csv/source

Built-in source adapter for person records: its manifest, needs, provides, config keys, cost, and an example step

What you'll learn

You're about to put csv/source in a step and want its keys, fields, credentials, and cost before plan tells you.

what the adapter needs and provides
every with: key it accepts
what it costs and which credential it demands

A built-in source adapter for person records. It publishes no per-record estimate, so plan prints ?, and it demands no credential.

Manifest

Key	Value
id	`csv/source`
version	1
role	source
entity type	person
credentials	none
cost estimate per record	not published

Needs

Nothing. A source starts a run, so no record precedes it.

Provides

None declared. What it writes depends on its with: config, and plan reports the fields as the step resolves them.

Config keys

The keys with: accepts. Plan rejects any other key.

Key	Type	Required	What it sets
`columns`	object		Ingress mapping (ADR-018): canonical field name → CSV header as written. Headers already matching canonical names auto-map; unmapped non-canonical headers become csv.<normalized_header>.
`entity_type`	one of person, company
`limit`	integer
`path`	string	yes	CSV file to read; header row names the fields

Example

From Your CSV:

source:
  use: csv/source
  with:
    path: contacts.csv
    columns:
      full_name: Full Name
      email: Email
      title: Title
      company_domain: Company Website

Used in

The minimal-offline example in gtme help --agent
The human-review-then-cron example in gtme help --agent
The csv-sourced-compose-with-uses example in gtme help --agent
The persona-file-and-text-compose example in gtme help --agent