These functions help construct and explain get_diver_data() calls (from
the certedb package) using the active LLM. Each call uses an isolated
chat context so the main conversation history is not affected.
llm_diver_query() generates a get_diver_data() call from a natural
language description.
llm_diver_explain() explains an existing get_diver_data() call in
plain Dutch.
The LLM is primed with a static schema hint describing known Diver column names and conventions. When the exact column name is uncertain, the generated code includes a comment noting the assumption.
llm_diver_query(
description,
table = NULL,
columns_hint = NULL,
new = FALSE,
...
)
llm_diver_explain(query, new = FALSE, ...)Natural language description of the data to retrieve,
e.g. "All E. coli urine cultures from 2024 in Groningen".
Optional hint for the Diver table to query.
Optional character vector of known column names to assist the LLM in forming the query.
Logical. If TRUE, re-initiates the LLM before the task call.
Default is FALSE.
Additional arguments passed to initiate_llm().
An existing get_diver_data() call as a character string,
to be explained.
A character string: the generated R code (llm_diver_query()) or
the Dutch explanation (llm_diver_explain()).