Repository navigation
Conversation
0.5.148 was pinned because 0.5.149 depended on an unpublished @tensorlakeai/tensorlake build. Every release from 0.5.150 on is published with its native packages, and 0.5.158 makes the default Sandbox.create() wait on the server instead of polling every 2 s, which was adding ~2.2 s of fixed client-side sleep to burst TTI. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Contributor License AgreementAll contributors are covered by a CLA. |
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Unpins
tensorlakefrom the exact0.5.148back to a caret range, now^0.5.158, and refreshespnpm-lock.yaml.The exact pin was added in #506 because
tensorlake@0.5.149depended on an unpublished@tensorlakeai/tensorlake@0.5.149. That was a broken publish on our side (sorry). Every release from0.5.150on ships with its@tensorlakeai/tensorlakeand@tensorlakeai/native-*packages, and the lockfile here resolves0.5.158with all six native packages.@computesdk/tensorlakeis unchanged: its^0.5.135range already admits0.5.158. No other lockfile entries move.Why this matters for the burst TTI numbers
From
0.5.136through0.5.157, the defaultSandbox.create()in the TypeScript SDK sent the create withwait: false, got a 202, polledGET /sandboxes/{id}once while the sandbox was still starting, then slept a fixed 2 s before polling again. On the 2026-10-20 run that put ~2.2 s of fixed client-side sleep into every Tensorlake TTI sample: our server promoted the 100 sandboxes to Running at p50 219 ms / p95 285 ms after the POST, while the harness measured p50 2765 ms / p95 2858 ms, with a 2.35 to 2.58 s spread between the create POST and the destroy DELETE on every sample.tensorlake@0.5.158(tensorlakeai/tensorlake#1040) makesSandbox.create()wait on the server's own state change instead, so the create returns as soon as the sandbox is Running. Nothing about what the benchmark measures changes; the adapter still callsSandbox.create()and thenrun("sh", ["-c", "node -v"]).Verification
pnpm install --frozen-lockfile --ignore-scripts,pnpm -r --filter './packages/**' run build, andpnpm typecheckpass with the new lockfile (pnpm 10.33.0, Node 24).tti.bench.ts --iterations 10 --concurrency 10 --provider tensorlake --no-ingestfrom a laptop (so each request carries ~300 ms more than the Namespace runner would see), same shape both runs:0.5.148(current pin)0.5.158(this PR)The result files those local runs wrote were discarded; this PR only touches
package.jsonandpnpm-lock.yaml.🤖 Generated with Claude Code