Improve guidance_prompt. Cleanup code

app.py

@@ -23,6 +23,8 @@ from langchain_core.vectorstores import VectorStoreRetriever
 
 import time 
 
+from str2escaped import str2escaped
+
 
 # feature_text = "Brand: Duckly. \nProduct name: Duck runner pro. \nKey properties: t-shirt, for running, sweat wicking, for marathon, 100% cotton."
 
@@ -105,6 +107,184 @@ models = ["gpt-4-turbo",
 
 openai.api_key = os.environ["OPENAI_API_KEY"]
 
+struct_copy_prompt = """Generate {nversions} versions of the product description for a product with the following information.
+Write in a way that target the customer.
+For each version, write at least {min_length} words and at most {max_length} words.
+Make sure that the structure of each output follows the reference structure. Do not add any additional sentences or structures that are not in the reference structure.
+Make sure to use the tone of voice, rythm, cadence and style of the reference copy for each output.
+Use markdown format for each output.
+Do not include any part of the reference structure in the output.
+The structure of the output should follow the reference structure.
+Do not use the structure of the reference copy in the output. 
+Do not use any of the excluded words in the output.
+Include all included words in the output.
+Make sure to use product features and intended use in the output.
+Do not hallucinate any information about the product, use only the provided key features and intended use to write about the product.
+Note that the reference copy should be used for style and tone only, do not use any part of the reference copy in the output. 
+Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
+Use creative language in each output, do not use the common ways of starting product descriptions.
+Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
+For each version, try to write in different style.
+Put each item in the list on a separate line. Use '-' character to start each item.  
+Rate the quality of each version based on the following criteria:
+    - how faithful it describes the product features. 
+    - how well it follows the reference structure. 
+    - how well it follows the tone of voice, rythm, cadence and style of the reference copy.
+    - how well it avoid the excluded words.                                 
+    - how well it includes the included words.                                 
+    - how creative the language is.                                    
+The score should be a number between 0 and 10 with 10 being the best quality.                              
+Return the result in the following JSON format:
+{{
+    "versions": [                                 
+        {{
+            "id": 1,
+            "content": The first product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,                         
+            "score": The score of the first product description
+        }},
+        {{
+            "id": 2,
+            "content": The second product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,                                     
+            "score": The score of the second product description
+        }},
+        ...                                 
+    ],                                      
+    "best_version": {{
+        "explanation": Explanation for why this version is the best,                                                              
+        "id": The id of the best version                                                             
+    }}                                                          
+}}                              
+Make sure that the output is in JSON format, no extra text should be included in the output.                                
+                                 
+Product information:
+Key features: {key_features}
+
+Reference structure: {structure}
+
+Reference copy: {copy}
+
+Included words: {included_phrases}
+
+Excluded words: {excluded_phrases}"""
+
+
+copy_prompt = """Generate {nversions} versions of the product description for a product with the following information.
+Write in a way that target the customer.
+For each version, write at least {min_length} words and at most {max_length} words.
+Make sure that the structure of each output follows the structure of the reference copy.
+Make sure to use the tone of voice, rythm, cadence and style of the reference copy for each output.
+Use markdown format for each output.
+Make sure that the structure of each output follows the structure of the reference copy.
+Do not use any of the excluded words in the output.
+Include all included words in the output.
+Make sure to use product features and intended use in the output.
+Do not hallucinate any information about the product, use only the provided key features and intended use to write about the product.
+Note that the reference copy should be used for style and tone only, do not use any part of the reference copy in the output. 
+Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
+Use creative language in each output, do not use the common ways of starting product descriptions.
+Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
+For each version, try to write in different style.
+Put each item in the list on a separate line. Use '-' character to start each item.  
+Rate the quality of each version based on the following criteria:
+    - how faithful it describes the product features. 
+    - how well it follows the tone of voice, rythm, cadence and style of the reference copy.
+    - how well it avoid the excluded words.
+    - how well it includes the included words.
+    - how creative the language is
+The score should be a number between 0 and 10 with 10 being the best quality.
+Return the result in the following JSON format:
+{{
+    "versions": [
+        {{
+            "id": 1,
+            "content": The first product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,
+            "score": The score of the first product description
+        }},
+        {{
+            "id": 2,
+            "content": The second product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,
+            "score": The score of the second product description
+        }},
+        ...
+    ],
+    "best_version": {{
+        "explanation": Explanation for why this version is the best,                                                              
+        "id": The id of the best version
+    }}
+}}
+Make sure that the output is in JSON format, no extra text should be included in the output.
+
+Product information:
+Key features: {key_features}
+
+Reference copy: {copy}
+
+Included words: {included_phrases}
+
+Excluded words: {excluded_phrases}"""
+
+
+struct_prompt = """Generate {nversions} versions of the product description for a product with the following information.
+Write in a way that target the customer.  
+For each version, write at least {min_length} words and at most {max_length} words.
+Make sure that the structure of each output follows the reference structure. Do not add any additional sentences or structures that are not in the reference structure.
+Use markdown format for each output.
+Do not include any part of the reference structure in the output.
+Make sure that the structure of each output follows the reference structure.
+Do not use any of the excluded words in the output.
+Include all included words in the output.
+Make sure to use product features and intended use in the output.
+Do not hallucinate any information about the product, use only the provided key features and intended use to write about the product.
+Note that the reference structure should be used for structure only, do not use any part of the reference structure in the output. 
+Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
+Use creative language in each output, do not use the common ways of starting product descriptions.
+Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
+For each version, try to write in different style.
+Put each item in the list on a separate line. Use '-' character to start each item.  
+Rate the quality of each version based on the following criteria:
+    - how faithful it describes the product features. 
+    - how well it follows the reference structure. 
+    - how well it avoid the excluded words.                                 
+    - how well it includes the included words.                                 
+    - how creative the language is.                                    
+The score should be a number between 0 and 10 with 10 being the best quality.                              
+Return the result in the following JSON format:
+{{
+    "versions": [                                 
+        {{
+            "id": 1,
+            "content": The first product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,                         
+            "score": The score of the first product description
+        }},
+        {{
+            "id": 2,
+            "content": The second product description,
+            "explanation": A less than 20 word explanation of the score of the first product description,                                     
+            "score": The score of the second product description
+        }},
+        ...                                 
+    ],                                      
+    "best_version": {{
+        "explanation": Explanation for why this version is the best,                                                              
+        "id": The id of the best version                                                                
+    }}                                                          
+}}                              
+Make sure that the output is in JSON format, no extra text should be included in the output.                                
+                                 
+Product information:
+Key features: {key_features}
+
+Reference structure: {structure}
+
+Included words: {included_phrases}
+
+Excluded words: {excluded_phrases}"""
+
 
