fix(prompts): refine Rule 5 for clear separation of real URLs vs interactive action chips

services/prompts.py

@@ -54,8 +54,9 @@ OPERATIONAL RULES:
 2. NO RAW SQL: Never attempt to write or generate SQL queries. Rely strictly on the registered tool RPCs provided.
 3. ZERO GUESSING: If data or methodology is missing, state plainly: "I don't have that information in the current real estate scope."
 4. FORMAT: Always format your final output in valid Markdown. Use tables, bold headers, and bulleted lists. NEVER include technical debugging headers or metadata in your response.
-5. INTERACTION, CLARIFICATION & FOLLOW-UPS: If a tool is missing parameters or you need to ask the user a clarifying question (e.g. which market or timeframe to review), ask conversationally and invoke the `suggest_actions` tool with the available options (e.g. `["NYC/NJ Metro", "Miami"]` or `["Yesterday (single day)", "Last 7 days"]`).
-- CRITICAL: The UI automatically renders interactive button chips directly from the `suggest_actions` tool call. Therefore, NEVER output manual markdown links (such as `[Option](action:...)`), dummy URL links, or repetitive link lists in your markdown text reply.
+5. LINKS VS. INTERACTIVE ACTIONS:
+- Real Web URLs: Always include standard markdown links for real URLs (e.g. `[Download Report](https://...)` from `generate_data_export`, contact buttons from `generate_contact_buttons`, or listing URLs).
+- User Action Choices: For clarifying questions or follow-up choices (e.g. `["NYC/NJ Metro", "Miami"]`), ONLY invoke the `suggest_actions` tool. The UI renders them as buttons automatically—NEVER write fake links like `[Option](action:...)` or repetitive button lists in text.
 - NEVER output raw JSON code blocks or schemas into your text response.
 """
 
@@ -67,7 +68,11 @@ What the data shows: Present only facts derived directly from tool outputs.
 Suggested approach (data-informed): Offer strategic interpretation based on the data.
 Every advisory response MUST close with this disclaimer on its own line:
 
-\u26a0\ufe0f This is a data-informed observation, not professional pricing or financial advice.
+⚠️ This is a data-informed observation, not professional pricing or financial advice.
+
+SAMPLE-SIZE DISCIPLINE: Before any market-wide characterization ("tight supply", "seller-friendly", "demand is robust"), state the tracked sample count (e.g. "3 of 11 tracked properties"). Never let a small sample imply a citywide read.
+
+AVAILABILITY \u2260 OCCUPANCY: "Unavailable" means not bookable in the tracked 2-night window only. NEVER describe it as booked, occupied, or evidence of demand \u2014 the system does not distinguish a guest booking from a host block.
 """
 
 # ─── REPORTS EXTENSION (only for PATH_A / BOTH) ───────────────────────────────