@nick3/copilot-api 1.3.5 → 1.4.0

package/dist/{server-CM_0PrbK.js → server-D3sySKxC.js}

@@ -1,8 +1,8 @@
 import { PATHS } from "./paths-DoT4SZ8f.js";
 import { listAccountsFromRegistry } from "./accounts-registry-c7rs5Ed9.js";
-import { HTTPError, accountFromState, cacheModels, copilotBaseUrl, copilotHeaders, forwardError, generateRequestIdFromPayload, getCopilotUsage, getRootSessionId, getUUID, isNullish, prepareInteractionHeaders, sleep, state } from "./utils-BUJfM1V2.js";
-import "./get-copilot-token-BwP_PxV5.js";
-import { PROVIDER_TYPE_ANTHROPIC, accountsManager, getAliasTargetSet, getConfig, getExtraPromptForModel, getModelAliases, getModelAliasesInfo, getModelRefreshIntervalMs, getProviderConfig, getReasoningEffortForModel, getSmallModel, isForceAgentEnabled, isFreeModelLoadBalancingEnabled, isMessageStartInputTokensFallbackEnabled, isMessagesApiEnabled, isResponsesApiContextManagementModel, mergeConfigWithDefaults, shouldCompactUseSmallModel } from "./accounts-manager-DjGzZIcp.js";
+import { HTTPError, accountFromState, cacheModels, copilotBaseUrl, copilotHeaders, forwardError, generateRequestIdFromPayload, getCopilotUsage, getRootSessionId, getUUID, isNullish, parseUserIdMetadata, prepareForCompact, prepareInteractionHeaders, sleep, state } from "./utils-D8j9lvS0.js";
+import "./get-copilot-token-BA1FaCgQ.js";
+import { PROVIDER_TYPE_ANTHROPIC, accountsManager, getAliasTargetSet, getConfig, getExtraPromptForModel, getModelAliases, getModelAliasesInfo, getModelRefreshIntervalMs, getProviderConfig, getReasoningEffortForModel, getSmallModel, isForceAgentEnabled, isFreeModelLoadBalancingEnabled, isMessageStartInputTokensFallbackEnabled, isMessagesApiEnabled, isResponsesApiContextManagementModel, mergeConfigWithDefaults, shouldCompactUseSmallModel } from "./accounts-manager-BQCAoBZp.js";
 import consola from "consola";
 import fs, { readFile } from "node:fs/promises";
 import * as path$1 from "node:path";
@@ -12,6 +12,7 @@ import { Hono } from "hono";
 import { cors } from "hono/cors";
 import { logger } from "hono/logger";
 import fs$1, { existsSync } from "node:fs";
+import { AsyncLocalStorage } from "node:async_hooks";
 import { Database } from "bun:sqlite";
 import { fileURLToPath } from "node:url";
 import { streamSSE } from "hono/streaming";
@@ -104,6 +105,40 @@ function createAuthMiddleware(options = {}) {
 	};
 }
 
+//#endregion
+//#region src/lib/request-context.ts
+const TRACE_ID_MAX_LENGTH = 64;
+const TRACE_ID_PATTERN = /^\w[\w.-]*$/;
+const asyncLocalStorage = new AsyncLocalStorage();
+const requestContext = {
+	getStore: () => asyncLocalStorage.getStore(),
+	run: (context, callback) => asyncLocalStorage.run(context, callback)
+};
+function generateTraceId() {
+	const timestamp = Date.now().toString(36);
+	const random = Math.random().toString(36).slice(2, 8);
+	return `${timestamp}-${random}`;
+}
+function resolveTraceId(traceId) {
+	const candidate = traceId?.trim();
+	if (!candidate || candidate.length > TRACE_ID_MAX_LENGTH || !TRACE_ID_PATTERN.test(candidate)) return generateTraceId();
+	return candidate;
+}
+
+//#endregion
+//#region src/lib/trace.ts
+const traceIdMiddleware = async (c, next) => {
+	const traceId = resolveTraceId(c.req.header("x-trace-id"));
+	c.header("x-trace-id", traceId);
+	const context = {
+		traceId,
+		startTime: Date.now()
+	};
+	await requestContext.run(context, async () => {
+		await next();
+	});
+};
+
 //#endregion
 //#region src/lib/admin-db.ts
 const DEFAULT_DB_PATH = path.join(PATHS.APP_DIR, "admin.sqlite");
@@ -2148,12 +2183,14 @@ const createHandlerLogger = (name) => {
 			cleanupOldLogs();
 			lastCleanup = Date.now();
 		}
+		const traceId = requestContext.getStore()?.traceId;
 		const date = logObj.date;
 		const dateKey = date.toLocaleDateString("sv-SE");
 		const timestamp = date.toLocaleString("sv-SE", { hour12: false });
 		const filePath = path.join(LOG_DIR, `${sanitizedName}-${dateKey}.log`);
 		const message = formatArgs(logObj.args);
-		const line = `[${timestamp}] [${logObj.type}] [${logObj.tag || name}]${message ? ` ${message}` : ""}`;
+		const traceIdStr = traceId ? ` [${traceId}]` : "";
+		const line = `[${timestamp}] [${logObj.type}] [${logObj.tag || name}]${traceIdStr}${message ? ` ${message}` : ""}`;
 		appendLine(filePath, line);
 	} });
 	return instance;
@@ -2402,568 +2439,108 @@ const getTokenCount = async (payload, model) => {
 };
 
 //#endregion
-//#region src/services/copilot/create-responses.ts
-const createResponses = async (payload, { vision, initiator, upstreamRequestId, subagentMarker, sessionId }, account) => {
+//#region src/services/copilot/create-chat-completions.ts
+function isGpt5MiniFamily(modelId) {
+	return modelId === "gpt-5-mini" || modelId.startsWith("gpt-5-mini-");
+}
+function applyDefaultReasoningEffort(payload) {
+	if (!isGpt5MiniFamily(payload.model)) return payload;
+	if (payload.reasoning_effort !== null && payload.reasoning_effort !== void 0) return payload;
+	return {
+		...payload,
+		reasoning_effort: getReasoningEffortForModel("gpt-5-mini")
+	};
+}
+const getChatInitiator = (messages) => {
+	if (isForceAgentEnabled()) return messages.some((msg) => ["assistant", "tool"].includes(msg.role)) ? "agent" : "user";
+	const lastMessage = messages.at(-1);
+	if (!lastMessage) return "user";
+	return ["assistant", "tool"].includes(lastMessage.role) ? "agent" : "user";
+};
+const createChatCompletions = async (payload, account, options) => {
 	const ctx = account ?? accountFromState();
 	if (!ctx.copilotToken) throw new Error("Copilot token not found");
+	const enableVision = payload.messages.some((x) => typeof x.content !== "string" && x.content?.some((x$1) => x$1.type === "image_url"));
+	const initiator = options?.initiator ?? getChatInitiator(payload.messages);
 	const headers = {
-		...copilotHeaders(ctx, vision, upstreamRequestId),
-		"x-initiator": initiator
+		...copilotHeaders(ctx, enableVision, options?.upstreamRequestId),
+		"x-initiator": options?.subagentMarker ? "agent" : initiator
 	};
-	prepareInteractionHeaders(sessionId, Boolean(subagentMarker), headers);
-	payload.service_tier = null;
-	const response = await fetch(`${copilotBaseUrl(ctx)}/responses`, {
+	prepareInteractionHeaders(options?.sessionId, Boolean(options?.subagentMarker), headers);
+	const upstreamPayload = applyDefaultReasoningEffort(payload);
+	prepareForCompact(headers, options?.isCompact);
+	const response = await fetch(`${copilotBaseUrl(ctx)}/chat/completions`, {
 		method: "POST",
 		headers,
-		body: JSON.stringify(payload)
+		body: JSON.stringify(upstreamPayload)
 	});
 	if (!response.ok) {
-		consola.error("Failed to create responses", response);
-		throw new HTTPError("Failed to create responses", response);
+		consola.error("Failed to create chat completions", response);
+		throw new HTTPError("Failed to create chat completions", response);
 	}
 	if (payload.stream) return events(response);
 	return await response.json();
 };
 
 //#endregion
-//#region src/routes/messages/responses-translation.ts
-const MESSAGE_TYPE = "message";
-const COMPACTION_SIGNATURE_PREFIX = "cm1#";
-const COMPACTION_SIGNATURE_SEPARATOR = "@";
-const THINKING_TEXT$1 = "Thinking...";
-const translateAnthropicMessagesToResponsesPayload = (payload, modelOverride) => {
-	const model = modelOverride ?? payload.model;
-	const input = [];
-	const applyPhase = shouldApplyPhase(payload.model);
-	for (const message of payload.messages) input.push(...translateMessage(message, payload.model, applyPhase));
-	const translatedTools = convertAnthropicTools(payload.tools);
-	const toolChoice = convertAnthropicToolChoice(payload.tool_choice);
-	const { safetyIdentifier, promptCacheKey } = parseUserId(payload.metadata?.user_id);
-	return {
-		model,
-		input,
-		instructions: translateSystemPrompt(payload.system, model),
-		temperature: 1,
-		top_p: payload.top_p ?? null,
-		max_output_tokens: Math.max(payload.max_tokens, 12800),
-		tools: translatedTools,
-		tool_choice: toolChoice,
-		metadata: payload.metadata ? { ...payload.metadata } : null,
-		safety_identifier: safetyIdentifier,
-		prompt_cache_key: promptCacheKey,
-		stream: payload.stream ?? null,
-		store: false,
-		parallel_tool_calls: true,
-		reasoning: {
-			effort: getReasoningEffortForModel(model),
-			summary: "auto"
-		},
-		include: ["reasoning.encrypted_content"]
-	};
-};
-const encodeCompactionCarrierSignature = (compaction) => {
-	return `${COMPACTION_SIGNATURE_PREFIX}${compaction.encrypted_content}${COMPACTION_SIGNATURE_SEPARATOR}${compaction.id}`;
-};
-const decodeCompactionCarrierSignature = (signature) => {
-	if (signature.startsWith(COMPACTION_SIGNATURE_PREFIX)) {
-		const raw = signature.slice(4);
-		const separatorIndex = raw.indexOf(COMPACTION_SIGNATURE_SEPARATOR);
-		if (separatorIndex <= 0 || separatorIndex === raw.length - 1) return;
-		const encrypted_content = raw.slice(0, separatorIndex);
-		const id = raw.slice(separatorIndex + 1);
-		if (!encrypted_content) return;
-		return {
-			id,
-			encrypted_content
-		};
-	}
-};
-const translateMessage = (message, model, applyPhase) => {
-	if (message.role === "user") return translateUserMessage(message);
-	return translateAssistantMessage(message, model, applyPhase);
-};
-const translateUserMessage = (message) => {
-	if (typeof message.content === "string") return [createMessage("user", message.content)];
-	if (!Array.isArray(message.content)) return [];
-	const items = [];
-	const pendingContent = [];
-	for (const block of message.content) {
-		if (block.type === "tool_result") {
-			flushPendingContent(pendingContent, items, { role: "user" });
-			items.push(createFunctionCallOutput(block));
-			continue;
-		}
-		const converted = translateUserContentBlock(block);
-		if (converted) pendingContent.push(converted);
+//#region src/routes/chat-completions/handler.ts
+const logger$6 = createHandlerLogger("chat-completions-handler");
+const CHAT_COMPLETIONS_ENDPOINT$1 = "/chat/completions";
+async function handleCompletion$1(c) {
+	await checkRateLimit(state);
+	const store = getRequestHistoryStore();
+	const request = buildRequestContext$1(c);
+	const payload = await c.req.json();
+	const clientModel = payload.model;
+	const streamRequested = Boolean(payload.stream);
+	const initiator = getChatInitiator(payload.messages);
+	const userId = payload.user ?? void 0;
+	const { safetyIdentifier, sessionId: promptCacheKey } = parseUserIdMetadata(userId);
+	const normalizedSafetyIdentifier = safetyIdentifier ?? void 0;
+	const normalizedPromptCacheKey = promptCacheKey ?? void 0;
+	request.userId = userId;
+	request.safetyIdentifier = normalizedSafetyIdentifier;
+	request.promptCacheKey = normalizedPromptCacheKey;
+	request.initiator = initiator;
+	if (getAliasTargetSet().has(clientModel.toLowerCase())) {
+		recordSelectionFailure$2(store, {
+			request,
+			clientModel,
+			stream: streamRequested,
+			reason: "MODEL_NOT_SUPPORTED"
+		});
+		return selectionFailureResponse$2(c, {
+			clientModel,
+			reason: "MODEL_NOT_SUPPORTED"
+		});
 	}
-	flushPendingContent(pendingContent, items, { role: "user" });
-	return items;
-};
-const translateAssistantMessage = (message, model, applyPhase) => {
-	const assistantPhase = resolveAssistantPhase(model, message.content, applyPhase);
