Conversation
noahlwest
left a comment
There was a problem hiding this comment.
Thank you very much! Overall looks good, just two minor comments.
…bectl-ai into session_persistance
…t failures (GoogleCloudPlatform#525) * Update statefulset-lifecycle prompt and verify to make eval less flaky * Update list-images-for-pods eval to work with kind * Remove unneeded verify.sh file
… to verify. Refactor verify.sh (GoogleCloudPlatform#524)
|
@noahlwest @droot |
Getting this in today. Thank you! |
presubmits are failing, you may want to run |
…Platform#521) MCP servers can return object schemas with additionalProperties but no properties field, which is valid JSON Schema. The current code panics when trying to convert such schemas. This fix adds proper nil/existence checks before accessing the properties field and gracefully handles object schemas that rely on additionalProperties without explicit properties. Fixes GoogleCloudPlatform#520
* install mockgen * wrong mockgen
* fix: update Kind cluster setup action and add error handling in periodic CI workflow * fix: reduce timeout values and improve retry logic in periodic eval workflow
* update k8s-bench results * remove gemma results
* add krew releaser step * add generated from * remove comment
…ogleCloudPlatform#542) * Add --show-failures flag to `analyze-evals`, update scripts * Add string formatter to AddFailure call * add --show-tool-output to eval runs
…eCloudPlatform#553) * Add --create-kind-cluster flag to k8s-bench * Make cluster name fixed, add logging
Conflicts: go.mod Changes to be committed: modified: go.mod new file: pkg/sessions/session_test.go
…latform#518 now looks closed, also closes GoogleCloudPlatform#497
…rmat.sh" This reverts commit 123954a.
…ormat.sh" This reverts commit 325734a.
…/tasks/format.sh"" This reverts commit 68b0066.
…bectl-ai into session_persistance
@noahlwest
Just implement Session Tests for #466
It covers the following:
A modification I made in session.go is to change the implementation of session file loading from entire JSON at one time to line by line so any corrupt JSON lines can be omitted
Another issue that I found out during testing is that GetLatestSession gets the first session, however it gets the latest session when a call to ListSession has already been made which sorts the slice of sessions, at the moment I am not sure if there is a use case of calling GetLatestSession without calling the ListSession(), because if there is a case for that then we should fix this.