Skip to content

perf(coding-agent): request short evolution prompt caching - #259

Open
Issacboy123 wants to merge 1 commit into
fix/retryable-evidence-processingfrom
fix/evolution-prompt-cache
Open

Issacboy123 wants to merge 1 commit into
fix/retryable-evidence-processingfrom
fix/evolution-prompt-cache

Conversation

@Issacboy123

Copy link
Copy Markdown
Contributor

Summary

Evolution model calls resend stable request content across tool rounds but previously opted out of provider prompt caching. This change requests short prompt-cache retention for every evolution model call while preserving the existing stage idempotency key as the provider session ID.

This PR is stacked on the retry-safe evidence-processing fix.

Reasoning

The centralized evolution model executor applies the retention request consistently to every stage without changing response handling or durable checkpoint state. Short retention lets providers that support it reuse stable prompt prefixes while retaining the existing session identity and avoiding new cache-usage persistence or provider-specific behavior.

Changes

  • Changed evolution model calls to request short provider prompt-cache retention.
  • Preserved the stage idempotency key as the provider session ID.
  • Added a faux-provider regression that verifies both retention and session-ID propagation.

Testing

  • node ../../node_modules/vitest/dist/cli.js --run test/evolution-runtime.test.ts
  • npm run check

@Issacboy123 Issacboy123 self-assigned this Oct 6, 2026
@Issacboy123
Issacboy123 requested a review from Pepps233 October 6, 2026 04:54

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants