refactor(app): replace voice hooks with voice runtime and audio engine architecture
This commit is contained in:
@@ -103,6 +103,12 @@ export function buildToolCallDisplayModel(input: ToolCallDisplayInput): ToolCall
|
||||
summary = isRecord(input.metadata) ? readString(input.metadata.subAgentActivity) : undefined;
|
||||
} else if (lowerName === "thinking" && input.detail.type === "unknown") {
|
||||
displayName = "Thinking";
|
||||
} else if (lowerName === "terminal") {
|
||||
displayName = "Interacted with terminal";
|
||||
summary =
|
||||
input.detail.type === "plain_text"
|
||||
? readString(input.detail.label)
|
||||
: undefined;
|
||||
}
|
||||
|
||||
const errorText = input.status === "failed" ? formatErrorText(input.error) : undefined;
|
||||
|
||||
Reference in New Issue
Block a user