Update chatbot_prompts.py

chatbot_prompts.py

@@ -1,82 +1,70 @@
 # chatbot_prompts.py
 import logging
 
-def get_initial_insight_and_suggestions(plot_id: str, plot_label: str, plot_data_summary: str = None):
+def get_initial_insight_prompt_and_suggestions(plot_id: str, plot_label: str, plot_data_summary: str = None):
     """
-    Generates an initial insight (now including a data summary) and suggested questions for a given plot.
+    Generates an initial prompt for the LLM to provide insights on a plot and suggested questions.
     Args:
         plot_id (str): The unique identifier for the plot.
         plot_label (str): The display label for the plot.
         plot_data_summary (str, optional): A textual summary of the data for the plot.
     Returns:
-        tuple: (initial_chat_message_dict, list_of_suggestion_strings)
+        tuple: (prompt_for_llm_str, list_of_suggestion_strings)
     """
-    logging.info(f"Generating initial insight for plot_id: {plot_id}, label: {plot_label}")
-    # logging.debug(f"Data summary for initial insight of '{plot_label}':\n{plot_data_summary}")
+    logging.info(f"Generating initial insight prompt for plot_id: {plot_id}, label: {plot_label}")
 
+    base_persona_prompt = "You are an expert in Employer Branding and LinkedIn social media strategy. Analyze the following data for the chart '{plot_label}' and provide key insights and actionable advice. Focus on interpreting the provided data snapshot."
+    prompt_text = f"{base_persona_prompt.format(plot_label=plot_label)}\n\n"
 
-    base_persona = "As an expert in Employer Branding and LinkedIn social media strategy, here's an insight on your "
-    insight_text = f"{base_persona}**{plot_label}**:\n\n"
-    
-    # Include the data summary if available
-    if plot_data_summary and plot_data_summary.strip() and "No data summary available" not in plot_data_summary and "Error generating data summary" not in plot_data_summary and "Accesso negato" not in plot_data_summary:
-        insight_text += f"Here is a snapshot of the data for your '{plot_label}' chart:\n```text\n{plot_data_summary}\n```\n\n"
+    if plot_data_summary and plot_data_summary.strip() and \
+       "No data summary available" not in plot_data_summary and \
+       "Error generating data summary" not in plot_data_summary and \
+       "Accesso negato" not in plot_data_summary and \
+       f"Nessun sommario dati specifico disponibile per '{plot_label}'" not in plot_data_summary : # Added check for this specific no data message
+        prompt_text += f"Data Snapshot for '{plot_label}':\n```text\n{plot_data_summary}\n```\n\n"
+        prompt_text += "Based on this data and your expertise, what are the most important observations and what steps can be taken to improve or capitalize on these trends? Please provide a concise initial analysis."
     else:
-        insight_text += f"Currently, no specific data snapshot is available for '{plot_label}' to include here. General insights apply.\n\n"
+        prompt_text += f"No specific data snapshot is available for '{plot_label}'. Provide general insights and advice for improving performance related to '{plot_label}' on LinkedIn, assuming typical scenarios. Please provide a concise initial analysis."
 
-    # Default suggestions - these can be made more dynamic later if needed
+    # Default suggestions
     suggestions = [
         f"What are the key drivers for {plot_label.lower()} based on the data?",
         f"How can I improve my {plot_label.lower()} according to these trends?",
         f"What does good performance look like for {plot_label.lower()}?"
     ]
 