-	if (typeof message.content === "string") return [createMessage("assistant", message.content, assistantPhase)];
-	if (!Array.isArray(message.content)) return [];
-	const items = [];
-	const pendingContent = [];
-	for (const block of message.content) {
-		if (block.type === "tool_use") {
-			flushPendingContent(pendingContent, items, {
-				role: "assistant",
-				phase: assistantPhase
-			});
-			items.push(createFunctionToolCall(block));
-			continue;
-		}
-		if (block.type === "thinking" && block.signature) {
-			const compactionContent = createCompactionContent(block);
-			if (compactionContent) {
-				flushPendingContent(pendingContent, items, {
-					role: "assistant",
-					phase: assistantPhase
-				});
-				items.push(compactionContent);
-				continue;
-			}
-			if (block.signature.includes("@")) {
-				flushPendingContent(pendingContent, items, {
-					role: "assistant",
-					phase: assistantPhase
-				});
-				items.push(createReasoningContent(block));
-				continue;
-			}
-		}
-		const converted = translateAssistantContentBlock(block);
-		if (converted) pendingContent.push(converted);
+	logger$6.debug("Request payload:", JSON.stringify(payload).slice(-400));
+	const selection = await accountsManager.selectAccountForRequest([{
+		modelId: clientModel,
+		endpoint: CHAT_COMPLETIONS_ENDPOINT$1
+	}]);
+	if (!selection.ok) {
+		recordSelectionFailure$2(store, {
+			request,
+			clientModel,
+			stream: streamRequested,
+			reason: selection.reason
+		});
+		return selectionFailureResponse$2(c, {
+			clientModel,
+			reason: selection.reason
+		});
 	}
-	flushPendingContent(pendingContent, items, {
-		role: "assistant",
-		phase: assistantPhase
-	});
-	return items;
-};
-const translateUserContentBlock = (block) => {
-	switch (block.type) {
-		case "text": return createTextContent(block.text);
-		case "image": return createImageContent(block);
-		default: return;
-	}
-};
-const translateAssistantContentBlock = (block) => {
-	switch (block.type) {
-		case "text": return createOutPutTextContent(block.text);
-		default: return;
-	}
-};
-const flushPendingContent = (pendingContent, target, message) => {
-	if (pendingContent.length === 0) return;
-	const messageContent = [...pendingContent];
-	target.push(createMessage(message.role, messageContent, message.phase));
-	pendingContent.length = 0;
-};
-const createMessage = (role, content, phase) => ({
-	type: MESSAGE_TYPE,
-	role,
-	content,
-	...role === "assistant" && phase ? { phase } : {}
-});
-const resolveAssistantPhase = (_model, content, applyPhase) => {
-	if (!applyPhase) return;
-	if (typeof content === "string") return "final_answer";
-	if (!Array.isArray(content)) return;
-	if (!content.some((block) => block.type === "text")) return;
-	return content.some((block) => block.type === "tool_use") ? "commentary" : "final_answer";
-};
-const shouldApplyPhase = (model) => {
-	return getExtraPromptForModel(model).includes("## Intermediary updates");
-};
-const createTextContent = (text) => ({
-	type: "input_text",
-	text
-});
-const createOutPutTextContent = (text) => ({
-	type: "output_text",
-	text
-});
-const createImageContent = (block) => ({
-	type: "input_image",
-	image_url: `data:${block.source.media_type};base64,${block.source.data}`,
-	detail: "auto"
-});
-const createReasoningContent = (block) => {
-	const { encryptedContent, id } = parseReasoningSignature(block.signature);
-	const thinking = block.thinking === THINKING_TEXT$1 ? "" : block.thinking;
-	return {
-		id,
-		type: "reasoning",
-		summary: thinking ? [{
-			type: "summary_text",
-			text: thinking
-		}] : [],
-		encrypted_content: encryptedContent
-	};
-};
-const createCompactionContent = (block) => {
-	const compaction = decodeCompactionCarrierSignature(block.signature);
-	if (!compaction) return;
-	return {
-		id: compaction.id,
-		type: "compaction",
-		encrypted_content: compaction.encrypted_content
-	};
-};
-const parseReasoningSignature = (signature) => {
-	const splitIndex = signature.lastIndexOf("@");
-	if (splitIndex <= 0 || splitIndex === signature.length - 1) return {
-		encryptedContent: signature,
-		id: ""
-	};
-	return {
-		encryptedContent: signature.slice(0, splitIndex),
-		id: signature.slice(splitIndex + 1)
-	};
-};
-const createFunctionToolCall = (block) => ({
-	type: "function_call",
-	call_id: block.id,
-	name: block.name,
-	arguments: JSON.stringify(block.input),
-	status: "completed"
-});
-const createFunctionCallOutput = (block) => ({
-	type: "function_call_output",
-	call_id: block.tool_use_id,
-	output: convertToolResultContent(block.content),
-	status: block.is_error ? "incomplete" : "completed"
-});
-const translateSystemPrompt = (system, model) => {
-	if (!system) return null;
-	const extraPrompt = getExtraPromptForModel(model);
-	if (typeof system === "string") return system + extraPrompt;
-	const text = system.map((block, index) => {
-		if (index === 0) return block.text + extraPrompt;
-		return block.text;
-	}).join(" ");
-	return text.length > 0 ? text : null;
-};
-const convertAnthropicTools = (tools) => {
-	if (!tools || tools.length === 0) return null;
-	return tools.map((tool) => ({
-		type: "function",
-		name: tool.name,
-		parameters: tool.input_schema,
-		strict: false,
-		...tool.description ? { description: tool.description } : {}
-	}));
-};
-const convertAnthropicToolChoice = (choice) => {
-	if (!choice) return "auto";
-	switch (choice.type) {
-		case "auto": return "auto";
-		case "any": return "required";
-		case "tool": return choice.name ? {
-			type: "function",
-			name: choice.name
-		} : "auto";
-		case "none": return "none";
-		default: return "auto";
-	}
-};
-const translateResponsesResultToAnthropic = (response) => {
-	const contentBlocks = mapOutputToAnthropicContent(response.output);
-	const usage = mapResponsesUsage(response);
-	let anthropicContent = fallbackContentBlocks(response.output_text);
-	if (contentBlocks.length > 0) anthropicContent = contentBlocks;
-	const stopReason = mapResponsesStopReason(response);
-	return {
-		id: response.id,
-		type: "message",
-		role: "assistant",
-		content: anthropicContent,
-		model: response.model,
-		stop_reason: stopReason,
-		stop_sequence: null,
-		usage
-	};
-};
-const mapOutputToAnthropicContent = (output) => {
-	const contentBlocks = [];
-	for (const item of output) switch (item.type) {
-		case "reasoning": {
-			const thinkingText = extractReasoningText(item);
-			if (thinkingText.length > 0) contentBlocks.push({
-				type: "thinking",
-				thinking: thinkingText,
-				signature: (item.encrypted_content ?? "") + "@" + item.id
-			});
-			break;
-		}
-		case "function_call": {
-			const toolUseBlock = createToolUseContentBlock(item);
-			if (toolUseBlock) contentBlocks.push(toolUseBlock);
-			break;
-		}
-		case "message": {
-			const combinedText = combineMessageTextContent(item.content);
-			if (combinedText.length > 0) contentBlocks.push({
-				type: "text",
-				text: combinedText
-			});
-			break;
-		}
-		case "compaction": {
-			const compactionBlock = createCompactionThinkingBlock(item);
-			if (compactionBlock) contentBlocks.push(compactionBlock);
-			break;
-		}
-		default: {
-			const combinedText = combineMessageTextContent(item.content);
-			if (combinedText.length > 0) contentBlocks.push({
-				type: "text",
-				text: combinedText
-			});
-		}
-	}
-	return contentBlocks;
-};
-const combineMessageTextContent = (content) => {
-	if (!Array.isArray(content)) return "";
-	let aggregated = "";
-	for (const block of content) {
-		if (isResponseOutputText(block)) {
-			aggregated += block.text;
-			continue;
-		}
-		if (isResponseOutputRefusal(block)) {
-			aggregated += block.refusal;
-			continue;
-		}
-		if (typeof block.text === "string") {
-			aggregated += block.text;
-			continue;
-		}
-		if (typeof block.reasoning === "string") {
-			aggregated += block.reasoning;
-			continue;
-		}
-	}
-	return aggregated;
-};
-const extractReasoningText = (item) => {
-	const segments = [];
-	const collectFromBlocks = (blocks) => {
-		if (!Array.isArray(blocks)) return;
-		for (const block of blocks) if (typeof block.text === "string") {
-			segments.push(block.text);
-			continue;
-		}
-	};
-	if (!item.summary || item.summary.length === 0) return THINKING_TEXT$1;
-	collectFromBlocks(item.summary);
-	return segments.join("").trim();
-};
-const createToolUseContentBlock = (call) => {
-	const toolId = call.call_id;
-	if (!call.name || !toolId) return null;
-	const input = parseFunctionCallArguments(call.arguments);
-	return {
-		type: "tool_use",
-		id: toolId,
-		name: call.name,
-		input
-	};
-};
-const createCompactionThinkingBlock = (item) => {
-	if (!item.id || !item.encrypted_content) return null;
-	return {
-		type: "thinking",
-		thinking: THINKING_TEXT$1,
-		signature: encodeCompactionCarrierSignature({
-			id: item.id,
-			encrypted_content: item.encrypted_content
-		})
-	};
-};
-const parseFunctionCallArguments = (rawArguments) => {
-	if (typeof rawArguments !== "string" || rawArguments.trim().length === 0) return {};
-	try {
-		const parsed = JSON.parse(rawArguments);
-		if (Array.isArray(parsed)) return { arguments: parsed };
-		if (parsed && typeof parsed === "object") return parsed;
-	} catch (error) {
-		consola.warn("Failed to parse function call arguments", {
-			error,
-			rawArguments
-		});
-	}
-	return { raw_arguments: rawArguments };
-};
-const fallbackContentBlocks = (outputText) => {
-	if (!outputText) return [];
-	return [{
-		type: "text",
-		text: outputText
-	}];
-};
-const mapResponsesStopReason = (response) => {
-	const { status, incomplete_details: incompleteDetails } = response;
-	if (status === "completed") {
-		if (response.output.some((item) => item.type === "function_call")) return "tool_use";
-		return "end_turn";
-	}
-	if (status === "incomplete") {
-		if (incompleteDetails?.reason === "max_output_tokens") return "max_tokens";
-		if (incompleteDetails?.reason === "content_filter") return "end_turn";
-	}
-	return null;
-};
-const mapResponsesUsage = (response) => {
-	const inputTokens = response.usage?.input_tokens ?? 0;
-	const outputTokens = response.usage?.output_tokens ?? 0;
-	const inputCachedTokens = response.usage?.input_tokens_details?.cached_tokens;
-	return {
-		input_tokens: inputTokens - (inputCachedTokens ?? 0),
-		output_tokens: outputTokens,
-		...response.usage?.input_tokens_details?.cached_tokens !== void 0 && { cache_read_input_tokens: response.usage.input_tokens_details.cached_tokens }
-	};
-};
-const isRecord = (value) => typeof value === "object" && value !== null;
-const isResponseOutputText = (block) => isRecord(block) && "type" in block && block.type === "output_text";
-const isResponseOutputRefusal = (block) => isRecord(block) && "type" in block && block.type === "refusal";
-const parseUserId = (userId) => {
-	if (!userId || typeof userId !== "string") return {
-		safetyIdentifier: null,
-		promptCacheKey: null
-	};
-	const userMatch = userId.match(/user_([^_]+)_account/);
-	const safetyIdentifier = userMatch ? userMatch[1] : null;
-	const sessionMatch = userId.match(/_session_(.+)$/);
-	const promptCacheKey = sessionMatch ? sessionMatch[1] : null;
-	return {
-		safetyIdentifier,
-		promptCacheKey
-	};
-};
