@hsupu/copilot-api 0.7.0 → 0.8.0

package/dist/main.js

@@ -12,9 +12,11 @@ import { getProxyForUrl } from "proxy-from-env";
 import { Agent, ProxyAgent, setGlobalDispatcher } from "undici";
 import { execSync } from "node:child_process";
 import process$1 from "node:process";
+import { Box, Text, render, useInput, useStdout } from "ink";
+import React, { useEffect, useState } from "react";
+import { Fragment, jsx, jsxs } from "react/jsx-runtime";
 import { Hono } from "hono";
 import { cors } from "hono/cors";
-import { logger } from "hono/logger";
 import { streamSSE } from "hono/streaming";
 import { events } from "fetch-event-stream";
 
@@ -45,7 +47,8 @@ const state = {
 	accountType: "individual",
 	manualApprove: false,
 	rateLimitWait: false,
-	showToken: false
+	showToken: false,
+	autoCompact: false
 };
 
 //#endregion
@@ -104,6 +107,27 @@ var HTTPError = class HTTPError extends Error {
 		return new HTTPError(message, response.status, text);
 	}
 };
+/** Parse token limit info from error message */
+function parseTokenLimitError(message) {
+	const match = message.match(/prompt token count of (\d+) exceeds the limit of (\d+)/);
+	if (match) return {
+		current: Number.parseInt(match[1], 10),
+		limit: Number.parseInt(match[2], 10)
+	};
+	return null;
+}
+/** Format Anthropic-compatible error for token limit exceeded */
+function formatTokenLimitError(current, limit) {
+	const excess = current - limit;
+	const percentage = Math.round(excess / limit * 100);
+	return {
+		type: "error",
+		error: {
+			type: "invalid_request_error",
+			message: `prompt is too long: ${current} tokens > ${limit} maximum (${excess} tokens over, ${percentage}% excess)`
+		}
+	};
+}
 async function forwardError(c, error) {
 	consola.error("Error occurred:", error);
 	if (error instanceof HTTPError) {
@@ -114,6 +138,15 @@ async function forwardError(c, error) {
 			errorJson = error.responseText;
 		}
 		consola.error("HTTP error:", errorJson);
+		const copilotError = errorJson;
+		if (copilotError.error?.code === "model_max_prompt_tokens_exceeded") {
+			const tokenInfo = parseTokenLimitError(copilotError.error.message ?? "");
+			if (tokenInfo) {
+				const formattedError = formatTokenLimitError(tokenInfo.current, tokenInfo.limit);
+				consola.info("Returning formatted token limit error:", formattedError);
+				return c.json(formattedError, 400);
+			}
+		}
 		return c.json({ error: {
 			message: error.responseText,
 			type: "error"
@@ -476,7 +509,7 @@ const logout = defineCommand({
 
