Update the prompts

app.py

@@ -212,7 +212,10 @@ def set_up_chat_ui():
 
             # Update the progress bar
             progress_percentage = min(len(response) / APPROX_TARGET_LENGTH, 0.95)
-            progress_bar.progress(progress_percentage, text='Streaming content...')
+            progress_bar.progress(
+                progress_percentage,
+                text='Streaming content...this might take a while...'
+            )
 
         history.add_user_message(prompt)
         history.add_ai_message(response)
@@ -226,7 +229,7 @@ def set_up_chat_ui():
             'Cleaned JSON response:: original length: %d | cleaned length: %d',
             len(response), len(response_cleaned)
         )
-        logger.debug('Cleaned JSON: %s', response_cleaned)
+        # logger.debug('Cleaned JSON: %s', response_cleaned)
 
         # Now create the PPT file
         progress_bar.progress(

langchain_templates/chat_prompts/initial_template_v4_two_cols_img.txt

@@ -1,24 +1,25 @@
-You are a helpful, intelligent chatbot. You are experienced with PowerPoint.
+You are a helpful, intelligent assistant. You are experienced with PowerPoint.
 
 Create the slides for a presentation on the given topic.
 Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
-If relevant, add one or two EXAMPLES to illustrate the concept.
+Add relevant, detailed content to each slide. When relevant, add one or two EXAMPLES to illustrate the concept.
 For two or three important slides, generate the key message that those slides convey.
 
-Identify if a slide describes a step-by-step/sequential process, then begin the bullet points with a special marker >>. Limit this to max two or three slides.
+Identify if a slide describes a step-by-step/sequential process, then begin the bullet points with a special marker >>.
+Limit this to max two or three slides.
+
 Also, add at least one slide with a double column layout by generating appropriate content based on the description in the JSON schema provided below.
 In addition, for each slide, add image keywords based on the content of the respective slides.
 These keywords will be later used to search for images from the Web relevant to the slide content.
 
-Create one slide with rich visual experience, containing icons (pictograms) and short texts.
-This particular slide will illustrate some key ideas/aspects/concepts relevant to the topic.
-Here, you will generate 4 TO 6 short lines of text (hence, select 4 TO 6 icon names).
-Each such line will have the name of an icon enclosed between [[ and ]].
-The name of the icons should be relevant to the aspects described and MUST BE chosen from the <ICONS> section provided below.
-You must specify the name of the icon exactly as mentioned in <ICONS>.
-You MUST NEVER generate any icon name not mentioned in the <ICONS> section.
+In addition, create one slide containing 4 TO 6 icons (pictograms) illustrating some key ideas/aspects/concepts relevant to the topic.
+In this slide, each line of text will begin with the name of an icon enclosed between [[ and ]].
+The name of an icon MUST BE chosen exactly as it is from the <ICONS> section provided below and should be relevant to the aspects described.
+If you need an icon that is unavailable in the <ICONS> section, you may select a conceptually similar icon's name from <ICONS>.
+For example, if an icon for "neural network" is unavailable, you may choose the "deep-learning" icon from <ICONS>.
+However, you MUST NEVER generate any icon name not mentioned in the <ICONS> section.
+
+The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
 
 ALWAYS add a concluding slide at the end, containing a list of the key takeaways and an optional call-to-action if relevant to the context.
 Unless explicitly instructed, create 10 TO 12 SLIDES in total.
@@ -63,9 +64,10 @@ The output must be only a valid and syntactically correct JSON adhering to the f
         {{
             "heading": "A slide illustrating key ideas/aspects/concepts (Hint: generate an appropriate heading)",
             "bullet_points": [
-                "[[icon_name]] Some short text",
+                "[[icon_name]] Some text",
                 "[[another_icon_name]] Some words describing this aspect",
-                "[[icon_other_name]] Another aspect highlighted in a few words",
+                "[[icon_other_name]] Another aspect highlighted here",
+                "[[an_icon]] Another point here",
             ],
             "key_message": "",
             "img_keywords": ""

langchain_templates/chat_prompts/refinement_template_v4_two_cols_img.txt

@@ -1,12 +1,11 @@
-You are a helpful, intelligent chatbot. You are experienced with PowerPoint.
+You are a helpful, intelligent assistant. You are experienced with PowerPoint.
 
 A list of user instructions is provided below in sequential order -- from the oldest to the latest.
 The previously generated content of the slide deck in JSON format is also provided.
 Follow the instructions to revise the content of the previously generated slides of the presentation on the given topic.
+You will not repeat any slide.
 Include main headings for each slide, detailed bullet points for each slide.
-Add relevant content to each slide.
-The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
-If relevant, add one or two EXAMPLES to illustrate the concept.
+Add relevant, detailed content to each slide. When relevant, add one or two EXAMPLES to illustrate the concept.
 For two or three important slides, generate the key message that those slides convey.
 
 Identify if a slide describes a step-by-step/sequential process, then begin the bullet points with a special marker >>. Limit this to max two or three slides.
@@ -14,13 +13,14 @@ Also, add at least one slide with a double column layout by generating appropria
 In addition, for each slide, add image keywords based on the content of the respective slides.
 These keywords will be later used to search for images from the Web relevant to the slide content.
 
-Create one slide with rich visual experience, containing icons (pictograms) and short texts.
-This particular slide will illustrate some key ideas/aspects/concepts relevant to the topic.
-Here, you will generate 4 TO 6 short lines of text (hence, select 4 TO 6 icon names).
-Each such line will have the name of an icon enclosed between [[ and ]].
-The name of the icons should be relevant to the aspects described and MUST BE chosen from the <ICONS> section provided below.
-You must specify the name of the icon exactly as mentioned in <ICONS>.
-You MUST NEVER generate any icon name not mentioned in the <ICONS> section.
+In addition, create one slide containing 4 TO 6 icons (pictograms) illustrating some key ideas/aspects/concepts relevant to the topic.
+In this slide, each line of text will begin with the name of an icon enclosed between [[ and ]].
+The name of an icon MUST BE chosen exactly as it is from the <ICONS> section provided below and should be relevant to the aspects described.
+If you need an icon that is unavailable in the <ICONS> section, you may select a conceptually similar icon's name from <ICONS>.
+For example, if an icon for "neural network" is unavailable, you may choose the "deep-learning" icon from <ICONS>.
+However, you MUST NEVER generate any icon name not mentioned in the <ICONS> section.
+
+The content of each slide should be VERBOSE, DESCRIPTIVE, and very DETAILED.
 
 ALWAYS add a concluding slide at the end, containing a list of the key takeaways and an optional call-to-action if relevant to the context.
 Unless explicitly instructed, create 10 TO 12 SLIDES in total.
@@ -69,9 +69,10 @@ The output must be only a valid and syntactically correct JSON adhering to the f
         {{
             "heading": "A slide illustrating key ideas/aspects/concepts (Hint: generate an appropriate heading)",
             "bullet_points": [
-                "[[icon_name]] Some short text",
+                "[[icon_name]] Some text",
                 "[[another_icon_name]] Some words describing this aspect",
-                "[[icon_other_name]] Another aspect highlighted in a few words",
+                "[[icon_other_name]] Another aspect highlighted here",
+                "[[an_icon]] Another point here",
             ],
             "key_message": "",
             "img_keywords": ""