Reconstruct LLM analyses against their targets

Joins llm_results rows back to the row they analysed. The target table is named per row by target_type (asset, item or collection; unknown and missing targets have no resolvable row), so this cannot be one SQL join: each row's target is looked up in the table its type names and returned as a target list-column (a one-row tibble, or NULL when unresolvable). The result JSON is parsed into a list-column and timestamps become POSIXct.

Usage

entropia_reconstruct_analysis(con, target = NULL, job_type = NULL)

Arguments

con

A connection returned by entropia_connect().

target

Optional single target id (an asset/item/collection id) to keep. NULL (the default) keeps all targets.

job_type

Optional character vector of job types to keep. NULL (the default) keeps all.

Value

A tibble of class entropia_reconstruction.

Details

Materialised (one row per llm_results row) and ordered deterministically by id; the result carries the entropia_reconstruction class.

Examples

con <- entropia_connect(system.file("extdata", "entropia-example.sqlite",
  package = "entropiaR"
))
entropia_reconstruct_analysis(con)
#> Reconstructed LLM analyses: 1 row(s)
#>   item       1
#> 
#> # A tibble: 1 × 7
#>   id    target_id target_type job_type result       created_at          target  
#>   <chr> <chr>     <chr>       <chr>    <list>       <dttm>              <list>  
#> 1 llr-… 22222222… item        summary  <named list> 2026-01-15 12:11:40 <tibble>
entropia_disconnect(con)