Dashboard › opencode-lore › Distillation
de2a77b5-8baf-47d9-8de4-d7a76124307e["lore_tm_v1_yvNKxTCgwn-GsjogBi50TNyov4UL3HXdM-1tsmIJ0tE"]
Date: Sep 16, 2026
packages/gateway/test/openai-responses-recall-aware-stream.test.ts add response.reasoning_summary_text.done for reasoning item rs_0 with output_index: 0, summary_index: 0, and text "reasoned" before recallCall(1, { query: "architecture" }), strengthening the incomplete-principal recall regression scenario.finalizeResponsesAcc calls via acc.rawItems.get(0), and that their expected parts have length scenario.partIndex + 1.event: response.failed.maxRetainedStateBytes, including an indexed-SSE-data-sized limit and a 128 * 1024 byte limit, plus scenarios covering many retained items and terminal usage flushing.