v0.56.0

@pipelex/sdk on npm

Changed

• listRuns and iterateRuns take the method as an optional filter (Breaking): listRuns(query?) and iterateRuns(query?) take the method as methodId in their query, sent as method_id, where it was a required first argument, and without it they list the organization's runs of every method, newest first, with createdBy, status, createdFrom and createdTo applying as they do with a method. Move the id into the query, listRuns(methodId, query) becoming listRuns({ ...query, methodId }), and pass a cursor back with the same methodId, since the cursor of one list is refused on another. A call still passing the id first is rejected with a RequestArgumentError before any request, for iterateRuns at the call rather than when iteration begins, instead of listing every method's runs.
• A run-history row names its method and its author (Breaking): RunHistoryItem gains method_id, the bare catalog id of the method the run ran, and created_by_user_id, the canonical user id of the member who started it, both required, so a list across methods can name each row's method from the catalog and its author from the organization's members; a row built by hand, as in a test, now carries both.

pipelex-sdk on PyPI

Changed

• list_runs and iterate_runs take the method as an optional filter (Breaking): method_id is a keyword-only argument among the filters, None by default, where it was a required first positional argument, and without it they list the organization's runs of every method, newest first, with created_by, status, created_from and created_to applying as they do with a method. Call list_runs(method_id=method_id, …) where you called list_runs(method_id, …), and pass a cursor back with the same method_id, since the cursor of one list is refused on another.
• A run-history row names its method and its author (Breaking): RunHistoryItem gains method_id, the bare catalog id of the method the run ran, and created_by_user_id, the canonical user id of the member who started it, both required, so a list across methods can name each row's method from the catalog and its author from the organization's members; a row served without either is an answer the SDK cannot read, raised as ApiResponseError.

Fixed

• iterate_runs checks its status at the call: a status given as a one-shot iterable, such as a generator, is checked and copied once and filters every page, where the second page found it exhausted and raised RequestArgumentError; an empty or unknown status raises RequestArgumentError, and a bare str TypeError, at the call rather than once iterated, as iterate_files checks its filters. list_runs and iterate_runs type status as Iterable[RunStatus], which they read once, where it was typed Sequence[RunStatus].