Changelog¶

v0.6 Adapter Workflow Split¶

v0.6 makes the adapter API more explicit. Earlier versions centered the public workflow on ingest, which did two jobs at once:

standardize raw planner output into canonical Route objects
collect those routes onto benchmark targets and write routes.json.gz

That shape worked for benchmark runs, but it made RetroCast feel less like a general route-standardization library. In v0.6, standardization and benchmark collection are separate library workflows. ingest still exists as the project-mode convenience command that runs both.

What Changed¶

Before v0.6	v0.6 replacement	Why
`adapt_single_route(raw, target, adapter_name)`	`adapt_route(raw_route, adapter)`	The one-route API no longer requires target context when the raw route carries its own target.
`adapt_routes(raw, target, adapter_name)`	`adapt_provider_output(raw_provider_output, adapter)`	Standardization can now handle one provider output without benchmark target context.
`retrocast ingest` as the main way to adapt benchmark predictions	`retrocast adapt` then `retrocast collect` for ad-hoc use	Exposes standardization and benchmark collection as separate steps.
`retrocast ingest` for project-mode benchmark runs	still `retrocast ingest`	Project mode keeps the one-command convenience wrapper.
`Route.rank`	list order	Keeps canonical `Route` free of benchmark/list-position metadata.

Prediction manifest content_hash values are now order-sensitive within each target because route list order is the ranking signal. Re-exported manifests can therefore differ from older manifests even when the route structures are otherwise identical.

adapt_single_route(...) and adapt_routes(...) emit RetroCastFutureWarning in v0.6 and are scheduled for removal in v0.7.

1-5 Minute Migration¶

For one raw route-like payload:

from retrocast import adapt_route, get_adapter

route = adapt_route(raw_route, get_adapter("dms"))

For one raw provider output containing one or many routes:

from retrocast import adapt_provider_output, get_adapter

routes = adapt_provider_output(raw_provider_output, get_adapter("aizynth"))

For raw output already keyed by target ID or target SMILES:

from retrocast import adapt_target_keyed_provider_output, get_adapter, load_benchmark

benchmark = load_benchmark("benchmark.json.gz")
routes = adapt_target_keyed_provider_output(raw_mapping, benchmark, get_adapter("aizynth"))

To produce benchmark-keyed routes for scoring:

from retrocast import collect_benchmark_predictions

collected = collect_benchmark_predictions(routes, benchmark)
routes_by_target = collected.routes_by_target

CLI Migration¶

Ad-hoc CLI users can now run the two steps explicitly:

retrocast adapt \
  --input raw_predictions.json.gz \
  --adapter aizynth \
  --input-kind provider-output \
  --output route-corpus.jsonl.gz

retrocast collect \
  --input route-corpus.jsonl.gz \
  --benchmark benchmark.json.gz \
  --output routes.json.gz

Project-mode users can keep using retrocast ingest. In v0.6 it remains the one-command wrapper that adapts raw output and then collects routes onto the selected benchmark.