 import base64
 import requests
@@ -239,16 +419,17 @@ def get_language(struct_lang, copy_lang):
     return "English"
 
 
-def post_process(text: str, guidance_prompt: str, chat: ChatOpenAI):
+def post_process(text: str, guidance_prompt: str, language: str, chat: ChatOpenAI):
     messages = [
-        SystemMessage(content=f"""You are a helpful assistant that edit documents based on the guidlines provided. Make sure to write in English language."""),
+        SystemMessage(content=f"""You are a helpful assistant that edit documents based on the guidlines provided. Make sure to write in {language} language."""),
         HumanMessage(content=f"""Given the following product description, your task is to
 make minimal modification to the product description such that the resulting description 
 follows the rules defined in the guidlines. Make sure to preserve the structure of the 
 original text as much as possible. Do not modify the structure of the original text.
+Do not change the language of the original text.                     
 Output only the modified text in markdown format.
 
-Text: 
+Product description: 
 {text}
 
 Guildlines: 
@@ -323,7 +504,7 @@ def generate(*data):
     print("visible", visible)
     
     nargs = 11 
-    feature, image, garment_type, model, temperature, nversions, excluded_words, included_words, glossary_upload, debug, guidance_prompt = data[:nargs]
+    feature, image, garment_type, model, temperature, nversions, excluded_phrases, included_phrases, glossary_upload, debug, guidance_prompt = data[:nargs]
     struct_ref = data[nargs:]
 
     print(f"{feature=}")
@@ -331,8 +512,8 @@ def generate(*data):
     print(f"{garment_type=}")
     print(f"{model=}")
     print(f"{temperature=}")
-    print(f"{excluded_words=}")
-    print(f"{included_words=}")
+    print(f"{excluded_phrases=}")
+    print(f"{included_phrases=}")
     print(f"{debug=}")
     # print(f"{glossary=}")
     print(f"{glossary_upload=}")
@@ -357,12 +538,15 @@ def generate(*data):
     detected_features = ""
     intended_use = ""
     alt_texts = []
+    key_features = feature
     if image_features is not None and len(image_features) > 0:
         alt_texts = image_features["alt_text"]
         detected_features = ", ".join(image_features["features"])
         intended_use = ", ".join(image_features["intended_use"])
         print(f"Detected features: {detected_features}, Intended use: {intended_use}, Alt text: {alt_texts}")
 
