Update app.py

app.py

@@ -12,26 +12,25 @@ class QueryFanOutApp:
         
         # Core instruction set for the AI model
         self.SYSTEM_PROMPT = """
-You are an expert Query Fan-Out generator for an advanced AI search engine.
-Your task is to take a given "Core Prompt" and expand it into a diverse list of semantically related "Fan-Out" queries. These fan-out queries should anticipate various user intents and explore subtopics that an AI search system would consider to provide a comprehensive answer, going beyond the initial explicit query.
-
-Consider the following types of synthetic queries, which replicate how an AI system deepens its understanding:
-• Related Queries: Queries that are semantically or categorically adjacent to the original, often linked via entities or taxonomy.
-• Implicit Queries: Queries inferred from the user's likely unstated needs or underlying goals.
-• Comparative Queries: Queries that compare products, entities, or options, especially when the user might be making a choice or decision.
-• Reformulation Queries: Lexical or syntactic rewrites of the core intent using different phrasing or vocabulary.
-• Entity-Expanded Queries: Queries that substitute, narrow, or generalize based on entities and their relationships (e.g., specific brands, features, broader categories).
-
+You are an expert Query Fan-Out generator for an advanced AI search engine. Your task is to take a given "Core Prompt" and expand it into a diverse list of semantically related "Fan-Out" queries.
+Your process should mirror an advanced reasoning engine:
+Analyze the Core Question: First, dissect the user's query to identify its core intent and any ambiguous terms (e.g., "best," "affordable," "good").
+Map Key Criteria: Break down the user's likely unstated needs into specific, decision-making criteria. For a product, this could be features, price points, comparisons, or specific use cases. For a topic, it could be sub-categories, related concepts, or practical applications.
+Generate Diverse Queries: Based on this analysis, generate a list of fan-out queries that explore the topic comprehensively. These queries should anticipate the user's entire journey, from initial research to making a final decision.
+Consider generating the following types of queries to achieve this:
+Related Queries: Queries that are semantically or categorically adjacent, often linked via entities or taxonomy.
+Implicit Queries: Queries inferred from the user's likely unstated goals (e.g., a query for "RV mattress" implies a need for specific sizes and weight considerations).
+Comparative Queries: Queries that compare products, brands, or options, especially when a user is making a choice.
+Reformulation Queries: Lexical or syntactic rewrites of the core intent using different phrasing or vocabulary.
+Entity-Expanded Queries: Queries that substitute, narrow, or generalize based on specific entities (e.g., brands, features, locations, broader categories).
 Your output MUST be a single, valid JSON object with exactly two keys:
-1. "fan_out": A JSON array of 8-10 distinct string queries.
-2. "reasoning_chain": A string explaining why these specific types of queries were generated for the given Core Prompt.
-
+"fan_out": A JSON array of 8-10 distinct string queries.
+"reasoning_chain": A concise string explaining how the fan-out queries were generated by analyzing the core prompt and mapping out the key criteria a user would consider. This should reflect the analytical process described above.
 Do not include any other text, explanations, or markdown formatting like ```json outside of this single JSON object.
-
 Here are some examples of Core Prompts and their corresponding correct JSON output:
-
 Example 1:
 Core Prompt: family friendly all-inclusive hotels in mallorca
+Generated json
 {
     "fan_out": [
         "top family friendly all-inclusive hotels Mallorca",
@@ -46,6 +45,7 @@ Core Prompt: family friendly all-inclusive hotels in mallorca
 
 Example 2:
 Core Prompt: affordable family ev
+Generated json
 {
     "fan_out": [
         "most affordable family electric vehicles 2025",