Skip to contents

Create a run project directory

Source: R/project.R
np_project_init.Rd

Scaffolds one directory per matching run: numbered stage folders, each with batches/, interim/ and logs/ subdirectories, a README.md explaining what the stage does, and (for the LLM stages) a PROMPT.md holding the agent instructions. Writes a config.yml, an empty manifest.csv, an AGENTS.md describing how to orchestrate the whole run, and a first RUN-STATUS.md.

Usage

np_project_init(
  path,
  run_id = NULL,
  bmf = NULL,
  sam = NULL,
  config = np_config(),
  templates = TRUE,
  use = TRUE,
  overwrite = FALSE,
  quiet = FALSE
)

Arguments

path

Directory to create for this run.

run_id

Short run label recorded in reports and in the run_id column of every stage output. Defaults to basename(path).

bmf

Optional path to (or vintage tag of) the reference used by this run, recorded in 00_bmf/SOURCE.md. NULL leaves a placeholder.

sam

Optional path to a SAM extract to record in 00_sams/README.md. The file is not copied.

config

An np_config() whose thresholds and weights are serialized to config.yml.

templates

Write the README / PROMPT / AGENTS documents. Default TRUE.

use

Set this project as the session default. Default TRUE.

overwrite

Overwrite existing template documents. Directories and data files are never removed. Default FALSE.

quiet

Suppress the summary message. Default FALSE.

Value

The project root, invisibly.

Details

All six stage folders are created unconditionally. Deploying a stage is optional — one that never runs simply leaves its outputs absent, which np_project_status() reports as not run rather than as an error.

Reference data is pointed at, not copied

The BMF is ~3.5 GB raw plus a ~470 MB normalized bundle, and is reused across runs. 00_bmf/ therefore holds only SOURCE.md (vintage, resolved path in the shared store, checksum, row count) and SIGNATURE.txt (the np_normalize_signature() hash of the cached artifact, so drift is detectable). The bytes stay in the np_data_path() store. SAM is genuinely per-run, so 00_sams/raw/ holds the real extract — paste one in, or let np_fetch_sam() download into it.

See also

np_project_status() for the run status report, np_project_path() for addressing files inside it, np_data_init() for the shared asset store.

Examples

p <- file.path(tempdir(), "run-demo")
np_project_init(p, run_id = "demo", quiet = TRUE, use = FALSE)
list.files(p)
#>  [1] "00_bmf"        "00_sams"       "01_stage1"     "02_stage2"    
#>  [5] "03_stage3"     "04_final"      "AGENTS.md"     "README.md"    
#>  [9] "RUN-STATUS.md" "config.yml"    "manifest.csv" 
unlink(p, recursive = TRUE)