-const convertToolResultContent = (content) => {
-	if (typeof content === "string") return content;
-	if (Array.isArray(content)) {
-		const result = [];
-		for (const block of content) switch (block.type) {
-			case "text":
-				result.push(createTextContent(block.text));
-				break;
-			case "image":
-				result.push(createImageContent(block));
-				break;
-			default: break;
-		}
-		return result;
-	}
-	return "";
-};
-
-//#endregion
-//#region src/services/copilot/create-chat-completions.ts
-function isGpt5MiniFamily(modelId) {
-	return modelId === "gpt-5-mini" || modelId.startsWith("gpt-5-mini-");
-}
-function applyDefaultReasoningEffort(payload) {
-	if (!isGpt5MiniFamily(payload.model)) return payload;
-	if (payload.reasoning_effort !== null && payload.reasoning_effort !== void 0) return payload;
-	return {
-		...payload,
-		reasoning_effort: getReasoningEffortForModel("gpt-5-mini")
-	};
-}
-const getChatInitiator = (messages) => {
-	if (isForceAgentEnabled()) return messages.some((msg) => ["assistant", "tool"].includes(msg.role)) ? "agent" : "user";
-	const lastMessage = messages.at(-1);
-	if (!lastMessage) return "user";
-	return ["assistant", "tool"].includes(lastMessage.role) ? "agent" : "user";
-};
-const createChatCompletions = async (payload, account, options) => {
-	const ctx = account ?? accountFromState();
-	if (!ctx.copilotToken) throw new Error("Copilot token not found");
-	const enableVision = payload.messages.some((x) => typeof x.content !== "string" && x.content?.some((x$1) => x$1.type === "image_url"));
-	const initiator = options?.initiator ?? getChatInitiator(payload.messages);
-	const headers = {
-		...copilotHeaders(ctx, enableVision, options?.upstreamRequestId),
-		"x-initiator": options?.subagentMarker ? "agent" : initiator
-	};
-	prepareInteractionHeaders(options?.sessionId, Boolean(options?.subagentMarker), headers);
-	const upstreamPayload = applyDefaultReasoningEffort(payload);
-	const response = await fetch(`${copilotBaseUrl(ctx)}/chat/completions`, {
-		method: "POST",
-		headers,
-		body: JSON.stringify(upstreamPayload)
-	});
-	if (!response.ok) {
-		consola.error("Failed to create chat completions", response);
-		throw new HTTPError("Failed to create chat completions", response);
-	}
-	if (payload.stream) return events(response);
-	return await response.json();
-};
-
-//#endregion
-//#region src/routes/chat-completions/handler.ts
-const logger$6 = createHandlerLogger("chat-completions-handler");
-const CHAT_COMPLETIONS_ENDPOINT$1 = "/chat/completions";
-async function handleCompletion$1(c) {
-	await checkRateLimit(state);
-	const store = getRequestHistoryStore();
-	const request = buildRequestContext$1(c);
-	const payload = await c.req.json();
-	const clientModel = payload.model;
-	const streamRequested = Boolean(payload.stream);
-	const initiator = getChatInitiator(payload.messages);
-	const userId = payload.user ?? void 0;
-	const { safetyIdentifier, promptCacheKey } = parseUserId(userId);
-	const normalizedSafetyIdentifier = safetyIdentifier ?? void 0;
-	const normalizedPromptCacheKey = promptCacheKey ?? void 0;
-	request.userId = userId;
-	request.safetyIdentifier = normalizedSafetyIdentifier;
-	request.promptCacheKey = normalizedPromptCacheKey;
-	request.initiator = initiator;
-	if (getAliasTargetSet().has(clientModel.toLowerCase())) {
-		recordSelectionFailure$2(store, {
-			request,
-			clientModel,
-			stream: streamRequested,
-			reason: "MODEL_NOT_SUPPORTED"
-		});
-		return selectionFailureResponse$2(c, {
-			clientModel,
-			reason: "MODEL_NOT_SUPPORTED"
-		});
-	}
-	logger$6.debug("Request payload:", JSON.stringify(payload).slice(-400));
-	const selection = await accountsManager.selectAccountForRequest([{
-		modelId: clientModel,
-		endpoint: CHAT_COMPLETIONS_ENDPOINT$1
-	}]);
-	if (!selection.ok) {
-		recordSelectionFailure$2(store, {
-			request,
-			clientModel,
-			stream: streamRequested,
-			reason: selection.reason
-		});
-		return selectionFailureResponse$2(c, {
-			clientModel,
-			reason: selection.reason
-		});
-	}
-	const { account, selectedModel } = selection;
-	const upstreamPayload = {
-		...payload,
-		model: selectedModel.id
-	};
-	const premiumRemainingBefore = account.premiumRemaining;
-	const premiumUnlimitedBefore = account.unlimited;
-	await logTokenCountForRequest({
-		payload: upstreamPayload,
-		selectedModel
+	const { account, selectedModel } = selection;
+	const upstreamPayload = {
+		...payload,
+		model: selectedModel.id
+	};
+	const premiumRemainingBefore = account.premiumRemaining;
+	const premiumUnlimitedBefore = account.unlimited;
+	await logTokenCountForRequest({
+		payload: upstreamPayload,
+		selectedModel
 	});
 	if (state.manualApprove) await awaitApproval();
 	const payloadWithMaxTokens = applyDefaultMaxTokens(upstreamPayload, selectedModel);
@@ -3311,600 +2888,1059 @@ async function handleNonStreamingRequest(params) {
 			errorMessage
 		});
 	}
-}
-const isNonStreaming$1 = (response) => Object.hasOwn(response, "choices");
-
-//#endregion
-//#region src/routes/chat-completions/route.ts
-const completionRoutes = new Hono();
-completionRoutes.post("/", async (c) => {
+}
+const isNonStreaming$1 = (response) => Object.hasOwn(response, "choices");
+
+//#endregion
+//#region src/routes/chat-completions/route.ts
+const completionRoutes = new Hono();
+completionRoutes.post("/", async (c) => {
+	try {
+		return await handleCompletion$1(c);
+	} catch (error) {
+		return await forwardError(c, error);
+	}
+});
+
+//#endregion
+//#region src/services/copilot/create-embeddings.ts
+const createEmbeddings = async (payload, account) => {
+	const ctx = account ?? accountFromState();
+	if (!ctx.copilotToken) throw new Error("Copilot token not found");
+	const response = await fetch(`${copilotBaseUrl(ctx)}/embeddings`, {
+		method: "POST",
+		headers: copilotHeaders(ctx),
+		body: JSON.stringify(payload)
+	});
+	if (!response.ok) throw new HTTPError("Failed to create embeddings", response);
+	return await response.json();
+};
+
+//#endregion
+//#region src/routes/embeddings/route.ts
+const embeddingRoutes = new Hono();
+const EMBEDDINGS_ENDPOINT = "/embeddings";
+embeddingRoutes.post("/", async (c) => {
+	try {
+		const store = getRequestHistoryStore();
+		const requestId = randomUUID();
+		const startedAtMs = Date.now();
+		const method = c.req.raw.method;
+		const path$2 = new URL(c.req.url, "http://local").pathname;
+		const { ip: clientIp, source: clientIpSource } = getClientIpInfo(c);
+		const userAgent = c.req.header("user-agent") ?? void 0;
+		const ctx = {
+			requestId,
+			startedAtMs,
+			method,
+			path: path$2,
+			clientIp,
+			clientIpSource,
+			userAgent
+		};
+		const payload = await c.req.json();
+		const clientModel = payload.model;
+		if (getAliasTargetSet().has(clientModel.toLowerCase())) {
+			recordSelectionFailure$1(store, {
+				ctx,
+				clientModel,
+				reason: "MODEL_NOT_SUPPORTED"
+			});
+			return selectionFailureResponse$1(c, clientModel, "MODEL_NOT_SUPPORTED");
+		}
+		const selection = await accountsManager.selectAccountForRequest([{
+			modelId: clientModel,
+			endpoint: EMBEDDINGS_ENDPOINT
+		}]);
+		if (!selection.ok) {
+			recordSelectionFailure$1(store, {
+				ctx,
+				clientModel,
+				reason: selection.reason
+			});
+			return selectionFailureResponse$1(c, clientModel, selection.reason);
+		}
+		const upstreamPayload = {
+			...payload,
+			model: selection.selectedModel.id
+		};
+		return await runEmbeddingsWithAccount({
+			c,
+			store,
+			ctx,
+			payload: upstreamPayload,
+			clientModel,
+			selection
+		});
+	} catch (error) {
+		return await forwardError(c, error);
+	}
+});
+function recordSelectionFailure$1(store, params) {
+	const { ctx, clientModel, reason } = params;
+	const finishedAtMs = Date.now();
+	store.insert({
+		requestId: ctx.requestId,
+		startedAtMs: ctx.startedAtMs,
+		finishedAtMs,
+		durationMs: finishedAtMs - ctx.startedAtMs,
+		method: ctx.method,
+		path: ctx.path,
+		upstreamEndpoint: EMBEDDINGS_ENDPOINT,
+		stream: false,
+		clientModel,
+		clientIp: ctx.clientIp,
+		clientIpSource: ctx.clientIpSource,
+		userAgent: ctx.userAgent,
+		httpStatus: reason === "MODEL_NOT_SUPPORTED" ? 400 : 429,
+		selectionFailureReason: reason
+	});
+}
+function selectionFailureResponse$1(c, clientModel, reason) {
+	if (reason === "MODEL_NOT_SUPPORTED") return c.json({ error: {
+		message: `Model "${clientModel}" is not available for any configured account.`,
+		type: "invalid_request_error"
+	} }, 400);
+	return c.json({ error: {
+		message: "All accounts have exhausted their quota. Please wait for quota refresh or add additional accounts.",
+		type: "rate_limit_error"
+	} }, 429);
+}
+async function runEmbeddingsWithAccount({ c, store, ctx, payload, clientModel, selection }) {
+	const { account, reservation, selectedModel, endpoint, costUnits } = selection;
+	const premiumRemainingBefore = account.premiumRemaining;
+	const premiumUnlimitedBefore = account.unlimited;
+	let httpStatus = 200;
+	let usage = {};
+	let errorName;
+	let errorStatus;
+	let errorMessage;
+	let finishedAtMs;
+	try {
+		const accountCtx = toAccountContext(account);
+		const response = await createEmbeddings(payload, accountCtx);
+		usage = normalizeEmbeddingsUsage(response.usage);
+		finishedAtMs = Date.now();
+		return c.json(response);
+	} catch (error) {
+		finishedAtMs = Date.now();
+		const details = extractErrorDetails(error);
+		httpStatus = details.httpStatus;
+		errorName = details.errorName;
+		errorStatus = details.errorStatus;
+		errorMessage = details.errorMessage;
+		if (details.unauthorized) accountsManager.markAccountFailed(account.id, "Unauthorized (401)");
+		throw error;
+	} finally {
+		const finishedAtMsFinal = finishedAtMs ?? Date.now();
+		await accountsManager.finalizeQuota(account, reservation);
+		const premiumRemainingAfter = account.premiumRemaining;
+		const premiumUnlimitedAfter = account.unlimited;
+		store.insert({
+			requestId: ctx.requestId,
+			startedAtMs: ctx.startedAtMs,
+			finishedAtMs: finishedAtMsFinal,
+			durationMs: finishedAtMsFinal - ctx.startedAtMs,
+			method: ctx.method,
+			path: ctx.path,
+			upstreamEndpoint: endpoint,
+			stream: false,
+			accountId: account.id,
+			accountType: account.accountType,
+			costUnits,
+			clientModel,
+			upstreamModel: selectedModel.id,
+			clientIp: ctx.clientIp,
+			clientIpSource: ctx.clientIpSource,
+			userAgent: ctx.userAgent,
+			...usage,
+			premiumRemainingBefore,
+			premiumRemainingAfter,
+			premiumRemainingDiff: computeDiff(premiumRemainingBefore, premiumRemainingAfter),
+			premiumUnlimitedBefore,
+			premiumUnlimitedAfter,
+			httpStatus,
+			errorName,
+			errorStatus,
+			errorMessage
+		});
+	}
+}
+
+//#endregion
+//#region src/lib/models.ts
+const findEndpointModel = (sdkModelId) => {
+	const models = state.models?.data ?? [];
+	const exactMatch = models.find((m) => m.id === sdkModelId);
+	if (exactMatch) return exactMatch;
+	const normalized = _normalizeSdkModelId(sdkModelId);
+	if (!normalized) return;
+	const modelName = `claude-${normalized.family}-${normalized.version}`;
+	const model = models.find((m) => m.id === modelName);
+	if (model) return model;
+};
+/**
+* Normalizes an SDK model ID to extract the model family and version.
