Run ID and Provenance
A run ID ties together snapshot, models, library version, and configuration so every number has a recipe.
Explanation
Each pricing or risk run is identified by a run ID that records the portfolio, snapshot ID, model set, and library version.
Run metadata also includes pricer settings such as numerical tolerances, seeds, and approximation flags.
Trade-level outputs and logs carry the run ID, making it possible to trace any PV or Greek back to its full context.
Run provenance turns isolated numbers into reproducible experiments rather than one-off calculations.
run-idprovenancemetadatareproducibility
Interactive visualisation
Run ID and provenance
A run ID is the handle. Provenance is the recipe: snapshot, library, models, and configuration.
Run inputs
Run ID
run_99d7faf4
Seed
42
abs tol
1e-1
rel tol
1e-5
What is captured?
Reproducibility
REPLAYABLE
Interpretation
- If any ingredient is missing, a run ID becomes a label, not a recipe.
- Seeds and tolerances are configuration. They belong in provenance.
- Trade-level logs carrying run_id let you trace “one wrong PV” to its exact context.
Run recipe
snapshot: snap_cd35e0
16:00
INFO
start run_id=run_99d7faf4
16:00
INFO
portfolio=PORTF_A
16:00
INFO
snapshot_id=snap_cd35e0
16:00
INFO
library=qwv-pricers@1.7.2
16:00
INFO
model_set=IR+FX v3
16:00
INFO
abs_tol=1e-1
16:00
INFO
rel_tol=1e-5
16:00
INFO
seed=42
16:00
INFO
timestamp=2026-02-01T16:00:00Z
Toggle off snapshot/library/config to see how provenance becomes incomplete.
Trace map
One run ID should connect snapshot → models → outputs → logs.
Snapshot
snap_cd35e0
Library
qwv-pricers@1.7.2
Configuration
seed=42
Outputs
replayable
Labels (toy)