`dodil data actions`

Curated Scriptum actions over a bucket’s data — list + run. Two subcommands ship: list and run (action is an accepted alias for the group).


dodil data actions list    # what actions fit a bucket/selection
dodil data actions run     # dispatch one and stream its progress

The queued and history calls from the API Reference (enqueue, runs/*) have no CLI verb today — reach them over HTTP or gRPC. The CLI covers the two interactive verbs.

Subcommand	RPC
`dodil data actions list`	`ListAvailableActions` — `POST /actions/list`
`dodil data actions run`	`RunAction` — `POST /actions/run` (NDJSON stream)

Both take --bucket / -b (required). The selection flags — --key, --table — decide which actions are offered and what the action runs against.

`dodil data actions list`


dodil data actions list -b BUCKET [--table TABLE] [--key KEY]... [--search TEXT]

Lists the actions that fit the selection. With no selection flags it returns every action available on the bucket; add --table or one-or-more --key to scope to a table or object selection and the list narrows to the actions that accept it.

Flag	Short	Type	Description
`--bucket`	`-b`	string	Required. Bucket to scope to
`--table`	—	string	Scope to a table (table selection)
`--key`	—	string (repeatable)	Scope to object key(s) (objects selection)
`--search`	—	string	Filter actions by free text (name + description)


# Everything available on the bucket
dodil data actions list -b kb-prod
 
# Actions that fit a single object
dodil data actions list -b kb-prod --key contracts/acme.pdf
 
# Actions that fit a table
dodil data actions list -b kb-prod --table invoices
 
# JSON — the full params_form per action, for scripting
dodil data actions list -b kb-prod -o json \
  | jq -r '.actions[] | "\(.actionId)\t\(.invoke)\t\(.description)"'

The list is server-driven (a filtered Scriptum catalog), so treat it as live rather than a fixed set — enumerate it instead of hard-coding ids. See Catalog for the shipping set.

`dodil data actions run`


dodil data actions run <action_id> -b BUCKET \
  [--table TABLE] [--key KEY]... [--param k=v]...

Dispatches one action over the selection and streams its ActionEvents to the terminal — started → progress… → completed | failed. Params are key=value pairs, one --param each.

Flag	Short	Type	Description
`--bucket`	`-b`	string	Required. Bucket to scope to
`--table`	—	string	Scope to a table (table selection)
`--key`	—	string (repeatable)	Scope to object key(s) (objects selection)
`--param`	—	`key=value` (repeatable)	Action form param


# Q&A over a single document (sync — completes inline)
dodil data actions run object_qa -b kb-prod \
  --key contracts/acme.pdf \
  --param question="What is the termination notice period?"
 
# Summarize a document, bullet style
dodil data actions run summarization -b kb-prod \
  --key reports/q3.pdf --param style=bullets
 
# Translate a document (the target language is a form param)
dodil data actions run translation -b kb-prod \
  --key docs/readme.md --param target_language=fr
 
# Natural-language question against a table (async — streams progress)
dodil data actions run table_qa -b kb-prod \
  --table invoices --param question="total revenue in 2025?"

Selection maps from the flags: --key (once) → object selection, --key repeated → objects selection, --table → table selection, none → agentic (plan_and_run). The action’s other inputs (bucket, key, table) are filled from the selection — you only pass the form params the action declares (browse them with actions list -o json).

Chunk and collection selections aren’t reachable from the CLI. list / run expose only --key and --table, so collection_topic_map (collection selection) and chunks_synthesize_answer (chunks selection) need the API with a CollectionSelection / ChunksSelection. plan_and_run runs from the CLI with no selection flags.

dodil data actions

dodil data actions list

dodil data actions run

See also

`dodil data actions`

`dodil data actions list`

`dodil data actions run`