+        key_features = key_features + ", " + detected_features + "\nIntended uses: " + intended_use
+
     # if glossary:
     #     print("Getting terms")
     #     terms = glossary.invoke(input=feature + detected_features)
@@ -381,131 +565,14 @@ def generate(*data):
                 print("Using both copy and structure")
                 messages = [
                     SystemMessage(content=f"""You are a helpful assistant that writes about products for ecommerce websites. Make sure to write in {languages[i]} language."""),
-                    HumanMessage(content=f"""Generate {nversions} versions of the product description for a product with the following information.
-Write in a way that target the customer.
-For each version, write at least {min_length} words and at most {max_length} words.
-Make sure that the structure of each output follows the reference structure. Do not add any additional sentences or structures that are not in the reference structure.
-Make sure to use the tone of voice, rythm, cadence and style of the reference copy for each output.
-Use markdown format for each output.
-Do not include any part of the reference structure in the output.
-The structure of the output should follow the reference structure.
-Do not use the structure of the reference copy in the output. 
-Do not use any of the excluded words in the output.
-Include all included words in the output.
-Do not hallucinate any information. 
-Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
-Use creative language in each output, do not use the common ways of starting product descriptions.
-Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
-For each version, try to write in different style.
-Put each item in the list on a separate line. Use '-' character to start each item.  
-Rate the quality of each version based on the following criteria:
-    - how well it follows the reference structure. 
-    - how well it follows the tone of voice, rythm, cadence and style of the reference copy.
-    - how faithful it describes the product features. 
-    - how well it avoid the excluded words.                                 
-    - how well it includes the included words.                                 
-    - how creative the language is.                                    
-The score should be a number between 0 and 10 with 10 being the best quality.                              
-Return the result in the following JSON format:
-{{
-    "versions": [                                 
-        {{
-            "id": 1,
-            "content": The first product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,                         
-            "score": The score of the first product description
-        }},
-        {{
-            "id": 2,
-            "content": The second product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,                                     
-            "score": The score of the second product description
-        }},
-        ...                                 
-    ],                                      
-    "best_version": {{
-        "explanation": Explanation for why this version is the best,                                                              
-        "id": The id of the best version                                                             
-    }}                                                          
-}}                              
-Make sure that the output is in JSON format, no extra text should be included in the output.                                
-                                 
-Product information:
-Key features: {feature} 
-{detected_features} 
-
-Intended use: {intended_use} 
-
-Reference structure: {structure}
-
-Reference copy: {copy}
-
-Included words: {included_words}
-
-Excluded words: {excluded_words}"""),]
+                    HumanMessage(content=struct_copy_prompt.format(nversions=nversions, min_length=min_length, max_length=max_length, key_features=key_features, structure=structure, copy=copy, included_phrases=included_phrases, excluded_phrases=excluded_phrases)),]
                 
             elif len(copy.strip()) > 0:
                 print('------------')
                 print("Using copy")
                 messages = [
                     SystemMessage(content=f"""You are a helpful assistant that writes about products for ecommerce websites. Make sure to write in {languages[i]} language."""),
-                    HumanMessage(content=f"""Generate {nversions} versions of the product description for a product with the following information.
-Write in a way that target the customer.
-For each version, write at least {min_length} words and at most {max_length} words.
-Make sure that the structure of each output follows the structure of the reference copy.
-Make sure to use the tone of voice, rythm, cadence and style of the reference copy for each output.
-Use markdown format for each output.
-Make sure that the structure of each output follows the structure of the reference copy.
-Do not use any of the excluded words in the output.
-Include all included words in the output.
-Do not hallucinate any information. 
-Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
-Use creative language in each output, do not use the common ways of starting product descriptions.
-Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
-For each version, try to write in different style.
-Put each item in the list on a separate line. Use '-' character to start each item.  
-Rate the quality of each version based on the following criteria:
-    - how well it follows the tone of voice, rythm, cadence and style of the reference copy.
-    - how faithful it describes the product features.
-    - how well it avoid the excluded words.
-    - how well it includes the included words.
-    - how creative the language is
-The score should be a number between 0 and 10 with 10 being the best quality.
-Return the result in the following JSON format:
-{{
-    "versions": [
-        {{
-            "id": 1,
-            "content": The first product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,
-            "score": The score of the first product description
-        }},
-        {{
-            "id": 2,
-            "content": The second product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,
-            "score": The score of the second product description
-        }},
-        ...
-    ],
-    "best_version": {{
-        "explanation": Explanation for why this version is the best,                                                              
-        "id": The id of the best version
-    }}
-}}
-Make sure that the output is in JSON format, no extra text should be included in the output.
-
-Product information:
-Key features: {feature}
-{detected_features}
-
-Intended use: {intended_use}
-
-Reference copy: {copy}
-
-Included words: {included_words}
-
-Excluded words: {excluded_words}"""),]
+                    HumanMessage(content=copy_prompt.format(nversions=nversions, min_length=min_length, max_length=max_length, key_features=key_features, structure=structure, copy=copy, included_phrases=included_phrases, excluded_phrases=excluded_phrases)),]
                 print(messages[1].content)
                 print('------------')
 
