ci: adapter sdk peer range - #150
Open
faustoq wants to merge 3 commits into
Open
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
Both framework adapters publish their
@ratel-ai/sdkpeer as^<in-repo SDK version>.At 0.x a caret pins the minor (
^0.6.0means>=0.6.0 <0.7.0), so every SDK minor orphans both adapters.npm install @ratel-ai/mastra @ratel-ai/sdk@latestfails withERESOLVE. Broken since SDK 0.7.0 (2026-08-07).The daily
verify-installcron caught it a few times (2026-08-08 -> 2026-08-13), then went green because #148 removed the SDK co-install from the adapter verify jobs. The check stopped exercising the failure mode but the breakage still exists.Solution
Publish a floor range instead of a caret:
SDK_ADAPTER_PEER_FLOORinscripts/release-units.mjsis the single source of truth.scripts/pin-adapter-sdk-peer.mjsapplies it in bothrelease.ymlpublish jobs and inpublish-rc.sh.Source manifests stay
workspace:^.The floor is
0.9.1: vercel's shipped code needsExperimentalPassthroughToolExposure, first exported in SDK 0.9.1 (verified 94/94 and 52/52 against 0.9.1).>=0.6.0would resolve and then fail on a missing export. ADR-0020 records the measurement.The SDK publish-order
npm viewguard is dropped: the peer no longer names the in-repo SDK version, so a floor-range check could never fail. Telemetry pins and their guards are unchanged.Notes
vercel-ai-sdk@0.4.0,mastra@0.3.0). This PR does not change any published package by itself.release.ymladded to thetspaths filter, so a release.yml-only PR still runs the contract tests that guard the pinpublish-rc.shrestores the manifests through a guardedgit checkoutEXIT trap. A core/telemetry-only run can't clobber an unrelated uncommitted edit^1.2.0already means>=1.2.0 <2.0.0, so we can delete it and letrelease.ymlgo back to a plain caret. Explained in ADR-0020.sdk@latestco-install toverify-install.ymlso the cron catches this class of drift again. It also needs theworkflow-contracts"does not interpolate the SDK version" assertion updated.