features + detected_features} \n{intended_use} \nReference structure: {structure}\n Reference text: {reference}""")]
-
-    print(output_types)
+
+    batch = []
+    for i in range(visible + 1):
+        structure = struct_ref[2 * i]
+        reference = struct_ref[2 * i + 1]   
+        messages = [
+            SystemMessage(content=f"""You are a helpful assistant that writes about products for ecommerce websites."""),
+            HumanMessage(content=f"""Write a product description with the following features.
+                                Make sure that the description follows the structure of the reference structure. 
+                                Make sure to use markdown format for the output.
+                                Make sure that the entire output is written entirely in language defined in the reference structure. 
+                                Use language that is suitable for the type of document specified in the reference structure. 
+                                Use a consistent tone of voice throughout the text. 
+                                If the reference text is not empty, write the product description in the tone of voice of the reference text. 
+                                Make sure to output the product description only, do not include any preceeding text like "Here is your product description" or any part of the reference structure in the output.
+                                \n\n {feature + detected_features} \n{intended_use} \nReference structure: {structure}\n Reference text: {reference}""")]
+        batch.append(messages)
 
     description = ""
     
-    print(all_messages[1].content)
+    response = chat.batch(batch, temperature=temperature)
+    print(response)
+
 
-    response = chat.invoke(desc_messages, temperature=temperature)
-    description = response.content
+    description = "\n\n".join([msg.content for msg in response])
     md_content = description 
 
     alt_texts_str = '\n\n### Alt text\n\n' + '\n- ' + '\n- '.join(alt_texts) if len(alt_texts) > 0 else ""
 
     alt_text_dict = {k[0]: v for (k, v) in zip(image, alt_texts)} if len(alt_texts) > 0 else {}
-    result_json = {"description": description, "alt_text": alt_text_dict, "language": language}
+    result_json = {"outputs": [msg.content for msg in response], "alt_text": alt_text_dict}
     result_md = md_content + alt_texts_str + '\n'.join([f'![Product photo](data:image/png;base64,{base64_image} "{alt_text}")' if base64_image != "" else "" for (base64_image, alt_text) in zip(base64_images, alt_texts)])
     return result_md, result_json
 
 
+visible = 1
+def add_output_click(*struct_ref):
+    global visible
+    print("Adding output ", visible)
+    # print(struct_ref)
+    visible += 1
+    structure_texts = struct_ref[::2]
+    reference_texts = struct_ref[1::2]
+    structures = [gr.Textbox(label=f"Structure {i}", lines=10, value=structure_texts[i], interactive=True, visible=i <= visible) for i in range(10)]
+    references = [gr.Textbox(label=f"Reference copy {i}", lines=3, value=reference_texts[i], interactive=True, visible=i <= visible) for i in range(10)]
 
+    struct_ref = [val for pair in zip(structures, references) for val in pair]
+    return struct_ref
 
-# demo = gr.Interface(
-#     fn=generate,
-#     inputs=[gr.Textbox(features, label="Features"), 
-#             gr.Gallery(label="Product image(s)", type="filepath"), 
-#             gr.Textbox(garment_type, label="Garment type"),
-#             gr.Textbox(structure, label="Structure"), 
-#             gr.Textbox(reference, label="Reference copy"),
-#             gr.Dropdown(["gpt-4-turbo", "gpt-3.5-turbo", "claude-3-sonnet-20240229", "claude-3-opus-20240229", "llama3-70b-8192"], value="llama3-70b-8192", label="Model"),
-#             gr.Slider(minimum=0., maximum=1.0, value=0.5, label="Temperature")],
-#     outputs=[gr.Markdown(label="Markdown"), gr.Textbox(label="Raw text")],
-# )
 
 with gr.Blocks() as demo:
-    inputs=[gr.Textbox(features, label="Features", lines=3, interactive=True), 
-            gr.Gallery(label="Product image(s)", type="filepath"), 
-            gr.Textbox(garment_type, label="Garment type"),
-            gr.Textbox(structure, label="Structure", lines=10, interactive=True), 
-            gr.Textbox(reference, label="Reference copy", lines=3, interactive=True),
-            gr.Dropdown(models, value="claude-3-sonnet-20240229", label="Model"),
-            gr.Dropdown(languages, label="Language", value="American English"),
-            gr.Slider(minimum=0., maximum=1.0, value=0.5, label="Temperature")]
-    outputs=[gr.Markdown(label="Markdown"), gr.JSON(label="JSON")]
-    submit_button = gr.Button("Submit")
-    submit_button.click(fn=generate, inputs=inputs, outputs=outputs)
-
-# import argparse
-
-# parser = argparse.ArgumentParser(description='Run the Gradio demo')
-# parser.add_argument('--public', action='store_true', help='Expose the demo to the public')
-# args = parser.parse_args()
-
-# if args.public:
-#     demo.launch(share=True)
-# else:
-#     demo.launch(server_name="0.0.0.0", server_port=7868)
-
-if __name__ == "__main__":
+    with gr.Row():
+        with gr.Column():
+            feature = gr.Textbox(label="Features", value=feature_text, lines=3, interactive=True)
+            image = gr.Gallery(label="Images")
+            garment_type = gr.Textbox(label="Garment Type", value="all", lines=1, interactive=True) 
+            # language = gr.Dropdown(languages, value="American English", interactive=True, label="Language")
+            model = gr.Dropdown(models, value="gpt-4-turbo", interactive=True, label="Model")
+            temperature = gr.Slider(minimum=0., maximum=1.0, value=0., interactive=True, label="Temperature")
+            submit = gr.Button(value="Submit")
+            
+        with gr.Column():
+            struct_ref = [val for i in range(10) for val in 
+                          [gr.Textbox(label=f"Structure {i}", lines=10, value="", interactive=True, visible=i <= visible), 
+                           gr.Textbox(label=f"Reference copy {i}", lines=3, value="", interactive=True, visible=i <= visible)]]
+            struct_ref[0].value = structure_text_1
+            struct_ref[2].value = structure_text_2
+            add_output = gr.Button(value="Add Output")
+            add_output.click(add_output_click, inputs=struct_ref, outputs=struct_ref)
+        with gr.Column():
+            md_output = gr.Markdown(label="Output", show_label=True)
+            json_output = gr.JSON(label="JSON Output")  
+        submit.click(generate, inputs=[feature, image, garment_type, model, temperature, *struct_ref], 
+                     outputs=[md_output, json_output])
+        
+
+
+if __name__ == '__main__':
     demo.launch()