@@ -514,63 +581,7 @@ Excluded words: {excluded_words}"""),]
                 print("Using structure")
                 messages = [
                     SystemMessage(content=f"""You are a helpful assistant that writes about products for ecommerce websites. Make sure to write in {languages[i]} language."""),
-                    HumanMessage(content=f"""Generate {nversions} versions of the product description for a product with the following information.
-Write in a way that target the customer.  
-For each version, write at least {min_length} words and at most {max_length} words.
-Make sure that the structure of each output follows the reference structure. Do not add any additional sentences or structures that are not in the reference structure.
-Use markdown format for each output.
-Do not include any part of the reference structure in the output.
-Make sure that the structure of each output follows the reference structure.
-Do not use any of the excluded words in the output.
-Include all included words in the output.
-Do not hallucinate any information. 
-Do not hallucinate information about size and weight. Write about size and weight only if it is available in the list of features.
-Use creative language in each output, do not use the common ways of starting product descriptions.
-Avoid common phrases and cliches such as "Step into something", "Elevate something", "Discover something", "Unleash something", "Embrace something", and similar phrases.
-For each version, try to write in different style.
-Put each item in the list on a separate line. Use '-' character to start each item.  
-Rate the quality of each version based on the following criteria:
-    - how well it follows the reference structure. 
-    - how faithful it describes the product features. 
-    - how well it avoid the excluded words.                                 
-    - how well it includes the included words.                                 
-    - how creative the language is.                                    
-The score should be a number between 0 and 10 with 10 being the best quality.                              
-Return the result in the following JSON format:
-{{
-    "versions": [                                 
-        {{
-            "id": 1,
-            "content": The first product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,                         
-            "score": The score of the first product description
-        }},
-        {{
-            "id": 2,
-            "content": The second product description,
-            "explanation": A less than 20 word explanation of the score of the first product description,                                     
-            "score": The score of the second product description
-        }},
-        ...                                 
-    ],                                      
-    "best_version": {{
-        "explanation": Explanation for why this version is the best,                                                              
-        "id": The id of the best version                                                                
-    }}                                                          
-}}                              
-Make sure that the output is in JSON format, no extra text should be included in the output.                                
-                                 
-Product information:
-Key features: {feature}
-{detected_features} 
-
-Intended use: {intended_use} 
-
-Reference structure: {structure}
-
-Included words: {included_words}
-
-Excluded words: {excluded_words}"""),]
+                    HumanMessage(content=struct_prompt.format(nversions=nversions, min_length=min_length, max_length=max_length, key_features=key_features, structure=structure, copy=copy, included_phrases=included_phrases, excluded_phrases=excluded_phrases)),]
                 print(messages[1].content)
                 print('------------')
             batch.append(messages)
@@ -583,7 +594,8 @@ Excluded words: {excluded_words}"""),]
     parser = JsonOutputParser()
     jresponse = [parser.parse(msg.content) for msg in response]
     descriptions = []
-    for jr in jresponse:
+    descriptions_post = []
+    for i, jr in enumerate(jresponse):
         print(f'{jr=}')
         bestid = jr["best_version"]["id"]
         for d in jr["versions"]:
@@ -604,14 +616,17 @@ Excluded words: {excluded_words}"""),]
         #         bests = d["score"]
         #         bestd = d["content"] + (f"\n\nDebug info:\n\nScore: {d['score']}\n\nExplanation: {d['explanation']}" if debug else "")
         #         break
+    
+        bestd_post = post_process(text=bestd, guidance_prompt=guidance_prompt, language=languages[i], chat=chat)
 
         descriptions.append(bestd)
+        descriptions_post.append(bestd_post)
 
