- Gateway/models: skip external OpenRouter and LiteLLM pricing refreshes for local/self-hosted model endpoints so startup does not wait on remote pricing catalogs for local-only Ollama, vLLM, and compatible providers. Thanks @codex.

- CLI/plugins: stop security-blocked plugin installs from retrying as hook packs, so normal plugin packages report the scanner failure without a misleading "not a valid hook pack" follow-up. Fixes #61175; supersedes #64102. Thanks @KonsultDigital and @ziyincody.

- Agents/Anthropic: strip stale trailing assistant prefill turns from outbound replay so context-engine short circuits cannot send unsupported assistant-prefill payloads to provider APIs. Fixes #72556. Thanks @Veda-openclaw.

- Agents/Google: strip stale trailing assistant/model prefill turns from Gemini outbound replay so Google Generative AI requests end with a user turn or function response. Follow-up to #72556. Thanks @Veda-openclaw.

- Control UI/Dreaming: require explicit confirmation before applying restart-impacting Dreaming mode changes, with restart warning copy and loading feedback. Fixes #63804. (#63807) Thanks @bbddbb1.

- CLI/update: keep the automatic post-update completion refresh on the core-command tree so it no longer stages bundled plugin runtime deps before the Gateway restart path, avoiding `.24` update hangs and 1006 disconnect cascades. Fixes #72665. Thanks @sakalaboator and @He-Pin.

- Agents/Bedrock: stop heartbeat runs from persisting blank user transcript turns and repair existing blank user text messages before replay, preventing AWS Bedrock `ContentBlock` blank-text validation failures. Fixes #72640 and #72622. Thanks @goldzulu.

expect(seenContext.messages).not.toBe(messages);

});

it("strips trailing assistant prefill turns for Gemini outbound replay", async () => {

const messages = [

{

role: "user",

content: [{ type: "text", text: "earlier question" }],

},

{

role: "assistant",

content: [{ type: "text", text: "stale model answer" }],

},

];

const baseFn = vi.fn((_model, _context) =>

createFakeStream({ events: [], resultMessage: { role: "assistant", content: [] } }),

);

const wrapped = wrapStreamFnSanitizeMalformedToolCalls(baseFn as never, new Set(["read"]), {

validateGeminiTurns: true,

preserveSignatures: true,

dropThinkingBlocks: false,

} as never);

const stream = wrapped(

{ api: "google-generative-ai" } as never,

{ messages } as never,

{} as never,

) as FakeWrappedStream | Promise<FakeWrappedStream>;

await Promise.resolve(stream);

expect(baseFn).toHaveBeenCalledTimes(1);

const seenContext = baseFn.mock.calls[0]?.[1] as { messages: unknown[] };

expect(seenContext.messages).toEqual([

{

role: "user",

content: [{ type: "text", text: "earlier question" }],

},

]);

expect(seenContext.messages).not.toBe(messages);

});

it("drops signed thinking turns when sibling replay tool calls are not allowlisted", async () => {

const messages = [

{

let nextMessages = replayInputsChanged

? sanitizeToolUseResultPairing(sanitized.messages)

: sanitized.messages;

let strippedTrailingAssistantPrefill = false;

if (transcriptPolicy?.validateAnthropicTurns) {

nextMessages = sanitizeAnthropicReplayToolResults(nextMessages, {

disallowEmbeddedUserToolResultsForSignedThinkingReplay: allowProviderOwnedThinkingReplay,

});

}

if (transcriptPolicy?.validateAnthropicTurns || transcriptPolicy?.validateGeminiTurns) {

const beforeStrip = nextMessages;

nextMessages = stripTrailingAssistantPrefillTurns(nextMessages);

strippedTrailingAssistantPrefill ||= nextMessages !== beforeStrip;

}

if (nextMessages === messages) {

return baseFn(model, context, options);

}

if (sanitized.droppedAssistantMessages > 0 || transcriptPolicy?.validateAnthropicTurns) {

if (

sanitized.droppedAssistantMessages > 0 ||

transcriptPolicy?.validateAnthropicTurns ||

strippedTrailingAssistantPrefill

) {

if (transcriptPolicy?.validateGeminiTurns) {

nextMessages = validateGeminiTurns(nextMessages);

}