Run record¶
Every call writes runs/<run_id>/run.json under its project folder, from the CLI or Python and also when it fails. The run ID is a UTC timestamp plus six random hex characters, for example 20261017T195308Z-510ada.
| Field | What |
|---|---|
run_id |
The run ID |
command |
The command name, for example data health |
params |
The validated params as JSON |
argv |
The equivalent command line, from to_argv |
penguineer_version, plantgraph_version, python_version |
Versions |
seed |
The random seed |
llm |
The LLM mode: none, local or api |
llm_provider, llm_model, llm_prompts |
The provider (local, anthropic or openai), the model and the prompt versions a run used, when it used an LLM |
inputs |
Every input file: path, sha256 and bytes |
outputs |
Paths of the files the command wrote |
started_at, duration_s |
When it started and how long it took |
status, exit_code, error |
ok or error, the exit code and the error message |
Alchemist seals the run record into its own run manifest.