-    print("\n\nRewriting with glossary")
-    rewrite = glossary_rewrite(chat=chat, glossary=glossary, text=descriptions[0])
-    if rewrite != "":
-        descriptions[0] = "Original:\n\n" + descriptions[0] + "\n\nRewritten:\n\n" + rewrite
-    print("\n\nDone rewriting with glossary\n\n")
+    # print("\n\nRewriting with glossary")
+    # rewrite = glossary_rewrite(chat=chat, glossary=glossary, text=descriptions[0])
+    # if rewrite != "":
+    #     descriptions[0] = "Original:\n\n" + descriptions[0] + "\n\nRewritten:\n\n" + rewrite
+    # print("\n\nDone rewriting with glossary\n\n")
 
     md_content = "\n\n---\n\n".join(descriptions)
 
@@ -620,9 +635,9 @@ Excluded words: {excluded_words}"""),]
     alt_text_dict = {k[0]: v for (k, v) in zip(image, alt_texts)} if len(alt_texts) > 0 else {}
     result_json = {"outputs": jresponse if debug else descriptions, "alt_text": alt_text_dict}
 
-    post_content = post_process(text=md_content, guidance_prompt=guidance_prompt, chat=chat)
+    # post_content = post_process(text=md_content, guidance_prompt=guidance_prompt, language=languages, chat=chat)
 
-    result_md = "Original\n\n" + md_content + "\n\n---\n\nModified\n\n" + post_content + "\n\n" \
+    result_md = "Original\n\n" + md_content + "\n\n---\n\nModified\n\n" + '\n'.join(descriptions_post) + "\n\n" \
                     + 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)])
 
@@ -681,8 +696,8 @@ with gr.Blocks() as demo:
                 model = gr.Dropdown(models, value="claude-3-5-sonnet-20241022", interactive=True, label="Model", visible=True)
                 temperature = gr.Slider(minimum=0., maximum=1.0, value=0., interactive=True, label="Temperature", visible=True)
                 nversions = gr.Slider(minimum=1, maximum=10, value=5, step=int, interactive=True, label="Number of versions", visible=True)
-                excluded_words = gr.Textbox(label="Excluded words", interactive=True, lines=2)
-                included_words = gr.Textbox(label="Included words", interactive=True, lines=2)
+                excluded_phrases = gr.Textbox(label="Excluded words", interactive=True, lines=2)
+                included_phrases = gr.Textbox(label="Included words", interactive=True, lines=2)
                 # glossary = gr.Dataframe(row_count = (2, "dynamic"), col_count=(2,"static"), headers=["Description", "Way of writing"], label="Glossary", interactive=True)
                 glossary_upload = gr.UploadButton(label="Upload Glossary", interactive=True, file_types=["csv"])
                 debug = gr.Checkbox(label="Debug", interactive=True, value=True)
@@ -708,7 +723,7 @@ with gr.Blocks() as demo:
             md_output = gr.Markdown(label="Output", show_label=True, line_breaks=True)
             json_output = gr.JSON(label="JSON Output")  
         submit.click(generate, inputs=[feature, image, garment_type, model, temperature, nversions,
-                                       excluded_words, included_words, glossary_upload, debug, guidance_prompt, *struct_ref], 
+                                       excluded_phrases, included_phrases, glossary_upload, debug, guidance_prompt, *struct_ref], 
                      outputs=[md_output, json_output])
         # advanced.click(show_advanced, inputs=[], outputs=[model, temperature])
         
@@ -731,4 +746,8 @@ def authf(username, password):
 
 if __name__ == '__main__':
     # demo.launch(server_name="0.0.0.0", auth=authf)
+    with open("formatted_prompts.txt", "w") as f:
+        f.writelines('"write_struct_copy": "' + str2escaped(struct_copy_prompt).strip() + '",')
+        f.writelines('"write_copy": "' + str2escaped(copy_prompt).strip() + '",')
+        f.writelines('"write_struct": "' + str2escaped(struct_prompt).strip() + '",')
     demo.launch()