Google Flights MCP Server

# Flight and Hotel API Fixes ## Issues Fixed ### 1. confirm_flight_price - Missing Mandatory Fields Error **Problem:** ``` Error 400: MANDATORY DATA MISSING Missing fields: travelerPricings, source, segment IDs Cause: Using simplified flight offer format instead of complete raw data ``` The flight pricing confirmation API (`/v1/shopping/flight-offers/pricing`) requires the COMPLETE raw flight offer data from the Amadeus API, but users were passing the simplified summary format returned by `search_flights`. **Root Cause:** - `search_flights` was returning only a simplified summary for readability - The simplified format was missing mandatory fields required by the pricing API: - `travelerPricings` - passenger pricing breakdown - `source` - offer source identifier (e.g., "GDS") - Segment `id` fields - unique identifiers for flight segments - Users didn't know they needed the complete raw data, not the summary **Solution:** Enhanced both `search_flights` and `confirm_flight_price`: 1. **search_flights now returns BOTH formats:** - `offers` - Simplified summary for easy reading - `raw_offers` - Complete raw data for use with confirm_flight_price 2. **confirm_flight_price validates input:** - Checks for missing mandatory fields - Provides clear error message if simplified data is used - Shows example of correct usage - Still automatically removes aircraft codes to prevent validation errors **Code Changes:** ```python # search_flights now includes raw offers summary = { "offers": [...], # Simplified summary "raw_offers": offers[:max_results] # Complete raw data } # confirm_flight_price validates input if "travelerPricings" not in offer or "source" not in offer: return helpful_error_message_with_solution() ``` **Updated Functions:** - `search_flights()` returns both simplified and raw offer data - `confirm_flight_price()` validates input format and provides helpful errors - Documentation updated with clear usage examples **Result:** ✅ Users have access to both simplified and complete flight offer data ✅ Clear error messages when wrong format is used ✅ Easy-to-follow examples in documentation ✅ Automatic aircraft code sanitization still works --- ### 2. get_hotel_ratings - Maximum Properties Limit Error **Problem:** ``` Error: "Number of properties exceeds maximum allowed" OR: "Properties not found in database" Cause: Incomplete/separated database, too many hotel IDs queried ``` The hotel sentiments API (`/v2/e-reputation/hotel-sentiments`) has: 1. A limit on how many hotels can be queried at once 2. An incomplete database - not all hotels have sentiment data **Root Cause:** - API limits the number of hotel IDs that can be queried in a single request - No validation was performed on the number of hotel IDs - No helpful error messages when hotels aren't in the sentiment database - Users didn't know if the error was due to too many IDs or missing data **Solution:** Enhanced `get_hotel_ratings()` with: 1. **Input Validation:** - Parses and validates comma-separated hotel IDs - Trims whitespace and converts to uppercase - Checks for empty input 2. **Limit Enforcement:** - Hard limit: 10 hotels per request - Recommended: 1-3 hotels per request - Returns clear error if limit exceeded 3. **Better Error Messages:** - Explains why no results were returned - Suggests using hotel IDs from search results - Lists which hotel IDs were requested - Differentiates between "too many IDs" and "no data found" 4. **Result Validation:** - Checks if any data was returned - Reports how many hotels had ratings vs. how many were requested - Provides helpful suggestions **Code Changes:** ```python @mcp.tool() async def get_hotel_ratings(hotel_ids: str) -> str: """ Get hotel ratings based on sentiment analysis of reviews. The API has a limit on the number of hotels that can be queried at once. For best results, query 1-3 hotels at a time. Args: hotel_ids: Comma-separated hotel IDs (max recommended: 3 hotels) """ # Input validation ids_list = [id.strip().upper() for id in hotel_ids.split(",") if id.strip()] # Limit checking if len(ids_list) > 10: return error_message_with_limit # Result validation if not data: return helpful_error_with_suggestions ``` **Result:** ✅ Clear error when too many hotel IDs provided ✅ Helpful guidance on limits (1-3 recommended, 10 max) ✅ Better error messages explaining why no results ✅ Suggestions to use IDs from search results ✅ Reports success rate (e.g., "2 out of 3 hotels found") --- ## Usage Examples ### Using confirm_flight_price (with validation and auto-sanitization) ```python # Get flight offers search_result = search_flights("JFK", "LAX", "2024-12-15") result = json.loads(search_result) # WRONG - Using simplified format (missing required fields) # first_offer = result["offers"][0] # ❌ This will fail! # CORRECT - Using complete raw format first_raw_offer = result["raw_offers"][0] # ✅ Has all required fields confirmed = confirm_flight_price(json.dumps(first_raw_offer)) # - Validates that all required fields are present # - Automatically removes invalid aircraft codes # - Returns confirmed pricing with tax details ``` ### Using get_hotel_ratings (with validation) ```python # GOOD: Query 1-3 hotels ratings = get_hotel_ratings("MCLONGHM,ADNYCCTB") # ✅ Works well # WARNING: Query 4-10 hotels ratings = get_hotel_ratings("ID1,ID2,ID3,ID4,ID5,ID6,ID7") # ⚠️ Works but some may not have data # ERROR: Query too many hotels ratings = get_hotel_ratings("ID1,ID2,ID3,...,ID15") # ❌ Returns clear error: "Too many hotel IDs (15). Please query 10 or fewer." # NO DATA: Invalid hotel IDs ratings = get_hotel_ratings("INVALID,BADID") # Returns helpful error with suggestions ``` --- ## Files Modified 1. **src/mcp_server_amadeus/server.py** - Added `sanitize_flight_offer_for_pricing()` function (lines 320-356) - Updated `confirm_flight_price()` to use sanitization (lines 359-405) - Enhanced `get_hotel_ratings()` with validation and better errors (lines 942-1005) 2. **FLIGHT_HOTEL_API_FIXES.md** (this file) - Complete documentation of both fixes --- ## Testing Recommendations ### Test confirm_flight_price ```bash # Test with complete raw offer data search_result = search_flights("JFK", "LAX", "2024-12-20") result = json.loads(search_result) # Test with raw offer (should work) confirm_flight_price(json.dumps(result["raw_offers"][0])) # ✅ Should succeed # Test with simplified offer (should give helpful error) confirm_flight_price(json.dumps(result["offers"][0])) # ❌ Should explain what's wrong ``` ### Test get_hotel_ratings ```bash # Test valid hotels (1-3) get_hotel_ratings("MCLONGHM") # Should work # Test multiple hotels get_hotel_ratings("MCLONGHM,ADNYCCTB,ELONMFS") # Should work # Test too many get_hotel_ratings("ID1,ID2,ID3,ID4,ID5,ID6,ID7,ID8,ID9,ID10,ID11") # Should error # Test invalid IDs get_hotel_ratings("INVALID,BADID") # Should give helpful error ``` --- ## Impact **confirm_flight_price:** - **Before:** Failed with Error 400 "MANDATORY DATA MISSING" when using simplified offer format - **After:** - `search_flights` now returns both simplified and raw offer data - Validates input and provides clear error if wrong format used - Automatically sanitizes aircraft codes - Succeeds with proper raw offer data **get_hotel_ratings:** - **Before:** Confusing errors about limits and missing data - **After:** Clear limits, helpful error messages, guidance on usage Both tools are now more robust and user-friendly!