Move CacheTtl to enums, export from package, raise maxTokens#207
Move CacheTtl to enums, export from package, raise maxTokens#207
Conversation
…factor The current-state section still described the architecture refactor as active work with a step-by-step table. That series finished at PR #199. Replace it with the three open PRs (#206, #207, #211) and a pointer to the backlog plans. Recent decisions gains entries for the PreviewEdit lineEdits/textEdits redesign and the CacheTtl enum move. Remove the stale cancelFn note — that work landed in #199.
bananabot9000
left a comment
bananabot9000
left a comment
There was a problem hiding this comment.
Holding for confirmation on two items 🍌
-
maxTokens8000 → 32000: This was explicitly lowered from 32768 to 8000 in PR #205 (merged today). Hellcar also mentioned Claude reverting his limit change tonight. Was this raise back to 32000 intentional, or did 006 sneak it in? -
SDK-level
cacheTtldefault changed from 5m to 1h inAgentRun.ts. This affects ALL consumers ofclaude-sdk, not just the CLI. The CLI was already overriding to'1h', so that side is unchanged. But any other consumer relying on the 5m default will silently get 1h caching. Intentional?
Everything else is clean -- CacheTtl enum move to enums.ts is good organisation, exports fixed from type-only to value, imports all updated consistently. No sensitive data.
Will approve once both items are confirmed intentional.
Reviewed by BananaBot9000 🍌
* Add sdk-tools and cli-features backlogs, link all GitHub issues * Update CLAUDE.md: current state reflects active PRs, not completed refactor The current-state section still described the architecture refactor as active work with a step-by-step table. That series finished at PR #199. Replace it with the three open PRs (#206, #207, #211) and a pointer to the backlog plans. Recent decisions gains entries for the PreviewEdit lineEdits/textEdits redesign and the CacheTtl enum move. Remove the stale cancelFn note — that work landed in #199.
bananabot9000
left a comment
bananabot9000
left a comment
There was a problem hiding this comment.
Confirmed with Hellcar -- both changes intentional 🍌
maxTokens 8k → 32k: The 8k limit was a cost band-aid during the broken-caching era. Claude was hitting the ceiling editing a 463-line (23kb) test file, failing 3-4 times and wasting more tokens on retries than the full output would have cost. With caching fixed (PR #205), the input cost savings dwarf any output increase.
cacheTtl 5m → 1h default: Aligns SDK default with what the CLI was already overriding to. 1h is the right default for most use cases.
CacheTtl enum move to enums.ts is clean. Exports fixed from type-only to value. All imports updated consistently.
No sensitive data.
Reviewed by BananaBot9000 🍌
# Conflicts: # .claude/plans/sdk-tools.md
2 participants
Summary
CacheTtlmoved fromtypes.tstoenums.tsalongsideAnthropicBetaAgentRunandrunAgentupdated to useCacheTtl.OneHourinstead of the'1h'string literalmaxTokensraised from 8000 to 32000 inrunAgent