gorazen 1a7ac1c98a Per-tab streaming robustness + release workflow hardening ... Tab streaming fix: - Add ensureAssistantStreamMessage() so a text_delta / thinking_start arriving while the assistant message hasn't been initialized yet (e.g. after switching back to a tab) creates one on the fly and continues streaming. - loadMessages() prefers the RPC getMessages() path while a tab is actively streaming, so the UI sees the in-flight tail even if the on-disk jsonl is lagging behind. Per-tab model/effort: - Add applyTabRuntimePreferences(tab) that re-pushes the saved modelProvider / modelId / thinkingLevel to that tab's client when the client is (re)created. - Add updateActiveTabModel() / updateActiveTabThinkingLevel() that write only to the active tab and to that tab's client, never to the global plugin settings. UI polish: - Move margin-inline-start: auto from the speed pill to the title wrapper, so hiding the pill no longer collapses the right-side header actions. Release workflow: - Pass --latest=true to gh release create so the highest tag gets the 'Latest' badge automatically (avoid manual promotion). Update CHANGELOG.		2026-06-10 18:36:47 +08:00
..
release.yml	Per-tab streaming robustness + release workflow hardening	2026-06-10 18:36:47 +08:00