Notebooks¶
Three reference notebooks live under notebooks/ in the
repository. They walk through the same NC1 / SC1 / partition
checks as the CLI but render them inline with plots so you can
read each step.
| Notebook | Topic | What it shows |
|---|---|---|
01_verify_nc1.ipynb |
Baseline NC1 verification | Calls run_baseline equivalents in cells, plots M (dB) per window with Mmin rule, prints the signed indicator payload. |
02_sc1_omega.ipynb |
SC1 via Ω power sag |
Reproduces an Ω window, overlays 𝓛_loop recovery, computes δ and τ_rec, calls sc1_evaluate. |
03_partition_sanity.ipynb |
Partition sanity checks | Exercises PartitionManager and greedy_suggest_C; plots before / after (C, Ex) membership. |
Running locally¶
Install the package in editable mode (notebooks import from
ldtc.*) and start Jupyter:
Then open the notebooks in the order above. Each notebook starts
with a small "set up the harness" cell that mirrors what
ldtc run does internally; if you have already done a CLI run,
the notebook will happily read from your existing
artifacts/audits/audit.jsonl.
Tips¶
- All three notebooks honor
seed,seed_py, andseed_npfrom the bundled R0 profile, so re-runs reproduce the same plots. - The plotting cells reuse the same theme as the CLI bundle; see
Reporting for the palette and
matplotlib
rcParams. - If a cell raises a smell-test exception, that is not a bug; it is the harness telling you the run was invalidated. Read the audit log alongside the notebook output.
See also¶
- Minimal run: a non-interactive equivalent.
- Runs: the corresponding CLI commands.
- Definitions: every symbol the notebooks use.