Submit tasks to the Databoard service

Submits one Databoard task per row of data, using the values of col as input. Task metadata (id, state) and results are stored back into data in the special columns .task_id, .task_state, and .task_result. After submission, results are unnested into regular columns and a progress summary is printed.

Usage

da_submit(data, col, task, options, wait = 0)

Arguments

data

A data frame containing the input data.

col

A column in data holding the text input. Tidy-evaluation is supported (i.e. pass the bare column name).

task

Character. The task the LLM is supposed to perform (e.g. "summarize", "coding", "annotate", "triples").

options

A named list of task-specific options passed to the API (e.g. coding rules, model parameters).

wait

Integer. Seconds to wait server-side for immediate completion of each task before returning. 0 (default) returns immediately with a PENDING state; larger values reduce the need for later da_fetch() calls.

Value

The input data frame with additional columns: .task_id, .task_state, and one column per field returned by the task (unnested from .task_result).

Details

Tasks are submitted sequentially, one per row. The function stops immediately if a submission fails.