dotanminh/travel-optimization-engine/flight-search

Flight Search - Tìm Route Combination Ẩn

Purpose

Find the best flight options by searching across multiple data sources, including virtual interlining combinations that standard search engines don't show. This is where the biggest savings happen - a combination of two separate tickets from different airlines can be 30-40% cheaper than any single-airline itinerary.

Input Required

From orchestrator:

• origin, destination (IATA codes)
• departure_date, return_date
• passengers (count + types)
• cabin_class
• preferred_airlines, avoided_airlines (optional)
• optimal_dates from date-optimization (if available)

Execution Steps

Step 1: Determine Search Strategy

Based on route type, choose data sources:

Step 2: AI-Knowledge Search

For each route, generate estimated options using airline knowledge:

Known carrier pricing tiers (estimate ranges):

Adjust estimates based on:

• Season and day-of-week (from date-optimization)
• Route competition level (more carriers = lower prices)
• Advance purchase timing

Step 3: Virtual Interlining Analysis

This is the KEY differentiator. Identify potential virtual interlining combinations:

Pattern: Route A→B has no cheap direct option, but:

• Carrier X has cheap A→C (transit hub)
• Carrier Y has cheap C→B
• Combined: A→C + C→B on separate tickets = cheaper than any A→B single ticket

Common virtual interlining opportunities:

Virtual interlining risks to flag:

• No protection if first flight is delayed (buy travel insurance)
• Separate check-in required at transit point
• Must self-transfer baggage
• Need sufficient layover time (minimum 3 hours international, 2 hours domestic)

Step 4: Normalize and Deduplicate Results

For ALL options regardless of source:

1. Price normalization: Ensure every price includes:

• Base fare
• Mandatory taxes and surcharges
• Fuel surcharge
• Airport fees
• Currency conversion to user's preferred currency

1. Deduplication: If same flight appears from multiple sources:

• Keep entry with lowest price
• Note the source in metadata

1. Tag each option:

• [DIRECT] - Point-to-point, no stops
• [1-STOP] - One connection, same airline/alliance
• [2-STOP] - Two connections
• [INTERLINING] - Virtual interlining (separate tickets)
• [LCC] - Low-cost carrier
• [LEGACY] - Full-service carrier

Step 5: Sort and Present

Sort by true_total_price (ascending). Present top 10:

#  Route                    Airlines        Adv.Price  True Total  Tags
1. HAN→ICN→SFO             VJ+UA           $520       $595        [INTERLINING][LCC+LEGACY]
2. HAN→NRT→SFO             VN+UA           $680       $710        [1-STOP][LEGACY]
3. HAN→BKK→SFO             VJ+CX+UA        $490       $620        [2-STOP][INTERLINING]
4. HAN→SFO                 VN              $1,100     $1,100      [DIRECT][LEGACY]
5. HAN→TPE→SFO             VJ+EVA           $550       $630        [INTERLINING][LCC+LEGACY]

Highlight: "Options 1, 3, 5 are virtual interlining - NOT available on Google Flights or standard aggregators."

Step 6: Pass Forward

Send to downstream skills:

• flight_options[] → fee-analysis (for true cost calculation)
• baseline_direct → route-optimization (for comparison)
• fare_classes[] → flexibility-analysis

API-Enhanced Mode (Kiwi Tequila)

When API is configured, execute:

1. Read references/api-integration.md for setup
2. Run ../../scripts/kiwi_client.py with parameters
3. Merge API results with AI-knowledge estimates
4. Real prices override estimates for matching routes

Important Notes

• Virtual interlining is NOT codeshare - user buys separate tickets
• Always warn: "If flight 1 is delayed, flight 2 is NOT protected"
• Recommend: "Consider travel insurance for virtual interlining itineraries"
• For families with children < 2: virtual interlining adds significant complexity