-    # Customize insights and suggestions per plot_id
-    # These specific insights can now refer to the data summary provided above.
+    # Customize suggestions per plot_id
     if plot_id == "followers_count":
-        insight_text += ("Tracking your 'Numero di Follower nel Tempo' is crucial. A steady increase, as potentially shown in the data snapshot, indicates growing brand reach and influence. "
-                         "If you observe stagnation or decline, it might be time to review your content strategy, posting frequency, or engagement tactics with your target audience. "
-                         "Consider running campaigns or promoting your LinkedIn page on other channels.")
         suggestions = [
             "Based on the follower data, what was our peak growth period?",
-            "How often should I post to maximize follower growth according to LinkedIn best practices?",
-            "What content typically resonates most with potential followers on LinkedIn?"
+            "How often should I post to maximize follower growth?",
+            "What content typically resonates most with potential followers?"
         ]
     elif plot_id == "engagement_rate":
-        insight_text += ("Your 'Tasso di Engagement nel Tempo' is a vital indicator of how compelling your audience finds your content. Higher rates, reflected in the data, signify that your posts are sparking meaningful interactions. "
-                         "If engagement is low, re-evaluate content relevance, the types of media used, and the clarity of your calls to action. Experiment with questions, polls, and interactive content.")
         suggestions = [
             "What does the engagement trend tell us about recent content performance?",
-            "What types of posts typically get the highest engagement on LinkedIn?",
-            "Can you give me examples of strong calls to action for LinkedIn posts?"
+            "What types of posts typically get the highest engagement?",
+            "Can you give me examples of strong calls to action?"
         ]
     elif plot_id == "reach_over_time":
-        insight_text += ("The 'Copertura nel Tempo' (Reach) shows how many unique LinkedIn members are seeing your posts. Expanding reach is fundamental for brand awareness. "
-                         "If this metric is flat or declining in the provided data, explore content sharing strategies, encourage employees to share company posts, utilize relevant hashtags, and consider targeted LinkedIn advertising.")
         suggestions = [
             "What does the reach data suggest about our content visibility?",
-            "What are effective organic strategies to increase post reach on LinkedIn?",
-            "How do hashtags and tagging strategies impact reach on LinkedIn?"
+            "What are effective organic strategies to increase post reach?",
+            "How do hashtags and tagging strategies impact reach?"
         ]
     elif plot_id == "impressions_over_time":
-        insight_text += ("'Visualizzazioni nel Tempo' (Impressions) indicates the total number of times your posts have been seen. "
-                         "High impressions are good, but also analyze them in conjunction with engagement (check the engagement rate chart) to ensure visibility translates to interaction.")
         suggestions = [
             "How do current impressions compare to previous periods based on the data?",
-            "What's the difference between reach and impressions, and which is more important?",
-            "Does LinkedIn's algorithm favor certain types of content for impressions?"
+            "What's the difference between reach and impressions?",
+            "Does LinkedIn's algorithm favor certain content for impressions?"
         ]
     elif plot_id == "comments_sentiment":
-        insight_text += ("Analyzing the 'Ripartizione Commenti per Sentiment' provides qualitative insights. The data snapshot should show the distribution. "
-                         "A high volume of positive sentiment is ideal. Pay close attention to neutral and negative comments to understand concerns or areas for improvement.")
         suggestions = [
             "What does the sentiment breakdown indicate about audience perception?",
-            "How can I encourage more positive comments on my posts?",
-            "What's the best way to respond to negative comments effectively?"
+            "How can I encourage more positive comments?",
+            "What's the best way to respond to negative comments?"
         ]
-    else: # Default fallback for other plots
-        insight_text += (f"This chart on '{plot_label}' provides valuable data. Analyzing its trends, peaks, and troughs, as seen in the snapshot, can help you refine your LinkedIn strategy. "
-                         f"Look for patterns and correlate them with your activities or external events to understand performance drivers.")
+    # Add more plot_id specific suggestions if needed
 
     # Ensure exactly 3 suggestions
     while len(suggestions) < 3:
@@ -84,5 +72,4 @@ def get_initial_insight_and_suggestions(plot_id: str, plot_label: str, plot_data
     if len(suggestions) > 3:
         suggestions = suggestions[:3]
 
-    initial_chat_message = {"role": "assistant", "content": insight_text.strip()}
-    return initial_chat_message, suggestions
+    return prompt_text, suggestions