assistantTexts: string[];

toolMetas: ToolMetaEntry[];

lastAssistant: AssistantMessage | undefined;

currentAssistant?: AssistantMessage | null;

lastToolError?: ToolErrorSummary;

config?: OpenClawConfig;

isCronTrigger?: boolean;

const useMarkdown = params.toolResultFormat === "markdown";

const suppressAssistantArtifacts =

params.didSendDeterministicApprovalPrompt === true || hasSourceReplyPayload;

const lastAssistantStopReason = params.lastAssistant?.stopReason;

const nonEmptyAssistantTexts = params.assistantTexts.filter((text) => text.trim().length > 0);

const currentAssistant = params.currentAssistant ?? undefined;

const assistantForPayload =

currentAssistant ?? (nonEmptyAssistantTexts.length === 1 ? undefined : params.lastAssistant);

const lastAssistantStopReason = assistantForPayload?.stopReason;

const lastAssistantErrored = lastAssistantStopReason === "error";

const lastAssistantAborted = lastAssistantStopReason === "aborted";

const runAborted = params.runAborted === true || lastAssistantAborted;

const lastAssistantNeedsErrorSurface = lastAssistantErrored || lastAssistantAborted;

const errorText =

params.lastAssistant && lastAssistantNeedsErrorSurface

assistantForPayload && lastAssistantNeedsErrorSurface

? suppressAssistantArtifacts

? undefined

: formatAssistantErrorText(params.lastAssistant, {

: formatAssistantErrorText(assistantForPayload, {

cfg: params.config,

sessionKey: params.sessionKey,

provider: params.provider,

model: params.model,

})

: undefined;

const rawErrorMessage = lastAssistantNeedsErrorSurface

? normalizeOptionalString(params.lastAssistant?.errorMessage)

? normalizeOptionalString(assistantForPayload?.errorMessage)

: undefined;

const rawErrorFingerprint = rawErrorMessage

? getApiErrorPayloadFingerprint(rawErrorMessage)

const reasoningText =

suppressAssistantArtifacts || runAborted

? ""

: params.lastAssistant && params.reasoningLevel === "on" && params.thinkingLevel !== "off"

? extractAssistantThinking(params.lastAssistant)

: assistantForPayload && params.reasoningLevel === "on" && params.thinkingLevel !== "off"

? extractAssistantThinking(assistantForPayload)

: "";

if (reasoningText) {

replyItems.push({ text: reasoningText, isReasoning: true });

}

const fallbackAnswerText = params.lastAssistant

? extractAssistantVisibleText(params.lastAssistant)

const fallbackAnswerText = assistantForPayload

? extractAssistantVisibleText(assistantForPayload)

: "";

const fallbackRawAnswerText = resolveRawAssistantAnswerText(params.lastAssistant);

const fallbackRawAnswerText = resolveRawAssistantAnswerText(assistantForPayload);

const shouldSuppressRawErrorText = (text: string) => {

if (!lastAssistantNeedsErrorSurface) {

return false;

const parsed = parseReplyDirectives(text);

return (parsed.mediaUrls?.length ?? 0) > 0 || parsed.audioAsVoice;

});

const nonEmptyAssistantTexts = params.assistantTexts.filter((text) => text.trim().length > 0);

const normalizedAssistantTexts = normalizeTextForComparison(nonEmptyAssistantTexts.join("\n\n"));

const normalizedRawAnswerText = normalizeTextForComparison(rawAnswerDirectiveState?.text ?? "");

const shouldPreferRawAnswerText =

? normalizeReplyTextForComparison(fallbackAnswerSourceText)

: "";

const shouldUseCanonicalFinalAnswer =

nonEmptyAssistantTexts.length > 1 &&

!lastAssistantNeedsErrorSurface &&

fallbackAnswerSourceText.length > 0 &&

normalizedFallbackAnswerSourceText.length > 0;

const hasAssistantTextPayload = nonEmptyAssistantTexts.length > 0;