Travel Butler Tool Reference

Detailed parameter documentation for the travel butler's domain tools. Consult this reference when you need full parameter signatures, field semantics, or deduplication rules.

record_booking

Parse and persist a booking confirmation or update email payload into the trip container. Links the leg, accommodation, or reservation to the correct trip with full structured field extraction.

Parameters:

• payload (dict, required): Structured booking data extracted from the email or message:
  • provider (str): Carrier or provider name (e.g., "United Airlines", "Marriott")
  • type (str): Entity type — "leg" (flight), "accommodation", "reservation", "car_rental"
  • departure (str): IATA code for origin airport (legs only)
  • arrival (str): IATA code for destination airport (legs only)
  • departure_at (str): ISO 8601 datetime with timezone (TIMESTAMPTZ) for departure
  • arrival_at (str): ISO 8601 datetime with timezone for arrival
  • pnr (str): Airline record locator / booking reference
  • confirmation_number (str): Provider-specific confirmation number
  • seat (str): Assigned seat (e.g., "22A")
  • terminal (str): Departure terminal
  • gate (str): Departure gate (if known at booking time)
  • source_message_id (str): Email or message ID — used for deduplication; always pass this
  • metadata (dict): Any additional fields that don't fit the schema
• trip_id (UUID, optional): If known, link directly. If omitted, the tool attempts to match by date/destination and creates a new trip if no match is found.

Key rules:

• Always pass source_message_id — the tool layer uses this for deduplication. Do not manually check for duplicates.
• PNRs are correlation hints, not global unique keys. Always pair with provider context.
• Never create floating bookings. If no matching trip exists, create one first with trip_create, then call record_booking.
• For rebookings, delays, and gate/seat changes, use update_itinerary — not record_booking.

update_itinerary

Apply itinerary changes to an existing trip — time changes, cancellations, seat/gate reassignments, and rebookings. Always preserves prior values in metadata.prior_values.

Parameters:

• trip_id (UUID, required): The trip container to update
• patch (dict, required): Fields to update on the target entity:
  • leg_id (UUID): If updating a specific flight leg
  • accommodation_id (UUID): If updating an accommodation
  • Any mutable field: departure_at, arrival_at, seat, gate, terminal, status
• reason (str, required): Human-readable reason for the change (e.g., "UA email: flight delay notification")
• source_message_id (str, optional): Source email ID for audit trail

Key rules:

• Always use update_itinerary for mutations (not re-calling record_booking).
• Prior values are automatically written to metadata.prior_values — do not overwrite manually.
• When processing rebooking emails (subject signals: "gate change", "trip update", "delay notification", "rebooking"), default to update_itinerary.

list_trips

Query trip containers by lifecycle status and/or date window.

Parameters:

• status (str, optional): One of "planned", "active", "completed", "cancelled". Omit to return all statuses.
• departure_after (str, optional): ISO date — filter trips departing after this date
• departure_before (str, optional): ISO date — filter trips departing before this date
• limit (int, optional): Maximum results (default: 50)

Status transitions:
planned → active → completed. Direct cancellation (→ cancelled) is allowed from planned or active. Never transition backward (e.g., completed → active).

trip_summary

Return a normalized trip timeline with all linked legs, accommodations, reservations, and document pointers. The single source of truth for a trip's current state.

Parameters:

• trip_id (UUID, required): Trip to summarize
• include_documents (bool, optional): Include attached documents (default: False)
• include_timeline (bool, optional): Return legs and accommodations sorted by date (default: False)

Returns: Structured dict with:

• trip: Core trip metadata (id, destination, status, dates)
• legs: Flight legs sorted by departure time
• accommodations: Hotel/lodging sorted by check-in
• reservations: Other reservations (car, restaurant, etc.)
• documents: Attached documents (if include_documents=True)
• alerts: Outstanding pre-trip actions (missing documents, pending check-ins)

upcoming_travel

Surface upcoming departures and check-ins within a configurable window, with urgency-ranked pre-trip actions.