+* this method from github copilot extension
+* Examples:
+* - "claude-opus-4-5-20251101" -> { family: "opus", version: "4.5" }
+* - "claude-3-5-sonnet-20241022" -> { family: "sonnet", version: "3.5" }
+* - "claude-sonnet-4-20250514" -> { family: "sonnet", version: "4" }
+* - "claude-haiku-3-5-20250514" -> { family: "haiku", version: "3.5" }
+* - "claude-haiku-4.5" -> { family: "haiku", version: "4.5" }
+*/
+const _normalizeSdkModelId = (sdkModelId) => {
+	const withoutDate = sdkModelId.toLowerCase().replace(/-\d{8}$/, "");
+	const pattern1 = withoutDate.match(/^claude-(\w+)-(\d+)-(\d+)$/);
+	if (pattern1) return {
+		family: pattern1[1],
+		version: `${pattern1[2]}.${pattern1[3]}`
+	};
+	const pattern2 = withoutDate.match(/^claude-(\d+)-(\d+)-(\w+)$/);
+	if (pattern2) return {
+		family: pattern2[3],
+		version: `${pattern2[1]}.${pattern2[2]}`
+	};
+	const pattern3 = withoutDate.match(/^claude-(\w+)-(\d+)\.(\d+)$/);
+	if (pattern3) return {
+		family: pattern3[1],
+		version: `${pattern3[2]}.${pattern3[3]}`
+	};
+	const pattern4 = withoutDate.match(/^claude-(\w+)-(\d+)$/);
+	if (pattern4) return {
+		family: pattern4[1],
+		version: pattern4[2]
+	};
+	const pattern5 = withoutDate.match(/^claude-(\d+)-(\w+)$/);
+	if (pattern5) return {
+		family: pattern5[2],
+		version: pattern5[1]
+	};
+};
+
+//#endregion
+//#region src/routes/messages/utils.ts
+function mapOpenAIStopReasonToAnthropic(finishReason) {
+	if (finishReason === null) return null;
+	return {
+		stop: "end_turn",
+		length: "max_tokens",
+		tool_calls: "tool_use",
+		content_filter: "end_turn"
+	}[finishReason];
+}
+const mergeContentWithText = (toolResult, textBlock) => {
+	if (typeof toolResult.content === "string") return {
+		...toolResult,
+		content: `${toolResult.content}\n\n${textBlock.text}`
+	};
+	return {
+		...toolResult,
+		content: [...toolResult.content, textBlock]
+	};
+};
+const mergeContentWithTexts = (toolResult, textBlocks) => {
+	if (typeof toolResult.content === "string") {
+		const appendedTexts = textBlocks.map((tb) => tb.text).join("\n\n");
+		return {
+			...toolResult,
+			content: `${toolResult.content}\n\n${appendedTexts}`
+		};
+	}
+	return {
+		...toolResult,
+		content: [...toolResult.content, ...textBlocks]
+	};
+};
+const mergeToolResult = (toolResults, textBlocks) => {
+	if (toolResults.length === textBlocks.length) return toolResults.map((toolResult, index) => mergeContentWithText(toolResult, textBlocks[index]));
+	const lastIndex = toolResults.length - 1;
+	return toolResults.map((toolResult, index) => index === lastIndex ? mergeContentWithTexts(toolResult, textBlocks) : toolResult);
+};
+const mergeToolResultForClaude = (anthropicPayload) => {
+	for (const msg of anthropicPayload.messages) {
+		if (msg.role !== "user" || !Array.isArray(msg.content)) continue;
+		const toolResults = [];
+		const textBlocks = [];
+		let valid = true;
+		for (const block of msg.content) if (block.type === "tool_result") toolResults.push(block);
+		else if (block.type === "text") textBlocks.push(block);
+		else {
+			valid = false;
+			break;
+		}
+		if (!valid || toolResults.length === 0 || textBlocks.length === 0) continue;
+		msg.content = mergeToolResult(toolResults, textBlocks);
+	}
+};
+const estimateInputTokens = async (payload, selectedModel, logger$7) => {
 	try {
-		return await handleCompletion$1(c);
+		return (await getTokenCount(payload, selectedModel)).input;
 	} catch (error) {
-		return await forwardError(c, error);
+		logger$7.warn("Failed to estimate input tokens for message_start", error);
+		return;
 	}
-});
-
-//#endregion
-//#region src/services/copilot/create-embeddings.ts
-const createEmbeddings = async (payload, account) => {
-	const ctx = account ?? accountFromState();
-	if (!ctx.copilotToken) throw new Error("Copilot token not found");
-	const response = await fetch(`${copilotBaseUrl(ctx)}/embeddings`, {
-		method: "POST",
-		headers: copilotHeaders(ctx),
-		body: JSON.stringify(payload)
+};
+const isWarmupProbeRequest = (payload) => {
+	const lastMsg = payload.messages.at(-1);
+	if (!lastMsg || lastMsg.role !== "user" || !Array.isArray(lastMsg.content)) return false;
+	const lastBlock = lastMsg.content.at(-1);
+	if (!lastBlock || lastBlock.type !== "text") return false;
+	const text = lastBlock.text.trim().toLowerCase();
+	if (!(lastBlock.cache_control?.type === "ephemeral")) return false;
+	if (text === "warmup") return true;
+	if (text === "hello") {
+		const preludeBlocks = lastMsg.content.slice(0, -1);
+		if (preludeBlocks.length === 0) return false;
+		return preludeBlocks.every((block) => block.type === "text" && block.text.trimStart().toLowerCase().startsWith("<system-reminder"));
+	}
+	return false;
+};
+const handleSelectionFailure = (context) => {
+	const { c, store, requestId, startedAtMs, method, path: path$2, streamRequested, clientModel, clientIp, clientIpSource, userAgent, userId, safetyIdentifier, promptCacheKey, initiator, selection } = context;
+	const finishedAtMs = Date.now();
+	store.insert({
+		requestId,
+		startedAtMs,
+		finishedAtMs,
+		durationMs: finishedAtMs - startedAtMs,
+		method,
+		path: path$2,
+		stream: streamRequested,
+		clientModel,
+		clientIp,
+		clientIpSource,
+		userAgent,
+		userId,
+		safetyIdentifier,
+		promptCacheKey,
+		initiator,
+		httpStatus: selection.reason === "MODEL_NOT_SUPPORTED" ? 400 : 429,
+		selectionFailureReason: selection.reason
+	});
+	if (selection.reason === "MODEL_NOT_SUPPORTED") return c.json({ error: {
+		message: `Model "${clientModel}" is not available for any configured account.`,
+		type: "invalid_request_error"
+	} }, 400);
+	return c.json({ error: {
+		message: "All accounts have exhausted their quota. Please wait for quota refresh or add additional accounts.",
+		type: "rate_limit_error"
+	} }, 429);
+};
+const maybeBlockOriginalModelName = (context) => {
+	if (!getAliasTargetSet().has(context.clientModel.toLowerCase())) return null;
+	return handleSelectionFailure({
+		...context,
+		selection: {
+			ok: false,
+			reason: "MODEL_NOT_SUPPORTED"
+		}
 	});
-	if (!response.ok) throw new HTTPError("Failed to create embeddings", response);
-	return await response.json();
 };
 
 //#endregion
-//#region src/routes/embeddings/route.ts
-const embeddingRoutes = new Hono();
-const EMBEDDINGS_ENDPOINT = "/embeddings";
-embeddingRoutes.post("/", async (c) => {
-	try {
-		const store = getRequestHistoryStore();
-		const requestId = randomUUID();
-		const startedAtMs = Date.now();
-		const method = c.req.raw.method;
-		const path$2 = new URL(c.req.url, "http://local").pathname;
-		const { ip: clientIp, source: clientIpSource } = getClientIpInfo(c);
-		const userAgent = c.req.header("user-agent") ?? void 0;
-		const ctx = {
-			requestId,
-			startedAtMs,
-			method,
-			path: path$2,
-			clientIp,
-			clientIpSource,
-			userAgent
-		};
-		const payload = await c.req.json();
-		const clientModel = payload.model;
-		if (getAliasTargetSet().has(clientModel.toLowerCase())) {
-			recordSelectionFailure$1(store, {
-				ctx,
-				clientModel,
-				reason: "MODEL_NOT_SUPPORTED"
-			});
-			return selectionFailureResponse$1(c, clientModel, "MODEL_NOT_SUPPORTED");
+//#region src/routes/messages/non-stream-translation.ts
+const THINKING_TEXT = "Thinking...";
+function translateToOpenAI(payload) {
+	const modelId = payload.model;
+	const model = state.models?.data.find((m) => m.id === modelId);
+	const thinkingBudget = getThinkingBudget(payload, model);
+	return {
+		model: modelId,
+		messages: translateAnthropicMessagesToOpenAI(payload, modelId, thinkingBudget),
+		max_tokens: payload.max_tokens,
+		stop: payload.stop_sequences,
+		stream: payload.stream,
+		temperature: payload.temperature,
+		top_p: payload.top_p,
+		user: payload.metadata?.user_id,
+		tools: translateAnthropicToolsToOpenAI(payload.tools),
+		tool_choice: translateAnthropicToolChoiceToOpenAI(payload.tool_choice),
+		thinking_budget: thinkingBudget
+	};
+}
+function getThinkingBudget(payload, model) {
+	const thinking = payload.thinking;
+	if (model && thinking) {
+		const maxThinkingBudget = Math.min(model.capabilities.supports.max_thinking_budget ?? 0, (model.capabilities.limits.max_output_tokens ?? 0) - 1);
+		thinking.budget_tokens ??= maxThinkingBudget;
+		if (maxThinkingBudget > 0) {
+			const budgetTokens = Math.min(thinking.budget_tokens, maxThinkingBudget);
+			return Math.max(budgetTokens, model.capabilities.supports.min_thinking_budget ?? 1024);
 		}
-		const selection = await accountsManager.selectAccountForRequest([{
-			modelId: clientModel,
-			endpoint: EMBEDDINGS_ENDPOINT
-		}]);
-		if (!selection.ok) {
-			recordSelectionFailure$1(store, {
-				ctx,
-				clientModel,
-				reason: selection.reason
+	}
+}
+function translateAnthropicMessagesToOpenAI(payload, modelId, _thinkingBudget) {
+	const systemMessages = handleSystemPrompt(payload.system);
+	const otherMessages = payload.messages.flatMap((message) => message.role === "user" ? handleUserMessage(message) : handleAssistantMessage(message, modelId));
+	return [...systemMessages, ...otherMessages];
+}
+function handleSystemPrompt(system) {
+	if (!system) return [];
+	if (typeof system === "string") return [{
+		role: "system",
+		content: system
+	}];
+	else return [{
+		role: "system",
+		content: system.map((block) => {
+			return block.text;
+		}).join("\n\n")
+	}];
+}
+function handleUserMessage(message) {
+	const newMessages = [];
+	if (Array.isArray(message.content)) {
+		const toolResultBlocks = message.content.filter((block) => block.type === "tool_result");
+		const otherBlocks = message.content.filter((block) => block.type !== "tool_result");
+		for (const block of toolResultBlocks) newMessages.push({
+			role: "tool",
+			tool_call_id: block.tool_use_id,
+			content: mapContent(block.content)
+		});
+		if (otherBlocks.length > 0) newMessages.push({
+			role: "user",
+			content: mapContent(otherBlocks)
+		});
+	} else newMessages.push({
+		role: "user",
+		content: mapContent(message.content)
+	});
+	return newMessages;
+}
+function handleAssistantMessage(message, modelId) {
+	if (!Array.isArray(message.content)) return [{
+		role: "assistant",
+		content: mapContent(message.content)
+	}];
+	const toolUseBlocks = message.content.filter((block) => block.type === "tool_use");
+	let thinkingBlocks = message.content.filter((block) => block.type === "thinking");
+	if (modelId.startsWith("claude")) thinkingBlocks = thinkingBlocks.filter((b) => b.thinking && b.thinking !== THINKING_TEXT && b.signature && !b.signature.includes("@"));
+	const thinkingContents = thinkingBlocks.filter((b) => b.thinking && b.thinking !== THINKING_TEXT).map((b) => b.thinking);
+	const allThinkingContent = thinkingContents.length > 0 ? thinkingContents.join("\n\n") : void 0;
+	const signature = thinkingBlocks.find((b) => b.signature)?.signature;
+	return toolUseBlocks.length > 0 ? [{
+		role: "assistant",
+		content: mapContent(message.content),
+		reasoning_text: allThinkingContent,
+		reasoning_opaque: signature,
+		tool_calls: toolUseBlocks.map((toolUse) => ({
+			id: toolUse.id,
+			type: "function",
+			function: {
+				name: toolUse.name,
+				arguments: JSON.stringify(toolUse.input)
+			}
+		}))
+	}] : [{
+		role: "assistant",
+		content: mapContent(message.content),
+		reasoning_text: allThinkingContent,
+		reasoning_opaque: signature
+	}];
+}
+function mapContent(content) {
+	if (typeof content === "string") return content;
+	if (!Array.isArray(content)) return null;
+	if (!content.some((block) => block.type === "image")) return content.filter((block) => block.type === "text").map((block) => block.text).join("\n\n");
