text_compound_query_detection_prompt = Analyze if this query contains multiple distinct questions or requests that should be processed separately.
Query: {{query}}
A compound query contains multiple independent questions joined by connectors like:
- "and", "then", "also", "as well as", "plus", "in addition"
Examples of compound queries:
- "pending orders and total revenue" → 2 queries: pending orders, total revenue
- "show me sales then customer count" → 2 queries: sales, customer count
- "revenue this month and best selling products" → 2 queries: revenue, best selling products
Examples of NON-compound queries:
- "revenue by product and category" → 1 query (grouping, not separate questions)
- "orders with status pending and shipped" → 1 query (filter conditions)
- "customers who bought product A and product B" → 1 query (combined filter)
- "price of the Samsung S25 then its SKU and EAN" → 1 query (multiple ATTRIBUTES of ONE entity = columns, NOT separate questions)
- "name, price and stock of product X" → 1 query (several attributes of one entity)
- "total revenue and average order value this month" → 1 query (several METRICS over the same scope)
KEY RULE: multiple ATTRIBUTES/COLUMNS or METRICS of the SAME entity are NOT compound — they are a single query returning several columns. A query is compound ONLY when it asks separate questions about different entities or different topics (e.g. "revenue AND out-of-stock products"). Conditions/filters on the same set are also NOT compound.
Respond in JSON format:
```json
{
    "is_compound": true/false,
    "confidence": 0.0-1.0,
    "reasoning": "explanation",
    "sub_queries": ["query1", "query2"]
}
```

text_compound_query_split_prompt = Split this compound query into separate, independent sub-queries.
Original Query: {{query}}
Rules:
1. Each sub-query should be a complete, standalone NATURAL LANGUAGE question (NOT SQL!)
2. Preserve the original intent of each part
3. Add context if needed for clarity (e.g., time periods, filters)
4. Keep the same language as the original query
5. DO NOT generate SQL queries - generate natural language questions only
6. If the original query has a time period (e.g., "last 6 months"), include it in EACH sub-query

IMPORTANT: The "query" field must be a natural language question, NOT a SQL query!

Examples:
- Original: "pending orders and revenue last 6 months"
  Sub-queries:
  - "pending orders" (natural language)
  - "revenue for the last 6 months" (natural language with time context)

- Original: "commandes en attente et chiffre d'affaires mensuel"
  Sub-queries:
  - "commandes en attente" (natural language)
  - "chiffre d'affaires mensuel" (natural language)

Respond in JSON format:
```json
{
    "sub_queries": [
        {
            "label": "Short descriptive label in English",
            "query": "The complete sub-query as a NATURAL LANGUAGE question (NOT SQL!)",
            "original_part": "The part from original query"
        }
    ]
}
```

text_compound_result_header = Combined Results for Your Query
text_compound_result_separator = ---
text_compound_result_label_prefix = Result for:
text_compound_partial_success = Some parts of your query could not be processed. Here are the available results:
text_compound_all_failed = Unable to process any part of your compound query. Please try rephrasing or asking each question separately.
text_compound_sub_query_error = Error processing this part:

text_compound_format_prompt = Format these combined results from a compound query into a clear, readable response.
Original Query: {{query}}
Results: {{results}}
Create a well-structured response that:
1. Clearly separates each sub-query result
2. Uses appropriate headers/labels
3. Summarizes key findings
4. Notes any errors or missing data
Respond with the formatted text (not JSON).
