feature/Public-execute-to-runresults-lift · L-260922-b7268d

[lchoquel]

The mapping from a blocking execute() result onto RunResults was a private function in client.py with a single caller, the bare-runner fallback inside start_and_wait. A consumer driving execute() itself had to re-read pipe_output.model_extra and re-validate the usage records by hand before it could reach summarize_usage, collect_artifacts or any of the parity fields — which is what the Python starter does today.

It moves to pipelex_sdk/execute_result.py as results_from_execute(result), beside the PipelexExecuteResult it takes and on the side of the import edge that already depends on runs. The mapping itself is unchanged, execute() still returns PipelexExecuteResult because that model carries the runner's whole typed envelope, and the docs now say how a direct caller gets from one to the other.

Closes L-260922-b7268d

🤖 Generated with Claude Code

---

Summary by cubic

Makes the mapping from a blocking execute() result onto RunResults public as results_from_execute(result) in pipelex_sdk/execute_result.py.

Previously only start_and_wait used this mapping; a consumer driving execute() directly had to re-read pipe_output.model_extra and re-validate usage records by hand to reach summarize_usage, collect_artifacts or the parity fields. The new function performs the same lift, so direct callers get the same RunResults shape the durable path hands back. execute() still returns PipelexExecuteResult because that model carries the runner's whole typed envelope. Adds unit tests and doc updates. Closes L-260922-b7268d.

^{Written for commit 4b5fb1a. Summary will update on new commits.}