+	const contentParts = [];
+	for (const block of content) switch (block.type) {
+		case "text":
+			contentParts.push({
+				type: "text",
+				text: block.text
 			});
-			return selectionFailureResponse$1(c, clientModel, selection.reason);
-		}
-		const upstreamPayload = {
-			...payload,
-			model: selection.selectedModel.id
-		};
-		return await runEmbeddingsWithAccount({
-			c,
-			store,
-			ctx,
-			payload: upstreamPayload,
-			clientModel,
-			selection
-		});
-	} catch (error) {
-		return await forwardError(c, error);
+			break;
+		case "image":
+			contentParts.push({
+				type: "image_url",
+				image_url: { url: `data:${block.source.media_type};base64,${block.source.data}` }
+			});
+			break;
 	}
-});
-function recordSelectionFailure$1(store, params) {
-	const { ctx, clientModel, reason } = params;
-	const finishedAtMs = Date.now();
-	store.insert({
-		requestId: ctx.requestId,
-		startedAtMs: ctx.startedAtMs,
-		finishedAtMs,
-		durationMs: finishedAtMs - ctx.startedAtMs,
-		method: ctx.method,
-		path: ctx.path,
-		upstreamEndpoint: EMBEDDINGS_ENDPOINT,
-		stream: false,
-		clientModel,
-		clientIp: ctx.clientIp,
-		clientIpSource: ctx.clientIpSource,
-		userAgent: ctx.userAgent,
-		httpStatus: reason === "MODEL_NOT_SUPPORTED" ? 400 : 429,
-		selectionFailureReason: reason
-	});
+	return contentParts;
 }
-function selectionFailureResponse$1(c, clientModel, reason) {
-	if (reason === "MODEL_NOT_SUPPORTED") return c.json({ error: {
-		message: `Model "${clientModel}" is not available for any configured account.`,
-		type: "invalid_request_error"
-	} }, 400);
-	return c.json({ error: {
-		message: "All accounts have exhausted their quota. Please wait for quota refresh or add additional accounts.",
-		type: "rate_limit_error"
-	} }, 429);
+function translateAnthropicToolsToOpenAI(anthropicTools) {
+	if (!anthropicTools) return;
+	return anthropicTools.map((tool) => ({
+		type: "function",
+		function: {
+			name: tool.name,
+			description: tool.description,
+			parameters: normalizeToolSchema(tool.input_schema)
+		}
+	}));
 }
-async function runEmbeddingsWithAccount({ c, store, ctx, payload, clientModel, selection }) {
-	const { account, reservation, selectedModel, endpoint, costUnits } = selection;
-	const premiumRemainingBefore = account.premiumRemaining;
-	const premiumUnlimitedBefore = account.unlimited;
-	let httpStatus = 200;
-	let usage = {};
-	let errorName;
-	let errorStatus;
-	let errorMessage;
-	let finishedAtMs;
-	try {
-		const accountCtx = toAccountContext(account);
-		const response = await createEmbeddings(payload, accountCtx);
-		usage = normalizeEmbeddingsUsage(response.usage);
-		finishedAtMs = Date.now();
-		return c.json(response);
-	} catch (error) {
-		finishedAtMs = Date.now();
-		const details = extractErrorDetails(error);
-		httpStatus = details.httpStatus;
-		errorName = details.errorName;
-		errorStatus = details.errorStatus;
-		errorMessage = details.errorMessage;
-		if (details.unauthorized) accountsManager.markAccountFailed(account.id, "Unauthorized (401)");
-		throw error;
-	} finally {
-		const finishedAtMsFinal = finishedAtMs ?? Date.now();
-		await accountsManager.finalizeQuota(account, reservation);
-		const premiumRemainingAfter = account.premiumRemaining;
-		const premiumUnlimitedAfter = account.unlimited;
-		store.insert({
-			requestId: ctx.requestId,
-			startedAtMs: ctx.startedAtMs,
-			finishedAtMs: finishedAtMsFinal,
-			durationMs: finishedAtMsFinal - ctx.startedAtMs,
-			method: ctx.method,
-			path: ctx.path,
-			upstreamEndpoint: endpoint,
-			stream: false,
-			accountId: account.id,
-			accountType: account.accountType,
-			costUnits,
-			clientModel,
-			upstreamModel: selectedModel.id,
-			clientIp: ctx.clientIp,
-			clientIpSource: ctx.clientIpSource,
-			userAgent: ctx.userAgent,
-			...usage,
-			premiumRemainingBefore,
-			premiumRemainingAfter,
-			premiumRemainingDiff: computeDiff(premiumRemainingBefore, premiumRemainingAfter),
-			premiumUnlimitedBefore,
-			premiumUnlimitedAfter,
-			httpStatus,
-			errorName,
-			errorStatus,
-			errorMessage
-		});
+/**
+* Ensures `type: "object"` schema has a `properties` field.
+* OpenAI's API rejects object schemas without it.
+*/
+const normalizeToolSchema = (schema) => {
+	if (schema.type === "object" && !schema.properties) return {
+		...schema,
+		properties: {}
+	};
+	return schema;
+};
+function translateAnthropicToolChoiceToOpenAI(anthropicToolChoice) {
+	if (!anthropicToolChoice) return;
+	switch (anthropicToolChoice.type) {
+		case "auto": return "auto";
+		case "any": return "required";
+		case "tool":
+			if (anthropicToolChoice.name) return {
+				type: "function",
+				function: { name: anthropicToolChoice.name }
+			};
+			return;
+		case "none": return "none";
+		default: return;
+	}
+}
+function translateToAnthropic(response) {
+	const assistantContentBlocks = [];
+	let stopReason = response.choices[0]?.finish_reason ?? null;
+	for (const choice of response.choices) {
+		const textBlocks = getAnthropicTextBlocks(choice.message.content);
+		const thinkBlocks = getAnthropicThinkBlocks(choice.message.reasoning_text, choice.message.reasoning_opaque);
+		const toolUseBlocks = getAnthropicToolUseBlocks(choice.message.tool_calls);
+		assistantContentBlocks.push(...thinkBlocks, ...textBlocks, ...toolUseBlocks);
+		if (choice.finish_reason === "tool_calls" || stopReason === "stop") stopReason = choice.finish_reason;
 	}
+	return {
+		id: response.id,
+		type: "message",
+		role: "assistant",
+		model: response.model,
+		content: assistantContentBlocks,
+		stop_reason: mapOpenAIStopReasonToAnthropic(stopReason),
+		stop_sequence: null,
+		usage: {
+			input_tokens: (response.usage?.prompt_tokens ?? 0) - (response.usage?.prompt_tokens_details?.cached_tokens ?? 0),
+			output_tokens: response.usage?.completion_tokens ?? 0,
+			...response.usage?.prompt_tokens_details?.cached_tokens !== void 0 && { cache_read_input_tokens: response.usage.prompt_tokens_details.cached_tokens }
+		}
+	};
+}
+function getAnthropicTextBlocks(messageContent) {
+	if (typeof messageContent === "string" && messageContent.length > 0) return [{
+		type: "text",
+		text: messageContent
+	}];
+	if (Array.isArray(messageContent)) return messageContent.filter((part) => part.type === "text").map((part) => ({
+		type: "text",
+		text: part.text
+	}));
+	return [];
+}
+function getAnthropicThinkBlocks(reasoningText, reasoningOpaque) {
+	if (reasoningText && reasoningText.length > 0) return [{
+		type: "thinking",
+		thinking: reasoningText,
+		signature: reasoningOpaque || ""
+	}];
+	if (reasoningOpaque && reasoningOpaque.length > 0) return [{
+		type: "thinking",
+		thinking: THINKING_TEXT,
+		signature: reasoningOpaque
+	}];
+	return [];
+}
+function getAnthropicToolUseBlocks(toolCalls) {
+	if (!toolCalls) return [];
+	return toolCalls.map((toolCall) => ({
+		type: "tool_use",
+		id: toolCall.id,
+		name: toolCall.function.name,
+		input: JSON.parse(toolCall.function.arguments)
+	}));
 }
 
 //#endregion
-//#region src/lib/models.ts
-const findEndpointModel = (sdkModelId) => {
-	const models = state.models?.data ?? [];
-	const exactMatch = models.find((m) => m.id === sdkModelId);
-	if (exactMatch) return exactMatch;
-	const normalized = _normalizeSdkModelId(sdkModelId);
-	if (!normalized) return;
-	const modelName = `claude-${normalized.family}-${normalized.version}`;
-	const model = models.find((m) => m.id === modelName);
-	if (model) return model;
-};
+//#region src/routes/messages/count-tokens-handler.ts
 /**
-* Normalizes an SDK model ID to extract the model family and version.
-* this method from github copilot extension
-* Examples:
-* - "claude-opus-4-5-20251101" -> { family: "opus", version: "4.5" }
-* - "claude-3-5-sonnet-20241022" -> { family: "sonnet", version: "3.5" }
-* - "claude-sonnet-4-20250514" -> { family: "sonnet", version: "4" }
-* - "claude-haiku-3-5-20250514" -> { family: "haiku", version: "3.5" }
-* - "claude-haiku-4.5" -> { family: "haiku", version: "4.5" }
+* Handles token counting for Anthropic messages
 */
-const _normalizeSdkModelId = (sdkModelId) => {
-	const withoutDate = sdkModelId.toLowerCase().replace(/-\d{8}$/, "");
-	const pattern1 = withoutDate.match(/^claude-(\w+)-(\d+)-(\d+)$/);
-	if (pattern1) return {
-		family: pattern1[1],
-		version: `${pattern1[2]}.${pattern1[3]}`
-	};
-	const pattern2 = withoutDate.match(/^claude-(\d+)-(\d+)-(\w+)$/);
-	if (pattern2) return {
-		family: pattern2[3],
-		version: `${pattern2[1]}.${pattern2[2]}`
-	};
-	const pattern3 = withoutDate.match(/^claude-(\w+)-(\d+)\.(\d+)$/);
-	if (pattern3) return {
-		family: pattern3[1],
-		version: `${pattern3[2]}.${pattern3[3]}`
-	};
-	const pattern4 = withoutDate.match(/^claude-(\w+)-(\d+)$/);
-	if (pattern4) return {
-		family: pattern4[1],
-		version: pattern4[2]
-	};
-	const pattern5 = withoutDate.match(/^claude-(\d+)-(\w+)$/);
-	if (pattern5) return {
-		family: pattern5[2],
-		version: pattern5[1]
+async function handleCountTokens(c) {
+	try {
+		const anthropicBeta = c.req.header("anthropic-beta");
+		const anthropicPayload = await c.req.json();
+		const openAIPayload = translateToOpenAI(anthropicPayload);
+		const selectedModel = findEndpointModel(anthropicPayload.model);
+		anthropicPayload.model = selectedModel?.id ?? anthropicPayload.model;
+		if (!selectedModel) {
+			consola.warn("Model not found, returning default token count");
+			return c.json({ input_tokens: 1 });
+		}
+		const tokenCount = await getTokenCount(openAIPayload, selectedModel);
+		if (anthropicPayload.tools && anthropicPayload.tools.length > 0) {
+			let addToolSystemPromptCount = false;
+			if (anthropicBeta) {
+				const toolsLength = anthropicPayload.tools.length;
+				addToolSystemPromptCount = !anthropicPayload.tools.some((tool) => tool.name.startsWith("mcp__") || tool.name === "Skill" && toolsLength === 1);
+			}
+			if (addToolSystemPromptCount) {
+				if (anthropicPayload.model.startsWith("claude")) tokenCount.input = tokenCount.input + 346;
+				else if (anthropicPayload.model.startsWith("grok")) tokenCount.input = tokenCount.input + 120;
+			}
+		}
+		let finalTokenCount = tokenCount.input + tokenCount.output;
+		if (anthropicPayload.model.startsWith("claude")) finalTokenCount = Math.round(finalTokenCount * 1.15);
+		consola.info("Token count:", finalTokenCount);
+		return c.json({ input_tokens: finalTokenCount });
+	} catch (error) {
+		consola.error("Error counting tokens:", error);
+		return c.json({ input_tokens: 1 });
+	}
+}
+
+//#endregion
+//#region src/services/copilot/create-responses.ts
+const createResponses = async (payload, { vision, initiator, upstreamRequestId, subagentMarker, sessionId, isCompact }, account) => {
+	const ctx = account ?? accountFromState();
+	if (!ctx.copilotToken) throw new Error("Copilot token not found");
+	const headers = {
+		...copilotHeaders(ctx, vision, upstreamRequestId),
+		"x-initiator": initiator
 	};
+	prepareInteractionHeaders(sessionId, Boolean(subagentMarker), headers);
+	prepareForCompact(headers, isCompact);
+	payload.service_tier = null;
+	const response = await fetch(`${copilotBaseUrl(ctx)}/responses`, {
+		method: "POST",
+		headers,
+		body: JSON.stringify(payload)
+	});
+	if (!response.ok) {
+		consola.error("Failed to create responses", response);
+		throw new HTTPError("Failed to create responses", response);
+	}
+	if (payload.stream) return events(response);
+	return await response.json();
 };
 
 //#endregion
-//#region src/routes/messages/utils.ts
-function mapOpenAIStopReasonToAnthropic(finishReason) {
-	if (finishReason === null) return null;
-	return {
