Entities

Adapter promotes recurring real-world things — people, companies, reservations, documents — into entities. Each entity has a fixed schema, a stable id, and is resolved across every evidence source it appears in. Entities are derived from evidence. A single person mentioned in an email, a calendar event, and a Slack message resolves to one person entity, with each mention preserved as a citation. Citations returned by Ask carry a type naming the kind of record each cited row is. Reservation entities are cited under a single reservation type, with the specific kind (one of the values below) in the row’s reservation_type field.

All entity_type values

[
  "actor",
  "deal",
  "document",
  "file",
  "organization",
  "person",
  "product",
  "purchase",
  "reservation",
  "table",
  "transaction"
]

All reservation_type values

[
  "car_rental",
  "dining",
  "event_reservation",
  "flight",
  "hotel_stay",
  "train"
]

All kind values

[
  "invoice",
  "order",
  "payment"
]

All entity types

Entity Type	Description
`actor`	A non-person participant in communication: a shared address that fans out to people, an automated sender, or a bookable resource
`deal`	The current belief about one pursuit — a sale, renewal or expansion the workspace is working
`document`	Searchable document/file entity with semantic search capabilities
`file`	A file or document ingested from a source system (Drive, Notion, attachments, etc.)
`organization`	Organization/company extracted from evidence
`person`	Individual person extracted from evidence
`product`	One real-world product or service (schema.org Product), whichever side of the exchange the user is on
`purchase`	Purchase/transaction entity
`reservation`	A booking extracted from a confirmation email — the specific kind is in `reservation_type`
`table`	Understanding of one table of structured data, independent of where it came from: a sheet of an uploaded workbook, a delimited file, or a table reached through a database connection
`transaction`	One exchange of money with a counterparty, or one billing period of a recurring charge — the orders, invoices and payments about it are listed in `documents`, each with its `kind`

Reservation kinds

Subtypes of reservation, carried in its reservation_type field:

`reservation_type`	Description
`car_rental`	Car rental reservation entity data
`dining`	Restaurant reservation entity data
`event_reservation`	Event/ticket reservation entity data
`flight`	Flight reservation entity representing a single itinerary (one-way, round-trip, or multi-city) with one or more legs, each containing one or more segments (connections)
`hotel_stay`	Hotel reservation entity data
`train`	Train reservation entity data

Transaction kinds

Subtypes of transaction, carried in its kind field:

`kind`	Description
`invoice`	A statement of money due as stated in one piece of evidence (schema.org Invoice): an invoice, a bill, a statement, or a payment-due reminder
`order`	An order as stated in one piece of evidence (schema.org Order): an order confirmation, a receipt, a purchase order, or a status update on one
`payment`	Money that moved from one party to another, as stated in one piece of evidence: a payment confirmation, an autopay charge, remittance advice, a refund, or a chargeback

What each entity carries

The summaries below describe the information you can expect each entity type to carry. Attributes may be added or removed in future releases.

Reservation

All bookings surface as a single reservation entity, with the specific kind in the reservation_type field. Every reservation carries the booking essentials when present: a confirmation code, the provider, start and end times (in the venue’s local timezone), a location, the named guests or travelers, the price, and the booking status. Each kind adds its own detail on top:

car_rental — Rental company, pickup and drop-off times and locations, vehicle class, renter name, and total price.
dining — Restaurant name and location, reservation time, party size, the name on the reservation, and cuisine type.
event_reservation — Event name, venue and address, start/end times, attendee name, seat or section, ticket count, and total price.
flight — The full itinerary: airline, flight numbers, departure/arrival airports and times for every leg and connection, every passenger on the booking, trip type (one-way, round-trip, multi-city), and total price.
hotel_stay — Property name and address, check-in/check-out times, guest name, room type, number of nights, and total price.
train — Operator, departure and arrival stations and times, passenger name, seat and travel class, and ticket price.

Transaction

Orders, invoices and payments surface as a single transaction: every document about the same exchange folds into it, so it carries one amount, the counterparty, the products on its lines, and a status for the money (pending, settled, partially settled, failed, cancelled, refunded or void). Each document stays on the transaction’s documents timeline with its kind:

invoice — A statement of money due as stated in one piece of evidence (schema.org Invoice): an invoice, a bill, a statement, or a payment-due reminder
order — An order as stated in one piece of evidence (schema.org Order): an order confirmation, a receipt, a purchase order, or a status update on one
payment — Money that moved from one party to another, as stated in one piece of evidence: a payment confirmation, an autopay charge, remittance advice, a refund, or a chargeback

Actor

A non-person participant in communication: a shared address that fans out to people, an automated sender, or a bookable resource

Deal

The current belief about one pursuit — a sale, renewal or expansion the workspace is working

Document

Title, document type, searchable text content with a preview, and file metadata (format, size, source location).

File

A file or document ingested from a source system (Drive, Notion, attachments, etc.)

Organization

Name and legal name, web domain, contact details, address, industry, parent organization, and public identifiers when available.

Person

Names (including given/family name variants), email addresses, phone numbers, job title, employer and other affiliations, linked profiles, and when they were first and last seen in your data.

Product

One real-world product or service (schema.org Product), whichever side of the exchange the user is on

Purchase

Order number, seller, line items, total amount and currency, order and delivery dates, fulfillment type, and tracking number.

Table

Understanding of one table of structured data, independent of where it came from: a sheet of an uploaded workbook, a delimited file, or a table reached through a database connection

Getting Started

Core Concepts

Triggers

All entity types

Reservation kinds

Transaction kinds

What each entity carries

Reservation

Transaction

Actor

Deal

Document

File

Organization

Person

Product

Purchase

Table

​All entity types

​Reservation kinds

​Transaction kinds

​What each entity carries

​Reservation

​Transaction

​Actor

​Deal

​Document

​File

​Organization

​Person

​Product

​Purchase

​Table

All entity types

Reservation kinds

Transaction kinds

What each entity carries

Reservation

Transaction

Actor

Deal

Document

File

Organization

Person

Product

Purchase

Table