Parameters:

• within_days (int, required): Look-ahead window in days (e.g., 2 for 48-hour scan)
• include_pretrip_actions (bool, optional): Include urgency-ranked pre-trip action items (default: False)

Pre-trip action types:

• boarding_pass_missing: Boarding pass not attached
• checkin_pending: Online check-in window open but not completed
• seat_unassigned: No seat selected
• tight_layover: Connection time < 60 minutes
• hotel_missing: No accommodation for an upcoming night

add_document

Attach a travel document reference to an existing trip.

Parameters:

• trip_id (UUID, required): Trip to attach the document to
• type (str, required): Document type — "boarding_pass", "visa", "travel_insurance", "passport", "booking_confirmation", "receipt", "other"
• blob_ref (str, required): Reference to the stored file/attachment
• expiry_date (str, optional): ISO date — document expiry (for passports, visas, insurance)
• metadata (dict, optional): Additional context:
  • flight (str): Flight number (for boarding passes)
  • seat (str): Seat assignment
  • gate (str): Gate (for boarding passes)
  • policy_number (str): Insurance policy number
  • holder (str): Document holder name

Key rules:

• Boarding passes, visas, and insurance should always include expiry_date when known.
• Pass source_message_id in metadata when ingesting from email for provenance tracking.

Memory Classification — Travel Domain

Subject

• User preferences and identity: "user" or the user's name
• Airlines: airline name (e.g., "United Airlines", "Delta", "ANA")
• Hotel chains: chain name (e.g., "Marriott", "Hilton")
• Destinations: city or country (e.g., "Tokyo", "Japan")
• Airports: IATA code (e.g., "SFO", "NRT")

Predicates

• preferred_airline: Preferred carrier for domestic or international travel
• preferred_seat: Seat type preference (window, aisle, bulkhead)
• passport_nationality: Country of passport (ISO alpha-2 code)
• passport_expiry: Passport expiry date (for document expiry alerts)
• frequent_flyer: Loyalty program name and membership number
• hotel_preference: Preferred chain or room type
• travel_style: Budget, business, or luxury
• known_airport: Home airport IATA code
• dietary_preference: In-flight meal preference
• tsa_precheck: TSA PreCheck or Global Entry known traveler number

Permanence

• stable: Passport info, frequent flyer numbers, home airport, TSA/Global Entry numbers, long-standing preferences
• standard (default): Current trip context, active booking patterns, recent destination preferences
• volatile: Real-time flight status, gate changes, delay notifications, live check-in reminders

Tags

travel-preference, loyalty, passport, flight, hotel, document, reminder, delay

Example Facts

# From: "I always fly United when I can"
memory_store_fact(
    subject="user",
    predicate="preferred_airline",
    content="United Airlines for both domestic and international travel",
    permanence="stable",
    importance=7.0,
    tags=["travel-preference", "flight"]
)

# From: "My passport expires June 2028"
memory_store_fact(
    subject="user",
    predicate="passport_expiry",
    content="US passport expires 2028-06-14",
    permanence="stable",
    importance=9.0,
    tags=["passport", "document"]
)

# From: "I'm a United MileagePlus member, number UA-7382910"
memory_store_fact(
    subject="United Airlines",
    predicate="frequent_flyer",
    content="MileagePlus member number UA-7382910",
    permanence="stable",
    importance=8.0,
    tags=["loyalty", "flight"]
)

# From: "I always book window seats"
memory_store_fact(
    subject="user",
    predicate="preferred_seat",
    content="window seat",
    permanence="stable",
    importance=6.0,
    tags=["travel-preference", "flight"]
)

# From: "I'm flying out of SFO most of the time"
memory_store_fact(
    subject="user",
    predicate="known_airport",
    content="SFO — San Francisco International (home airport)",
    permanence="stable",
    importance=8.0,
    tags=["travel-preference", "flight"]
)