-		stop: "end_turn",
-		length: "max_tokens",
-		tool_calls: "tool_use",
-		content_filter: "end_turn"
-	}[finishReason];
-}
-const mergeContentWithText = (toolResult, textBlock) => {
-	if (typeof toolResult.content === "string") return {
-		...toolResult,
-		content: `${toolResult.content}\n\n${textBlock.text}`
-	};
+//#region src/routes/messages/responses-translation.ts
+const MESSAGE_TYPE = "message";
+const COMPACTION_SIGNATURE_PREFIX = "cm1#";
+const COMPACTION_SIGNATURE_SEPARATOR = "@";
+const THINKING_TEXT$1 = "Thinking...";
+const translateAnthropicMessagesToResponsesPayload = (payload, modelOverride) => {
+	const model = modelOverride ?? payload.model;
+	const input = [];
+	const applyPhase = shouldApplyPhase(payload.model);
+	for (const message of payload.messages) input.push(...translateMessage(message, payload.model, applyPhase));
+	const translatedTools = convertAnthropicTools(payload.tools);
+	const toolChoice = convertAnthropicToolChoice(payload.tool_choice);
+	const { safetyIdentifier, sessionId: promptCacheKey } = parseUserIdMetadata(payload.metadata?.user_id);
 	return {
-		...toolResult,
-		content: [...toolResult.content, textBlock]
+		model,
+		input,
+		instructions: translateSystemPrompt(payload.system, model),
+		temperature: 1,
+		top_p: payload.top_p ?? null,
+		max_output_tokens: Math.max(payload.max_tokens, 12800),
+		tools: translatedTools,
+		tool_choice: toolChoice,
+		metadata: payload.metadata ? { ...payload.metadata } : null,
+		safety_identifier: safetyIdentifier,
+		prompt_cache_key: promptCacheKey,
+		stream: payload.stream ?? null,
+		store: false,
+		parallel_tool_calls: true,
+		reasoning: {
+			effort: getReasoningEffortForModel(model),
+			summary: "auto"
+		},
+		include: ["reasoning.encrypted_content"]
 	};
 };
-const mergeContentWithTexts = (toolResult, textBlocks) => {
-	if (typeof toolResult.content === "string") {
-		const appendedTexts = textBlocks.map((tb) => tb.text).join("\n\n");
+const encodeCompactionCarrierSignature = (compaction) => {
+	return `${COMPACTION_SIGNATURE_PREFIX}${compaction.encrypted_content}${COMPACTION_SIGNATURE_SEPARATOR}${compaction.id}`;
+};
+const decodeCompactionCarrierSignature = (signature) => {
+	if (signature.startsWith(COMPACTION_SIGNATURE_PREFIX)) {
+		const raw = signature.slice(4);
+		const separatorIndex = raw.indexOf(COMPACTION_SIGNATURE_SEPARATOR);
+		if (separatorIndex <= 0 || separatorIndex === raw.length - 1) return;
+		const encrypted_content = raw.slice(0, separatorIndex);
+		const id = raw.slice(separatorIndex + 1);
+		if (!encrypted_content) return;
 		return {
-			...toolResult,
-			content: `${toolResult.content}\n\n${appendedTexts}`
+			id,
+			encrypted_content
 		};
 	}
-	return {
-		...toolResult,
-		content: [...toolResult.content, ...textBlocks]
-	};
 };
-const mergeToolResult = (toolResults, textBlocks) => {
-	if (toolResults.length === textBlocks.length) return toolResults.map((toolResult, index) => mergeContentWithText(toolResult, textBlocks[index]));
-	const lastIndex = toolResults.length - 1;
-	return toolResults.map((toolResult, index) => index === lastIndex ? mergeContentWithTexts(toolResult, textBlocks) : toolResult);
+const translateMessage = (message, model, applyPhase) => {
+	if (message.role === "user") return translateUserMessage(message);
+	return translateAssistantMessage(message, model, applyPhase);
 };
-const mergeToolResultForClaude = (anthropicPayload) => {
-	for (const msg of anthropicPayload.messages) {
-		if (msg.role !== "user" || !Array.isArray(msg.content)) continue;
-		const toolResults = [];
-		const textBlocks = [];
-		let valid = true;
-		for (const block of msg.content) if (block.type === "tool_result") toolResults.push(block);
-		else if (block.type === "text") textBlocks.push(block);
-		else {
-			valid = false;
-			break;
+const translateUserMessage = (message) => {
+	if (typeof message.content === "string") return [createMessage("user", message.content)];
+	if (!Array.isArray(message.content)) return [];
+	const items = [];
+	const pendingContent = [];
+	for (const block of message.content) {
+		if (block.type === "tool_result") {
+			flushPendingContent(pendingContent, items, { role: "user" });
+			items.push(createFunctionCallOutput(block));
+			continue;
 		}
-		if (!valid || toolResults.length === 0 || textBlocks.length === 0) continue;
-		msg.content = mergeToolResult(toolResults, textBlocks);
-	}
-};
-const estimateInputTokens = async (payload, selectedModel, logger$7) => {
-	try {
-		return (await getTokenCount(payload, selectedModel)).input;
-	} catch (error) {
-		logger$7.warn("Failed to estimate input tokens for message_start", error);
-		return;
-	}
-};
-const isWarmupProbeRequest = (payload) => {
-	const lastMsg = payload.messages.at(-1);
-	if (!lastMsg || lastMsg.role !== "user" || !Array.isArray(lastMsg.content)) return false;
-	const lastBlock = lastMsg.content.at(-1);
-	if (!lastBlock || lastBlock.type !== "text") return false;
-	const text = lastBlock.text.trim().toLowerCase();
-	if (!(lastBlock.cache_control?.type === "ephemeral")) return false;
-	if (text === "warmup") return true;
-	if (text === "hello") {
-		const preludeBlocks = lastMsg.content.slice(0, -1);
-		if (preludeBlocks.length === 0) return false;
-		return preludeBlocks.every((block) => block.type === "text" && block.text.trimStart().toLowerCase().startsWith("<system-reminder"));
+		const converted = translateUserContentBlock(block);
+		if (converted) pendingContent.push(converted);
 	}
-	return false;
-};
-const handleSelectionFailure = (context) => {
-	const { c, store, requestId, startedAtMs, method, path: path$2, streamRequested, clientModel, clientIp, clientIpSource, userAgent, userId, safetyIdentifier, promptCacheKey, initiator, selection } = context;
-	const finishedAtMs = Date.now();
-	store.insert({
-		requestId,
-		startedAtMs,
-		finishedAtMs,
-		durationMs: finishedAtMs - startedAtMs,
-		method,
-		path: path$2,
-		stream: streamRequested,
-		clientModel,
-		clientIp,
-		clientIpSource,
-		userAgent,
-		userId,
-		safetyIdentifier,
-		promptCacheKey,
-		initiator,
-		httpStatus: selection.reason === "MODEL_NOT_SUPPORTED" ? 400 : 429,
-		selectionFailureReason: selection.reason
-	});
-	if (selection.reason === "MODEL_NOT_SUPPORTED") return c.json({ error: {
-		message: `Model "${clientModel}" is not available for any configured account.`,
-		type: "invalid_request_error"
-	} }, 400);
-	return c.json({ error: {
-		message: "All accounts have exhausted their quota. Please wait for quota refresh or add additional accounts.",
-		type: "rate_limit_error"
-	} }, 429);
+	flushPendingContent(pendingContent, items, { role: "user" });
+	return items;
 };
-const maybeBlockOriginalModelName = (context) => {
-	if (!getAliasTargetSet().has(context.clientModel.toLowerCase())) return null;
-	return handleSelectionFailure({
-		...context,
-		selection: {
-			ok: false,
-			reason: "MODEL_NOT_SUPPORTED"
+const translateAssistantMessage = (message, model, applyPhase) => {
+	const assistantPhase = resolveAssistantPhase(model, message.content, applyPhase);
+	if (typeof message.content === "string") return [createMessage("assistant", message.content, assistantPhase)];
+	if (!Array.isArray(message.content)) return [];
+	const items = [];
+	const pendingContent = [];
+	for (const block of message.content) {
+		if (block.type === "tool_use") {
+			flushPendingContent(pendingContent, items, {
+				role: "assistant",
+				phase: assistantPhase
+			});
+			items.push(createFunctionToolCall(block));
+			continue;
 		}
-	});
-};
-
-//#endregion
-//#region src/routes/messages/non-stream-translation.ts
-const THINKING_TEXT = "Thinking...";
-function translateToOpenAI(payload) {
-	const modelId = payload.model;
-	const model = state.models?.data.find((m) => m.id === modelId);
-	const thinkingBudget = getThinkingBudget(payload, model);
-	return {
-		model: modelId,
-		messages: translateAnthropicMessagesToOpenAI(payload, modelId, thinkingBudget),
-		max_tokens: payload.max_tokens,
-		stop: payload.stop_sequences,
-		stream: payload.stream,
-		temperature: payload.temperature,
-		top_p: payload.top_p,
-		user: payload.metadata?.user_id,
-		tools: translateAnthropicToolsToOpenAI(payload.tools),
-		tool_choice: translateAnthropicToolChoiceToOpenAI(payload.tool_choice),
-		thinking_budget: thinkingBudget
-	};
-}
-function getThinkingBudget(payload, model) {
-	const thinking = payload.thinking;
-	if (model && thinking) {
-		const maxThinkingBudget = Math.min(model.capabilities.supports.max_thinking_budget ?? 0, (model.capabilities.limits.max_output_tokens ?? 0) - 1);
-		thinking.budget_tokens ??= maxThinkingBudget;
-		if (maxThinkingBudget > 0) {
-			const budgetTokens = Math.min(thinking.budget_tokens, maxThinkingBudget);
-			return Math.max(budgetTokens, model.capabilities.supports.min_thinking_budget ?? 1024);
+		if (block.type === "thinking" && block.signature) {
+			const compactionContent = createCompactionContent(block);
+			if (compactionContent) {
+				flushPendingContent(pendingContent, items, {
+					role: "assistant",
+					phase: assistantPhase
+				});
+				items.push(compactionContent);
+				continue;
+			}
+			if (block.signature.includes("@")) {
+				flushPendingContent(pendingContent, items, {
+					role: "assistant",
+					phase: assistantPhase
+				});
+				items.push(createReasoningContent(block));
+				continue;
+			}
 		}
+		const converted = translateAssistantContentBlock(block);
+		if (converted) pendingContent.push(converted);
 	}
-}
-function translateAnthropicMessagesToOpenAI(payload, modelId, _thinkingBudget) {
-	const systemMessages = handleSystemPrompt(payload.system);
-	const otherMessages = payload.messages.flatMap((message) => message.role === "user" ? handleUserMessage(message) : handleAssistantMessage(message, modelId));
-	return [...systemMessages, ...otherMessages];
-}
-function handleSystemPrompt(system) {
-	if (!system) return [];
-	if (typeof system === "string") return [{
-		role: "system",
-		content: system
-	}];
-	else return [{
-		role: "system",
-		content: system.map((block) => {
-			return block.text;
-		}).join("\n\n")
-	}];
-}
-function handleUserMessage(message) {
-	const newMessages = [];
-	if (Array.isArray(message.content)) {
-		const toolResultBlocks = message.content.filter((block) => block.type === "tool_result");
-		const otherBlocks = message.content.filter((block) => block.type !== "tool_result");
-		for (const block of toolResultBlocks) newMessages.push({
-			role: "tool",
-			tool_call_id: block.tool_use_id,
-			content: mapContent(block.content)
-		});
-		if (otherBlocks.length > 0) newMessages.push({
-			role: "user",
-			content: mapContent(otherBlocks)
-		});
-	} else newMessages.push({
-		role: "user",
-		content: mapContent(message.content)
-	});
-	return newMessages;
-}
-function handleAssistantMessage(message, modelId) {
-	if (!Array.isArray(message.content)) return [{
-		role: "assistant",
-		content: mapContent(message.content)
-	}];
-	const toolUseBlocks = message.content.filter((block) => block.type === "tool_use");
-	let thinkingBlocks = message.content.filter((block) => block.type === "thinking");
-	if (modelId.startsWith("claude")) thinkingBlocks = thinkingBlocks.filter((b) => b.thinking && b.thinking !== THINKING_TEXT && b.signature && !b.signature.includes("@"));
-	const thinkingContents = thinkingBlocks.filter((b) => b.thinking && b.thinking !== THINKING_TEXT).map((b) => b.thinking);
-	const allThinkingContent = thinkingContents.length > 0 ? thinkingContents.join("\n\n") : void 0;
-	const signature = thinkingBlocks.find((b) => b.signature)?.signature;
