Skip to main content

AI-Powered Listing Search

Natural language search using vector embeddings + LLM. Combines with structured filters.

AI extracts intent from natural language, merges with explicit filters, runs full listing search.

POST /api/v1/crs/listings/ai-search

AI only:

{
  "query": "Pet-friendly villas in Goa for 6 adults this weekend under 20k"
}

AI + explicit filters (explicit wins on conflict):

{
  "query": "luxury beachfront villa with pool",
  "rawParams": [
    { "filterName": "channelId", "filterValues": ["B2C"] },
    { "filterName": "city", "filterValues": ["Goa"] },
    { "filterName": "adults", "filterValues": ["6"] }
  ]
}

Response:

{
  "intent": {
    "listingIds": ["lst_abc", "lst_def", "lst_ghi"],
    "checkInDate": "2026-04-05",
    "checkOutDate": "2026-04-07",
    "adultCount": 6,
    "city": "Goa",
    "propertyType": "Villa",
    "budgetPerNight": 20000,
    "petFriendly": true,
    "originalQuery": "Pet-friendly villas in Goa for 6 adults this weekend under 20k"
  },
  "results": {
    "singleListings": [ ... ],
    "combinations": []
  }
}

Extract Intent Only​

Populate filter UI from natural language — no listing search executed.

POST /api/v1/crs/listings/ai-search/intent
{ "query": "3 BHK villa in Lonavala for 8 people next weekend" }

Response:

{
  "listingIds": ["lst_abc", "lst_def"],
  "checkInDate": "2026-04-05",
  "checkOutDate": "2026-04-07",
  "adultCount": 8,
  "city": "Lonavala",
  "propertyType": "Villa",
  "originalQuery": "3 BHK villa in Lonavala for 8 people next weekend"
}
Use Case

User types in the search bar → frontend calls /ai-search/intent → auto-fills the filter chips (city, dates, guests, property type). User can then adjust filters manually before searching.

Caching​

Same query on the same day returns cached results (no tokens burned):

  • Key: 2026-04-03:pet-friendly villas in goa this weekend
  • TTL: 10 minutes
  • Why date in key: "this weekend" means different dates on different days