[kartojal]

Problem

HTTP reads retried by default, which conflicts with callers and query libraries that own retry policy.

Fix

• Make HTTP read retries opt-in: retries defaults to 0. Passing 1 to 5 enables retries for transport failures and short rate-limit waits. Mutations are never retried.
• Remove the unused mutation retry: false option and retryMutations configuration; update public TSDoc, README, AGENTS.md and focused retry tests to match.
• Preserve the existing per-attempt timeout, cancellation and shared read retry budget behavior.

Stack scope

The callback-owned AuthenticatedServiceClient transport is isolated in #481. Bearer/L2 credential types, strict input schemas and the concrete session lifecycle are introduced in #437.

Stack: #442 → #481 → #437 → #463 → #438 → #441.

🤖 Generated with Claude Code

---

Note

Medium Risk
This is a breaking default change for read retries that can surface more failures unless callers opt back in; transport code edits need careful validation even though mutations stay single-shot.

Overview
Summary
HTTP reads now start with zero transport retries unless callers pass retries > 0; parseRequestOptions and Ky initialization align with that default so mutations remain single-shot while opt-in reads share the existing retry budget.

Documentation & examples
README, AGENTS.md, decorator/action TSDoc, and the changeset call out the new default and show how to re-enable retries; RFQ tests expect the lighter default.

Tests
ServiceClient suites were reworked to assert single-attempt behavior for mutations and the opt-in retry path, dropping obsolete rate-limit/backoff scenarios.

^{Reviewed by Cursor Bugbot for commit c3f879b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit a92e6e8. Configure here.}

[cursor]

Workflow GETs lose transport retries

Medium Severity

Setting the shared ky instance retry to 0 also disables the previous GET transport recovery used by getWorkflow. Those command paths (API-key derivation, perps credential checks, and balance-allowance updates) no longer retry network failures and cannot opt in with retries. The new docs still say workflow retry policy is unchanged.

Additional Locations (2)

• packages/client/src/ServiceClient.ts#L323-L340
• packages/client/src/request-options.ts#L44-L49

 

^{Reviewed by Cursor Bugbot for commit a92e6e8. Configure here.}