fix(cli): exclude gpt-5.6-terra models from explicit prompt cache breakpoints - #13052
fix(cli): exclude gpt-5.6-terra models from explicit prompt cache breakpoints#13052chrarnoldus wants to merge 1 commit into
Conversation
|
|
||
| function supportsPromptCacheBreakpoint(model: Provider.Model): boolean { | ||
| if (isLikelyChatGPTSubscription(model)) return false | ||
| if (model.api.id.includes("terra")) return false |
There was a problem hiding this comment.
SUGGESTION: Consider matching -terra instead of a bare terra substring
includes("terra") matches anywhere in the model id, so a hypothetical future non-Terra model whose name merely contains the substring (e.g. a terraform-themed variant) would silently lose explicit breakpoints. includes("-terra") would still catch gpt-5.6-terra and prefixed gateway ids like openai/gpt-5.6-terra while avoiding that false positive. This is consistent with the broad includes checks already used nearby (claude, anthropic), and the failure mode is a graceful fallback to automatic caching, so feel free to reject.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
Code Review SummaryStatus: 1 Issue Found | Recommendation: Address before merge Overview
Issue Details (click to expand)SUGGESTION
The fix is correct and well-scoped: the Terra exclusion sits inside the existing Fix these issues in Kilo Cloud Files Reviewed (3 files)
Reviewed by kimi-k3 · Input: 58.4K · Output: 6.6K · Cached: 237.5K Review guidance: REVIEW.md from base branch |
Summary
prompt_cache_breakpoint is not supported on this modelonGPT-5.6 Terra).-terramodel variants from manualpromptCacheBreakpoint: { mode: 'explicit' }treatment on the OpenAI provider, falling back to standard automatic caching.