-	return toolUseBlocks.length > 0 ? [{
-		role: "assistant",
-		content: mapContent(message.content),
-		reasoning_text: allThinkingContent,
-		reasoning_opaque: signature,
-		tool_calls: toolUseBlocks.map((toolUse) => ({
-			id: toolUse.id,
-			type: "function",
-			function: {
-				name: toolUse.name,
-				arguments: JSON.stringify(toolUse.input)
-			}
-		}))
-	}] : [{
+	flushPendingContent(pendingContent, items, {
 		role: "assistant",
-		content: mapContent(message.content),
-		reasoning_text: allThinkingContent,
-		reasoning_opaque: signature
-	}];
-}
-function mapContent(content) {
-	if (typeof content === "string") return content;
-	if (!Array.isArray(content)) return null;
-	if (!content.some((block) => block.type === "image")) return content.filter((block) => block.type === "text").map((block) => block.text).join("\n\n");
-	const contentParts = [];
-	for (const block of content) switch (block.type) {
-		case "text":
-			contentParts.push({
-				type: "text",
-				text: block.text
-			});
-			break;
-		case "image":
-			contentParts.push({
-				type: "image_url",
-				image_url: { url: `data:${block.source.media_type};base64,${block.source.data}` }
-			});
-			break;
+		phase: assistantPhase
+	});
+	return items;
+};
+const translateUserContentBlock = (block) => {
+	switch (block.type) {
+		case "text": return createTextContent(block.text);
+		case "image": return createImageContent(block);
+		default: return;
 	}
-	return contentParts;
-}
-function translateAnthropicToolsToOpenAI(anthropicTools) {
-	if (!anthropicTools) return;
-	return anthropicTools.map((tool) => ({
+};
+const translateAssistantContentBlock = (block) => {
+	switch (block.type) {
+		case "text": return createOutPutTextContent(block.text);
+		default: return;
+	}
+};
+const flushPendingContent = (pendingContent, target, message) => {
+	if (pendingContent.length === 0) return;
+	const messageContent = [...pendingContent];
+	target.push(createMessage(message.role, messageContent, message.phase));
+	pendingContent.length = 0;
+};
+const createMessage = (role, content, phase) => ({
+	type: MESSAGE_TYPE,
+	role,
+	content,
+	...role === "assistant" && phase ? { phase } : {}
+});
+const resolveAssistantPhase = (_model, content, applyPhase) => {
+	if (!applyPhase) return;
+	if (typeof content === "string") return "final_answer";
+	if (!Array.isArray(content)) return;
+	if (!content.some((block) => block.type === "text")) return;
+	return content.some((block) => block.type === "tool_use") ? "commentary" : "final_answer";
+};
+const shouldApplyPhase = (model) => {
+	return getExtraPromptForModel(model).includes("## Intermediary updates");
+};
+const createTextContent = (text) => ({
+	type: "input_text",
+	text
+});
+const createOutPutTextContent = (text) => ({
+	type: "output_text",
+	text
+});
+const createImageContent = (block) => ({
+	type: "input_image",
+	image_url: `data:${block.source.media_type};base64,${block.source.data}`,
+	detail: "auto"
+});
+const createReasoningContent = (block) => {
+	const { encryptedContent, id } = parseReasoningSignature(block.signature);
+	const thinking = block.thinking === THINKING_TEXT$1 ? "" : block.thinking;
+	return {
+		id,
+		type: "reasoning",
+		summary: thinking ? [{
+			type: "summary_text",
+			text: thinking
+		}] : [],
+		encrypted_content: encryptedContent
+	};
+};
+const createCompactionContent = (block) => {
+	const compaction = decodeCompactionCarrierSignature(block.signature);
+	if (!compaction) return;
+	return {
+		id: compaction.id,
+		type: "compaction",
+		encrypted_content: compaction.encrypted_content
+	};
+};
+const parseReasoningSignature = (signature) => {
+	const splitIndex = signature.lastIndexOf("@");
+	if (splitIndex <= 0 || splitIndex === signature.length - 1) return {
+		encryptedContent: signature,
+		id: ""
+	};
+	return {
+		encryptedContent: signature.slice(0, splitIndex),
+		id: signature.slice(splitIndex + 1)
+	};
+};
+const createFunctionToolCall = (block) => ({
+	type: "function_call",
+	call_id: block.id,
+	name: block.name,
+	arguments: JSON.stringify(block.input),
+	status: "completed"
+});
+const createFunctionCallOutput = (block) => ({
+	type: "function_call_output",
+	call_id: block.tool_use_id,
+	output: convertToolResultContent(block.content),
+	status: block.is_error ? "incomplete" : "completed"
+});
+const translateSystemPrompt = (system, model) => {
+	if (!system) return null;
+	const extraPrompt = getExtraPromptForModel(model);
+	if (typeof system === "string") return system + extraPrompt;
+	const text = system.map((block, index) => {
+		if (index === 0) return block.text + extraPrompt;
+		return block.text;
+	}).join(" ");
+	return text.length > 0 ? text : null;
+};
+const convertAnthropicTools = (tools) => {
+	if (!tools || tools.length === 0) return null;
+	return tools.map((tool) => ({
 		type: "function",
-		function: {
-			name: tool.name,
-			description: tool.description,
-			parameters: tool.input_schema
-		}
+		name: tool.name,
+		parameters: normalizeToolSchema(tool.input_schema),
+		strict: false,
+		...tool.description ? { description: tool.description } : {}
 	}));
-}
-function translateAnthropicToolChoiceToOpenAI(anthropicToolChoice) {
-	if (!anthropicToolChoice) return;
-	switch (anthropicToolChoice.type) {
+};
+const convertAnthropicToolChoice = (choice) => {
+	if (!choice) return "auto";
+	switch (choice.type) {
 		case "auto": return "auto";
 		case "any": return "required";
-		case "tool":
-			if (anthropicToolChoice.name) return {
-				type: "function",
-				function: { name: anthropicToolChoice.name }
-			};
-			return;
-		case "none": return "none";
-		default: return;
-	}
-}
-function translateToAnthropic(response) {
-	const assistantContentBlocks = [];
-	let stopReason = response.choices[0]?.finish_reason ?? null;
-	for (const choice of response.choices) {
-		const textBlocks = getAnthropicTextBlocks(choice.message.content);
-		const thinkBlocks = getAnthropicThinkBlocks(choice.message.reasoning_text, choice.message.reasoning_opaque);
-		const toolUseBlocks = getAnthropicToolUseBlocks(choice.message.tool_calls);
-		assistantContentBlocks.push(...thinkBlocks, ...textBlocks, ...toolUseBlocks);
-		if (choice.finish_reason === "tool_calls" || stopReason === "stop") stopReason = choice.finish_reason;
+		case "tool": return choice.name ? {
+			type: "function",
+			name: choice.name
+		} : "auto";
+		case "none": return "none";
+		default: return "auto";
 	}
+};
+const translateResponsesResultToAnthropic = (response) => {
+	const contentBlocks = mapOutputToAnthropicContent(response.output);
+	const usage = mapResponsesUsage(response);
+	let anthropicContent = fallbackContentBlocks(response.output_text);
+	if (contentBlocks.length > 0) anthropicContent = contentBlocks;
+	const stopReason = mapResponsesStopReason(response);
 	return {
 		id: response.id,
 		type: "message",
 		role: "assistant",
+		content: anthropicContent,
 		model: response.model,
-		content: assistantContentBlocks,
-		stop_reason: mapOpenAIStopReasonToAnthropic(stopReason),
+		stop_reason: stopReason,
 		stop_sequence: null,
-		usage: {
-			input_tokens: (response.usage?.prompt_tokens ?? 0) - (response.usage?.prompt_tokens_details?.cached_tokens ?? 0),
-			output_tokens: response.usage?.completion_tokens ?? 0,
-			...response.usage?.prompt_tokens_details?.cached_tokens !== void 0 && { cache_read_input_tokens: response.usage.prompt_tokens_details.cached_tokens }
+		usage
+	};
+};
+const mapOutputToAnthropicContent = (output) => {
+	const contentBlocks = [];
+	for (const item of output) switch (item.type) {
+		case "reasoning": {
+			const thinkingText = extractReasoningText(item);
+			if (thinkingText.length > 0) contentBlocks.push({
+				type: "thinking",
+				thinking: thinkingText,
+				signature: (item.encrypted_content ?? "") + "@" + item.id
+			});
+			break;
+		}
+		case "function_call": {
+			const toolUseBlock = createToolUseContentBlock(item);
+			if (toolUseBlock) contentBlocks.push(toolUseBlock);
+			break;
+		}
+		case "message": {
+			const combinedText = combineMessageTextContent(item.content);
+			if (combinedText.length > 0) contentBlocks.push({
+				type: "text",
+				text: combinedText
+			});
+			break;
+		}
+		case "compaction": {
+			const compactionBlock = createCompactionThinkingBlock(item);
+			if (compactionBlock) contentBlocks.push(compactionBlock);
+			break;
+		}
+		default: {
+			const combinedText = combineMessageTextContent(item.content);
+			if (combinedText.length > 0) contentBlocks.push({
+				type: "text",
+				text: combinedText
+			});
+		}
+	}
+	return contentBlocks;
+};
+const combineMessageTextContent = (content) => {
+	if (!Array.isArray(content)) return "";
+	let aggregated = "";
+	for (const block of content) {
+		if (isResponseOutputText(block)) {
+			aggregated += block.text;
+			continue;
+		}
+		if (isResponseOutputRefusal(block)) {
+			aggregated += block.refusal;
+			continue;
+		}
+		if (typeof block.text === "string") {
+			aggregated += block.text;
+			continue;
+		}
+		if (typeof block.reasoning === "string") {
+			aggregated += block.reasoning;
+			continue;
+		}
+	}
+	return aggregated;
+};
+const extractReasoningText = (item) => {
+	const segments = [];
+	const collectFromBlocks = (blocks) => {
+		if (!Array.isArray(blocks)) return;
+		for (const block of blocks) if (typeof block.text === "string") {
+			segments.push(block.text);
+			continue;
 		}
 	};
-}
-function getAnthropicTextBlocks(messageContent) {
-	if (typeof messageContent === "string" && messageContent.length > 0) return [{
-		type: "text",
-		text: messageContent
-	}];
-	if (Array.isArray(messageContent)) return messageContent.filter((part) => part.type === "text").map((part) => ({
-		type: "text",
-		text: part.text
-	}));
-	return [];
-}
-function getAnthropicThinkBlocks(reasoningText, reasoningOpaque) {
-	if (reasoningText && reasoningText.length > 0) return [{
-		type: "thinking",
-		thinking: reasoningText,
-		signature: reasoningOpaque || ""
-	}];
-	if (reasoningOpaque && reasoningOpaque.length > 0) return [{
-		type: "thinking",
-		thinking: THINKING_TEXT,
-		signature: reasoningOpaque
-	}];
-	return [];
-}
-function getAnthropicToolUseBlocks(toolCalls) {
-	if (!toolCalls) return [];
-	return toolCalls.map((toolCall) => ({
+	if (!item.summary || item.summary.length === 0) return THINKING_TEXT$1;
+	collectFromBlocks(item.summary);
+	return segments.join("").trim();
+};
+const createToolUseContentBlock = (call) => {
+	const toolId = call.call_id;
+	if (!call.name || !toolId) return null;
+	const input = parseFunctionCallArguments(call.arguments);
+	return {
 		type: "tool_use",
-		id: toolCall.id,
-		name: toolCall.function.name,
-		input: JSON.parse(toolCall.function.arguments)
-	}));
-}
-
-//#endregion
-//#region src/routes/messages/count-tokens-handler.ts
-/**
-* Handles token counting for Anthropic messages
-*/
-async function handleCountTokens(c) {
+		id: toolId,
+		name: call.name,
+		input
+	};
+};
+const createCompactionThinkingBlock = (item) => {
+	if (!item.id || !item.encrypted_content) return null;
+	return {
+		type: "thinking",
+		thinking: THINKING_TEXT$1,
+		signature: encodeCompactionCarrierSignature({
+			id: item.id,
+			encrypted_content: item.encrypted_content
+		})
+	};
+};
+const parseFunctionCallArguments = (rawArguments) => {
+	if (typeof rawArguments !== "string" || rawArguments.trim().length === 0) return {};
 	try {
-		const anthropicBeta = c.req.header("anthropic-beta");
-		const anthropicPayload = await c.req.json();
-		const openAIPayload = translateToOpenAI(anthropicPayload);
