# =============================================================================
# MENU GUIDE TOOL (Deterministic / Data-Only)
# =============================================================================
from typing import List, Dict, Any
import json
import logging
from utils import fetch_all_tenant_contexts
logger = logging.getLogger(__name__)
def execute_menu_guide(query: str, tenant_key: str, config: dict = None) -> str:
"""
Returns the full menu context for the given tenant.
The calling AI (MCP Client) is responsible for filtering/finding items based on the user's query.
"""
if config is not None:
config["phase"] = "MENU_BROWSING"
# 1. Fetch Data
tenant_contexts = fetch_all_tenant_contexts()
tenant_data = tenant_contexts.get(tenant_key, {})
menu_data = tenant_data.get("menu", [])
if not menu_data:
return "No menu data available for this restaurant."
# 2. Format Menu for Raw Output
# We provide a clean list of items so the LLM can "read" the menu.
menu_summary = []
for m in menu_data:
name = m.get('name') or m.get('item_name', 'Unknown')
price = m.get('price', 0)
is_veg = m.get('is_veg', False)
category = m.get('category', 'General')
menu_id = m.get('menu_id')
menu_summary.append(f"- [{category}] {name} (₹{price}) {'(Veg)' if is_veg else '(Non-Veg)'} [ID: {menu_id}]")
# 3. Contextual Data
current_category = config.get("current_category", "None") if config else "None"
# Return raw text block
response = [
f"CONTEXT: User is strictly browsing. Current Category focus: {current_category}",
"--- MENU DATA START ---",
"\n".join(menu_summary),
"--- MENU DATA END ---",
"INSTRUCTION: You are the intelligent assistant. Use the above menu to answer the user's request."
]
return "\n".join(response)
