Struct & Type Casting - Sixtyfour AI Documentation

The struct parameter lists the fields you want back and, optionally, the type of each one, so People Intelligence, Company Intelligence and QA Agent return data your code can use without parsing.

Define a field

Each key in struct is a field name. Its value is either a description or a typed object.

{
  "struct": {
    "email": "The individual's work email address",
    "years_of_experience": {
      "description": "Total years of professional experience",
      "type": "int"
    }
  }
}

Form	Example	Returned as
Description	`"The individual's work email address"`	String
Typed object	`{"description": "...", "type": "int"}`	The given type

Field names must be letters, numbers and underscores. The description is what the agent researches, so say exactly what you want: "Current job title at the company in the input" beats "title". Use a typed object for every field that isn’t a string. It’s the only form that behaves the same on every endpoint (details below).

Types

Type	Returned as	Example value
`"str"` or `"string"`	String	`"Pacific View Studios"`
`"int"` or `"integer"`	Integer	`50`
`"float"` or `"number"`	Number	`95.5`
`"bool"` or `"boolean"`	Boolean	`true`
`"list"`, `"list[str]"` or `"list[string]"`	Array of strings	`["SaaS", "Fintech"]`
`"list[int]"` or `"list[integer]"`	Array of integers	`[2019, 2023]`
`"list[float]"` or `"list[number]"`	Array of numbers	`[1.5, 2.7]`
`"list[bool]"` or `"list[boolean]"`	Array of booleans	`[true, false]`
`"dict"`	JSON object serialized as a string	`"{\"plan\": \"pro\"}"`
`"enum"`	One of `options`	`"Series A"`
`"list[enum]"`	Array of values from `options`	`["Sales", "Marketing"]`
`"object"`	Object with the given `subfields`	`{"city": "Austin", "country": "US"}`
`"object_list"`	Array of objects with the given `subfields`	`[{"name": "Acme", "role": "CTO"}]`

Parse dict values client-side with json.loads() (Python) or JSON.parse() (JavaScript).

Enums

enum and list[enum] require a non-empty options list. The value is constrained to those options.

{
  "seniority": {
    "description": "Seniority of the current role",
    "type": "enum",
    "options": ["IC", "Manager", "Director", "VP", "C-level"]
  }
}

Nested objects

object and object_list take subfields: a list of {"key", "type", "description"} objects. A subfield can be any type except object and object_list; enum and list[enum] subfields take options. type defaults to string and description is optional.

{
  "work_history": {
    "description": "Previous roles, most recent first",
    "type": "object_list",
    "subfields": [
      {"key": "company", "type": "string", "description": "Company name"},
      {"key": "title", "type": "string", "description": "Job title"},
      {"key": "start_year", "type": "int", "description": "Year the role started"}
    ]
  }
}

Missing values

Every field you request is present in the response. A field the agent couldn’t establish is null (Company Intelligence returns [] for list fields). Values are never invented to satisfy a type.

How each endpoint resolves types

Field definition	People Intelligence	Company Intelligence
`{"type": "int", "description": "..."}`	`int`	`int`
`"int"` (a type name on its own)	`int`, with no description	String
`"Total headcount"` (a description)	String	String
`50` (an example value)	String	`int`

In Company Intelligence, a field without an explicit type whose key also appears in target_company with a value takes that value’s type. Company Intelligence follows the People Intelligence column when any field in the struct is an object or object_list. QA Agent accepts descriptions only and returns every field as a string. See QA Agent.

Type casting

When a researched value doesn’t match its field’s type, Sixtyfour converts it only if the conversion loses nothing:

Target type	Accepted
String	Strings, plus numbers and booleans, converted to text
Integer	Integers, whole-number floats (`5.0`), integer strings (`"5"`)
Number	Numbers and numeric strings (`"95.5"`)
Boolean	Booleans and the strings `"true"` and `"false"`
Array	Arrays; each item is converted to the item type

Nothing else is cast: "about 50" never becomes 50.

Input fields in the output

In People Intelligence, fields you send in lead_info but don’t list in struct come back in structured_data unchanged, with their original values and types. Fields you do list in struct are researched and returned with the struct’s type, even if you sent a value for them.

{
  "lead_info": {"name": "Jane Doe", "company": "Acme Corp", "crm_id": 48213},
  "struct": {
    "title": "Current job title",
    "years_of_experience": {"type": "int", "description": "Total years of professional experience"}
  }
}

{
  "structured_data": {
    "title": "VP of Sales",
    "years_of_experience": 12,
    "name": "Jane Doe",
    "company": "Acme Corp",
    "crm_id": 48213
  }
}

Generate a struct from a description

POST /struct-builder/generate turns a plain-English description of the data you want into a list of typed fields.

Parameter	Type	Required	Description
`description`	string	Yes	The fields you want, in plain English. 1–4,000 characters.
`existing_struct`	array	No	Fields you already have, in the response format. The generated struct extends or revises them.
`block_type`	string	No	Where the struct will be used: `lead_enrichment` (People Intelligence), `company_enrichment` (Company Intelligence), `linkedin_enrichment`, `research_agent`, `webhook` or `transform_data`.
`allow_nested`	boolean	No	Allow `object` and `object_list` fields. Default `false`.

curl -X POST "https://api.sixtyfour.ai/struct-builder/generate" \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "description": "Work email, seniority (IC, manager, director or executive) and years of experience",
    "block_type": "lead_enrichment"
  }'

import requests

response = requests.post(
    "https://api.sixtyfour.ai/struct-builder/generate",
    headers={"x-api-key": "YOUR_API_KEY", "Content-Type": "application/json"},
    json={
        "description": "Work email, seniority (IC, manager, director or executive) and years of experience",
        "block_type": "lead_enrichment",
    },
)
response.raise_for_status()
fields = response.json()["struct"]

const response = await fetch("https://api.sixtyfour.ai/struct-builder/generate", {
  method: "POST",
  headers: { "x-api-key": "YOUR_API_KEY", "Content-Type": "application/json" },
  body: JSON.stringify({
    description: "Work email, seniority (IC, manager, director or executive) and years of experience",
    block_type: "lead_enrichment"
  })
});
const { struct: fields } = await response.json();

The response is a list of fields:

{
  "struct": [
    {"name": "work_email", "description": "The individual's work email address", "type": "string"},
    {"name": "seniority", "description": "Seniority of the current role", "type": "enum", "options": ["IC", "Manager", "Director", "Executive"]},
    {"name": "years_of_experience", "description": "Total years of professional experience", "type": "int"}
  ]
}

Enrichment endpoints take struct as a map, so key each field by its name:

struct = {f.pop("name"): f for f in fields}

There is no image type. A field whose description asks for an image (a headshot, a logo, a screenshot) is an ordinary string field whose value is a permanent image URL hosted by Sixtyfour. See Image fields.

​Define a field

​Types

​Enums

​Nested objects

​Missing values

​How each endpoint resolves types

​Type casting

​Input fields in the output

​Generate a struct from a description

Define a field

Types

Enums

Nested objects

Missing values

How each endpoint resolves types

Type casting

Input fields in the output

Generate a struct from a description