-		const selectedModel = findEndpointModel(anthropicPayload.model);
-		anthropicPayload.model = selectedModel?.id ?? anthropicPayload.model;
-		if (!selectedModel) {
-			consola.warn("Model not found, returning default token count");
-			return c.json({ input_tokens: 1 });
-		}
-		const tokenCount = await getTokenCount(openAIPayload, selectedModel);
-		if (anthropicPayload.tools && anthropicPayload.tools.length > 0) {
-			let addToolSystemPromptCount = false;
-			if (anthropicBeta) {
-				const toolsLength = anthropicPayload.tools.length;
-				addToolSystemPromptCount = !anthropicPayload.tools.some((tool) => tool.name.startsWith("mcp__") || tool.name === "Skill" && toolsLength === 1);
-			}
-			if (addToolSystemPromptCount) {
-				if (anthropicPayload.model.startsWith("claude")) tokenCount.input = tokenCount.input + 346;
-				else if (anthropicPayload.model.startsWith("grok")) tokenCount.input = tokenCount.input + 120;
-			}
-		}
-		let finalTokenCount = tokenCount.input + tokenCount.output;
-		if (anthropicPayload.model.startsWith("claude")) finalTokenCount = Math.round(finalTokenCount * 1.15);
-		consola.info("Token count:", finalTokenCount);
-		return c.json({ input_tokens: finalTokenCount });
+		const parsed = JSON.parse(rawArguments);
+		if (Array.isArray(parsed)) return { arguments: parsed };
+		if (parsed && typeof parsed === "object") return parsed;
 	} catch (error) {
-		consola.error("Error counting tokens:", error);
-		return c.json({ input_tokens: 1 });
+		consola.warn("Failed to parse function call arguments", {
+			error,
+			rawArguments
+		});
 	}
-}
+	return { raw_arguments: rawArguments };
+};
+const fallbackContentBlocks = (outputText) => {
+	if (!outputText) return [];
+	return [{
+		type: "text",
+		text: outputText
+	}];
+};
+const mapResponsesStopReason = (response) => {
+	const { status, incomplete_details: incompleteDetails } = response;
+	if (status === "completed") {
+		if (response.output.some((item) => item.type === "function_call")) return "tool_use";
+		return "end_turn";
+	}
+	if (status === "incomplete") {
+		if (incompleteDetails?.reason === "max_output_tokens") return "max_tokens";
+		if (incompleteDetails?.reason === "content_filter") return "end_turn";
+	}
+	return null;
+};
+const mapResponsesUsage = (response) => {
+	const inputTokens = response.usage?.input_tokens ?? 0;
+	const outputTokens = response.usage?.output_tokens ?? 0;
+	const inputCachedTokens = response.usage?.input_tokens_details?.cached_tokens;
+	return {
+		input_tokens: inputTokens - (inputCachedTokens ?? 0),
+		output_tokens: outputTokens,
+		...response.usage?.input_tokens_details?.cached_tokens !== void 0 && { cache_read_input_tokens: response.usage.input_tokens_details.cached_tokens }
+	};
+};
+const isRecord = (value) => typeof value === "object" && value !== null;
+const isResponseOutputText = (block) => isRecord(block) && "type" in block && block.type === "output_text";
+const isResponseOutputRefusal = (block) => isRecord(block) && "type" in block && block.type === "refusal";
+const convertToolResultContent = (content) => {
+	if (typeof content === "string") return content;
+	if (Array.isArray(content)) {
+		const result = [];
+		for (const block of content) switch (block.type) {
+			case "text":
+				result.push(createTextContent(block.text));
+				break;
+			case "image":
+				result.push(createImageContent(block));
+				break;
+			default: break;
+		}
+		return result;
+	}
+	return "";
+};
 
 //#endregion
 //#region src/routes/messages/responses-stream-translation.ts
@@ -4441,6 +4477,7 @@ const createMessages = async (payload, account, options) => {
 		"x-initiator": options?.subagentMarker ? "agent" : initiator
 	};
 	prepareInteractionHeaders(options?.sessionId, Boolean(options?.subagentMarker), headers);
+	prepareForCompact(headers, options?.isCompact);
 	const anthropicBeta = buildAnthropicBetaHeader(options?.anthropicBetaHeader, payload.thinking);
 	if (anthropicBeta) headers["anthropic-beta"] = anthropicBeta;
 	const response = await fetch(`${copilotBaseUrl(ctx)}/v1/messages`, {
@@ -4794,7 +4831,7 @@ async function handleCompletion(c) {
 	const streamRequested = Boolean(anthropicPayload.stream);
 	const rawUserId = anthropicPayload.metadata?.user_id;
 	const userId = typeof rawUserId === "string" ? rawUserId : void 0;
-	const { safetyIdentifier, promptCacheKey } = parseUserId(userId);
+	const { safetyIdentifier, sessionId: promptCacheKey } = parseUserIdMetadata(userId);
 	const normalizedSafetyIdentifier = safetyIdentifier ?? void 0;
 	const normalizedPromptCacheKey = promptCacheKey ?? void 0;
 	const blockedResponse = maybeBlockOriginalModelName({
@@ -4887,7 +4924,8 @@ async function handleCompletion(c) {
 		subagentMarker,
 		sessionId,
 		instr,
-		selectedModel
+		selectedModel,
+		isCompact
 	});
 	if (endpoint === RESPONSES_ENDPOINT$1) return await handleWithResponsesApi({
 		c,
@@ -4897,7 +4935,8 @@ async function handleCompletion(c) {
 		subagentMarker,
 		sessionId,
 		selectedModel,
-		instr
+		instr,
+		isCompact
 	});
 	return await handleWithChatCompletions({
 		c,
@@ -4906,11 +4945,12 @@ async function handleCompletion(c) {
 		subagentMarker,
 		sessionId,
 		selectedModel,
-		instr
+		instr,
+		isCompact
 	});
 }
 const handleWithChatCompletions = async (params) => {
-	const { c, openAIPayload, initiatorOverride, subagentMarker, sessionId, selectedModel, instr } = params;
+	const { c, openAIPayload, initiatorOverride, subagentMarker, sessionId, selectedModel, instr, isCompact } = params;
 	logger$5.debug("Translated OpenAI request payload:", JSON.stringify(openAIPayload));
 	const ctx = toAccountContext(instr.account);
 	const initiator = initiatorOverride ?? getChatInitiator(openAIPayload.messages);
@@ -4921,7 +4961,8 @@ const handleWithChatCompletions = async (params) => {
 			upstreamRequestId: instr.upstreamRequestId,
 			initiator,
 			subagentMarker,
-			sessionId
+			sessionId,
+			isCompact
 		});
 	} catch (error) {
 		return await handleChatCompletionsCreateError({
@@ -4952,7 +4993,7 @@ const handleWithChatCompletions = async (params) => {
 	}));
 };
 const handleWithResponsesApi = async (params) => {
-	const { c, anthropicPayload, openAIPayload, initiatorOverride, subagentMarker, sessionId, selectedModel, instr } = params;
+	const { c, anthropicPayload, openAIPayload, initiatorOverride, subagentMarker, sessionId, selectedModel, instr, isCompact } = params;
 	const responsesPayload = translateAnthropicMessagesToResponsesPayload(anthropicPayload, selectedModel.id);
 	applyResponsesApiContextManagement(responsesPayload, selectedModel.capabilities.limits.max_prompt_tokens);
 	compactInputByLatestCompaction(responsesPayload);
@@ -4968,7 +5009,8 @@ const handleWithResponsesApi = async (params) => {
 			initiator: resolvedInitiator,
 			upstreamRequestId: instr.upstreamRequestId,
 			subagentMarker,
-			sessionId
+			sessionId,
+			isCompact
 		}, ctx);
 	} catch (error) {
 		return await handleResponsesCreateError({
@@ -5419,7 +5461,7 @@ async function streamMessagesAndLog(params) {
 	}
 }
 const handleWithMessagesApi = async (params) => {
-	const { c, anthropicPayload, anthropicBetaHeader, initiatorOverride, subagentMarker, sessionId, instr, selectedModel } = params;
+	const { c, anthropicPayload, anthropicBetaHeader, initiatorOverride, subagentMarker, sessionId, instr, selectedModel, isCompact } = params;
 	for (const msg of anthropicPayload.messages) if (msg.role === "assistant" && Array.isArray(msg.content)) msg.content = msg.content.filter((block) => {
 		if (block.type !== "thinking") return true;
 		return block.thinking && block.thinking !== "Thinking..." && block.signature && !block.signature.includes("@");
@@ -5443,7 +5485,8 @@ const handleWithMessagesApi = async (params) => {
 			upstreamRequestId: instr.upstreamRequestId,
 			initiator,
 			subagentMarker,
-			sessionId
+			sessionId,
+			isCompact
 		});
 	} catch (error) {
 		return await handleMessagesCreateError({
@@ -5665,14 +5708,37 @@ async function handleProviderMessages(c) {
 			provider
 		}));
 		const upstreamResponse = await forwardProviderMessages(providerConfig, payload, c.req.raw.headers);
+		if (!upstreamResponse.ok) {
+			logger$3.error("Failed to create responses", upstreamResponse);
+			throw new HTTPError("Failed to create responses", upstreamResponse);
+		}
 		const contentType = upstreamResponse.headers.get("content-type") ?? "";
 		if (Boolean(payload.stream) && contentType.includes("text/event-stream")) {
 			logger$3.debug("provider.messages.streaming");
 			return streamSSE(c, async (stream) => {
-				for await (const event of events(upstreamResponse)) {
-					const eventName = event.event;
-					const data = event.data ?? "";
-					logger$3.debug("provider.messages.raw_stream_event", data);
+				for await (const chunk of events(upstreamResponse)) {
+					logger$3.debug("provider.messages.raw_stream_event:", chunk.data);
+					const eventName = chunk.event;
+					if (eventName === "ping") {
+						await stream.writeSSE({
+							event: "ping",
+							data: "{\"type\":\"ping\"}"
+						});
+						continue;
+					}
+					let data = chunk.data;
+					if (!data) continue;
+					try {
+						const parsed = JSON.parse(data);
+						if (parsed.type === "message_start") adjustInputTokens(providerConfig, parsed.message.usage);
+						else if (parsed.type === "message_delta") adjustInputTokens(providerConfig, parsed.usage);
+						data = JSON.stringify(parsed);
+					} catch (error) {
+						logger$3.error("provider.messages.streaming.adjust_tokens_error", {
+							error,
+							originalData: data
+						});
+					}
 					await stream.writeSSE({
 						event: eventName,
 						data
@@ -5680,7 +5746,10 @@ async function handleProviderMessages(c) {
 				}
 			});
 		}
-		return createProviderProxyResponse(upstreamResponse);
+		const jsonBody = await upstreamResponse.json();
+		adjustInputTokens(providerConfig, jsonBody.usage);
+		logger$3.debug("provider.messages.no_stream result:", JSON.stringify(jsonBody));
+		return c.json(jsonBody);
 	} catch (error) {
 		logger$3.error("provider.messages.error", {
 			provider,
@@ -5689,6 +5758,11 @@ async function handleProviderMessages(c) {
 		throw error;
 	}
 }
+const adjustInputTokens = (providerConfig, usage) => {
+	if (!providerConfig.adjustInputTokens || !usage) return;
+	usage.input_tokens = Math.max(0, (usage.input_tokens ?? 0) - (usage.cache_read_input_tokens ?? 0) - (usage.cache_creation_input_tokens ?? 0));
+	logger$3.debug("provider.messages.adjusted_usage:", JSON.stringify(usage));
+};
 
 //#endregion
 //#region src/routes/provider/messages/route.ts
@@ -5788,7 +5862,7 @@ const handleResponses = async (c) => {
 	const streamRequested = Boolean(payload.stream);
 	const { initiator: initialInitiator } = getResponsesRequestOptions(payload);
 	const userId = payload.metadata?.user_id;
-	const { safetyIdentifier, promptCacheKey } = parseUserId(userId);
+	const { safetyIdentifier, sessionId: promptCacheKey } = parseUserIdMetadata(userId);
 	const normalizedSafetyIdentifier = safetyIdentifier ?? void 0;
 	const normalizedPromptCacheKey = promptCacheKey ?? void 0;
 	request.userId = userId;
@@ -6295,6 +6369,7 @@ usageRoute.get("/:accountIndex", async (c) => {
 //#endregion
 //#region src/server.ts
 const server = new Hono();
+server.use(traceIdMiddleware);
 server.use(logger());
 server.use(cors());
 server.use("*", createAuthMiddleware({
@@ -6320,4 +6395,4 @@ server.route("/:provider/v1/models", providerModelRoutes);
 
 //#endregion
 export { server };
-//# sourceMappingURL=server-CM_0PrbK.js.map
+//# sourceMappingURL=server-D3sySKxC.js.map