Repository navigation
1.43.2: Sonnet 5.5 pricing, and relay fresh fixes - #2
Merged
Merged
Conversation
…older project from its trusted folder, and relay cloud finds the shim behind where Three things a review of 1.43.0 found. The inline hand-off on Linux was checked against 120,000 characters, but the kernel's limit on one argument is 128 KiB of bytes. 70,000 two-byte characters (140,000 bytes) passed the check and the exec failed E2BIG here, so a hand-off heavy in accented text or dashes would have died in the window the same way on every retry. The POSIX cost is now UTF-8 bytes, and past it the first prompt points at the hand-off file, as it already did on Windows. The POSIX launcher cd'd into record.cwd even when the relay had chosen a trusted launch folder for a project in the home folder. The preflight answers the trust question for the launch folder only, and the new session's opening says it was started from one, so the window would have sat at the trust question. It now starts where the Windows launcher does, launchCwd, with --add-dir back to the project; the resume launcher shares the line. relay cloud took the first line of `where claude` as the claude a Windows shell runs, and on an npm install that line is the extensionless sh script beside claude.cmd, so the batch-shim warning never fired for the install it was written for. It now takes the first match with an extension. UNVERIFIED on a real Windows machine: the where output order is the documented one, fed to the function in the test. defer.md and relay.md also say the file is used when the hand-off is too long for one argument. The three new tests fail without the script changes. Suite 989/990, the skip being Task Scheduler, which is Windows only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
package.json, both plugin manifests and the VS Code package say 1.43.1, and vscode/lib is rebuilt from the scripts at that version - 32 scripts. 989/990, the one skip being Task Scheduler, which is Windows only; no usage-limits temp folder left behind. Patch, because these are fixes to 1.43.0. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
…th rule claude-sonnet-5-5 was missing from RATES, so its calls fell to the sonnet family average, $2.50/$12.50 from Sonnet 5 and 4.6 together - 25% over on every Sonnet 5.5 turn, and the model listed as assumed rather than known. It keeps Sonnet 5's prices (claude-api model table, cached 2026-09-25), and its $0.20 reads are a tenth of input, so no cacheRead override. Because nothing is cheaper, Sonnet 5 is not pointed at 5.5 as a newer sibling. One new test: rate, [1m] suffix, cost of a million in, out and read, and the parsed version. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
package.json, both plugin manifests and the VS Code package say 1.43.2. Patch: a price row for a model the table did not know. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
An unknown Sonnet id falls back to the average of the Sonnet rows, and that is three rows now: $7/3 in and $35/3 out, not $2.50/$12.50. The expected value was written for two. 990/991, the one skip being Task Scheduler. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Five commits on top of main (1.43.0).
Sonnet 5.5 (1.43.2)
claude-sonnet-5-5was missing fromRATES, so its calls fell back to the Sonnet family average ($2.50/$12.50). That overstated every Sonnet 5.5 turn by 25% and listed the model as "assumed".cacheReadoverride.newerSibling('claude-sonnet-5')stays null.[1m]suffix, the cost of 1M tokens in, out and read, and the parsed version;Relay fresh fixes (1.43.1)
relay freshcounts the POSIX argument limit in bytes.relay cloudfinds the shim behindwhere.Checks:
node --testgives 990/991 on Linux. The one skip is the Task Scheduler test, which only runs on Windows and has not been run here.🤖 Generated with Claude Code
https://claude.ai/code/session_01CPrhYxAHgTrw4QZVtWaHV8
Generated by Claude Code