Commit graph

2 commits

Author SHA1 Message Date
mateo-berri
04a318726f fix(interactions): forward continuation_token and handle requires_continuation for long decodes
Google's Interactions API cuts a long unary decode short with a new
requires_continuation status and a continuation_token that the client sends
back on a new create to resume. The proxy refused a stateful resume for a
missing input and dropped the token on a store=false resume, so the answer
restarted.

Add the status and token to the generated types, forward the token to Google
AI Studio and Vertex, skip the input check when a create carries a token,
reject the token for bridged providers unless drop_params is set, treat
requires_continuation as a finished interaction with output in cost polling,
and refresh the vendored spec snapshot.
2026-10-30 06:30:00 -07:00
Mateo Wang
08f04fbc97
test(interactions): read the spec from a vendored snapshot and canary the live one daily (#48002)
* test(interactions): assert the live status enum is parsable instead of pinning it

* test(interactions): read the spec from a vendored snapshot and canary the live one daily

* test(interactions): run the canary on PRs touching its inputs, file an issue on a red run, pin the regenerate command

* test(interactions): take the live spec URL as a pytest option, assert its source, split the canary's tracking-issue job

* test(interactions): file the canary issue whenever its compliance step did not pass, match the issue title exactly, cover UrlContextResult
2026-10-23 17:20:40 +00:00