🤖 fix: a new workspace waits for its branch list before it can be sent

src/browser/features/ChatInput/index.tsx

@@ -135,7 +135,7 @@ import {
 } from "@/browser/utils/chatEditing";
 
 import { type ThinkingLevel } from "@/common/types/thinking";
-import { normalizeRuntimeEnablement } from "@/common/types/runtime";
+import { normalizeRuntimeEnablement, RUNTIME_MODE } from "@/common/types/runtime";
 import {
   type MuxMessageMetadata,
   type ReviewNoteDataForDisplay,
@@ -936,6 +936,13 @@ const ChatInputInner: React.FC<ChatInputProps> = (props) => {
   // Disable send while Coder presets are loading (user could bypass preset validation)
   const coderPresetsLoading =
     coderState.enabled && !coderState.coderConfig?.existingWorkspace && coderState.loadingPresets;
+  // #6033: the trunk branch stays empty until the branch list loads, and the backend refuses
+  // every runtime except local without one, so a creation send must wait for the list.
+  const creationBranchesLoading =
+    variant === "creation" &&
+    props.kind !== "scratch" &&
+    !creationState.branchesLoaded &&
+    creationState.selectedRuntime.mode !== RUNTIME_MODE.LOCAL;
   const isProcessingAttachments = processingAttachmentCount > 0;
   const hasSendableDraft = hasTypedText || hasImages || hasReviews;
   // Workspace sends/edits mutate history relative to the visible transcript, so they wait for
@@ -958,6 +965,7 @@ const ChatInputInner: React.FC<ChatInputProps> = (props) => {
     !sendInFlightBlocksInput &&
     !isProcessingAttachments &&
     !coderPresetsLoading &&
+    !creationBranchesLoading &&
     !transcriptBlocksSend &&
     !editPreconditionInvalidated;
   const runningGoalActive =

src/browser/features/ChatInput/useCreationWorkspace.ts

@@ -355,32 +355,33 @@ export function useCreationWorkspace({
     let mounted = true;
     setBranchesLoaded(false);
     setRuntimeAvailabilityState({ status: "loading" });
-    const doLoad = async () => {
+    // Both requests run at once and settle independently. Send waits for the branch list (#6033),
+    // so it must not also wait for the runtime probe (Docker, Podman, devcontainer CLI), which can
+    // take several seconds after the branches are ready.
+    const loadBranches = async () => {
       try {
-        // Use allSettled so failures are independent - branches can load even if availability fails
-        const [branchResult, availabilityResult] = await Promise.allSettled([
-          api.projects.listBranches({ projectPath }),
-          api.projects.runtimeAvailability({ projectPath }),
-        ]);
+        const result = await api.projects.listBranches({ projectPath });
         if (!mounted) return;
-        if (branchResult.status === "fulfilled") {
-          setBranches(branchResult.value.branches);
-          setRecommendedTrunk(branchResult.value.recommendedTrunk);
-        } else {
-          console.error("Failed to load branches:", branchResult.reason);
-        }
-        if (availabilityResult.status === "fulfilled") {
-          setRuntimeAvailabilityState({ status: "loaded", data: availabilityResult.value });
-        } else {
-          setRuntimeAvailabilityState({ status: "failed" });
-        }
+        setBranches(result.branches);
+        setRecommendedTrunk(result.recommendedTrunk);
+      } catch (error) {
+        console.error("Failed to load branches:", error);
       } finally {
         if (mounted) {
           setBranchesLoaded(true);
         }
       }
     };
-    void doLoad();
+    const loadRuntimeAvailability = async () => {
+      try {
+        const data = await api.projects.runtimeAvailability({ projectPath });
+        if (mounted) setRuntimeAvailabilityState({ status: "loaded", data });
+      } catch {
+        if (mounted) setRuntimeAvailabilityState({ status: "failed" });
+      }
+    };
+    void loadBranches();
+    void loadRuntimeAvailability();
     return () => {
       mounted = false;
     };