 //#endregion
 //#region src/lib/history.ts
-function generateId() {
+function generateId$1() {
 	return Date.now().toString(36) + Math.random().toString(36).slice(2, 9);
 }
 const historyState = {
@@ -492,7 +525,7 @@ function initHistory(enabled, maxEntries) {
 	historyState.maxEntries = maxEntries;
 	historyState.entries = [];
 	historyState.sessions = /* @__PURE__ */ new Map();
-	historyState.currentSessionId = enabled ? generateId() : "";
+	historyState.currentSessionId = enabled ? generateId$1() : "";
 }
 function isHistoryEnabled() {
 	return historyState.enabled;
@@ -506,7 +539,7 @@ function getCurrentSession(endpoint) {
 			return historyState.currentSessionId;
 		}
 	}
-	const sessionId = generateId();
+	const sessionId = generateId$1();
 	historyState.currentSessionId = sessionId;
 	historyState.sessions.set(sessionId, {
 		id: sessionId,
@@ -526,7 +559,7 @@ function recordRequest(endpoint, request) {
 	const session = historyState.sessions.get(sessionId);
 	if (!session) return "";
 	const entry = {
-		id: generateId(),
+		id: generateId$1(),
 		sessionId,
 		timestamp: Date.now(),
 		endpoint,
@@ -543,7 +576,11 @@ function recordRequest(endpoint, request) {
 	historyState.entries.push(entry);
 	session.requestCount++;
 	if (!session.models.includes(request.model)) session.models.push(request.model);
-	while (historyState.entries.length > historyState.maxEntries) {
+	if (request.tools && request.tools.length > 0) {
+		if (!session.toolsUsed) session.toolsUsed = [];
+		for (const tool of request.tools) if (!session.toolsUsed.includes(tool.name)) session.toolsUsed.push(tool.name);
+	}
+	while (historyState.maxEntries > 0 && historyState.entries.length > historyState.maxEntries) {
 		const removed = historyState.entries.shift();
 		if (removed) {
 			if (historyState.entries.filter((e) => e.sessionId === removed.sessionId).length === 0) historyState.sessions.delete(removed.sessionId);
@@ -622,13 +659,13 @@ function getSessionEntries(sessionId) {
 function clearHistory() {
 	historyState.entries = [];
 	historyState.sessions = /* @__PURE__ */ new Map();
-	historyState.currentSessionId = generateId();
+	historyState.currentSessionId = generateId$1();
 }
 function deleteSession(sessionId) {
 	if (!historyState.sessions.has(sessionId)) return false;
 	historyState.entries = historyState.entries.filter((e) => e.sessionId !== sessionId);
 	historyState.sessions.delete(sessionId);
-	if (historyState.currentSessionId === sessionId) historyState.currentSessionId = generateId();
+	if (historyState.currentSessionId === sessionId) historyState.currentSessionId = generateId$1();
 	return true;
 }
 function getStats() {
@@ -800,16 +837,16 @@ function generateEnvScript(envVars, commandToRun = "") {
 	let commandBlock;
 	switch (shell) {
 		case "powershell":
-			commandBlock = filteredEnvVars.map(([key, value]) => `$env:${key} = "${value.replace(/"/g, "`\"")}"`).join("; ");
+			commandBlock = filteredEnvVars.map(([key, value]) => `$env:${key} = "${value.replaceAll("\"", "`\"")}"`).join("; ");
 			break;
 		case "cmd":
 			commandBlock = filteredEnvVars.map(([key, value]) => `set ${key}=${value}`).join(" & ");
 			break;
 		case "fish":
-			commandBlock = filteredEnvVars.map(([key, value]) => `set -gx ${key} "${value.replace(/"/g, "\\\"")}"`).join("; ");
+			commandBlock = filteredEnvVars.map(([key, value]) => `set -gx ${key} "${value.replaceAll("\"", String.raw`\"`)}"`).join("; ");
 			break;
 		default: {
-			const assignments = filteredEnvVars.map(([key, value]) => `${key}="${value.replace(/"/g, "\\\"")}"`).join(" ");
+			const assignments = filteredEnvVars.map(([key, value]) => `${key}="${value.replaceAll("\"", String.raw`\"`)}"`).join(" ");
 			commandBlock = filteredEnvVars.length > 0 ? `export ${assignments}` : "";
 			break;
 		}
@@ -819,68 +856,580 @@ function generateEnvScript(envVars, commandToRun = "") {
 }
 
 //#endregion
-//#region src/lib/approval.ts
-const awaitApproval = async () => {
-	if (!await consola.prompt(`Accept incoming request?`, { type: "confirm" })) throw new HTTPError("Request rejected", 403, JSON.stringify({ message: "Request rejected" }));
+//#region src/lib/tui/console-renderer.ts
+function formatDuration$1(ms) {
+	if (ms < 1e3) return `${ms}ms`;
+	return `${(ms / 1e3).toFixed(1)}s`;
+}
+function formatNumber$1(n) {
+	if (n >= 1e6) return `${(n / 1e6).toFixed(1)}M`;
+	if (n >= 1e3) return `${(n / 1e3).toFixed(1)}K`;
+	return String(n);
+}
+function formatTokens$1(input, output) {
+	if (input === void 0 || output === void 0) return "-";
+	return `${formatNumber$1(input)}/${formatNumber$1(output)}`;
+}
+/**
+* Console renderer that shows request lifecycle
+* Start: METHOD /path model-name
+* Complete: METHOD /path 200 1.2s 1.5K/500 model-name
+*/
+var ConsoleRenderer = class {
+	activeRequests = /* @__PURE__ */ new Map();
+	showActive;
+	constructor(options) {
+		this.showActive = options?.showActive ?? true;
+	}
+	onRequestStart(request) {
+		this.activeRequests.set(request.id, request);
+		if (this.showActive) {
+			const modelInfo = request.model ? ` ${request.model}` : "";
+			const queueInfo = request.queuePosition !== void 0 && request.queuePosition > 0 ? ` [q#${request.queuePosition}]` : "";
+			consola.log(`[....] ${request.method} ${request.path}${modelInfo}${queueInfo}`);
+		}
+	}
+	onRequestUpdate(id, update) {
+		const request = this.activeRequests.get(id);
+		if (!request) return;
+		Object.assign(request, update);
+		if (this.showActive && update.status === "streaming") {
+			const modelInfo = request.model ? ` ${request.model}` : "";
+			consola.log(`[<-->] ${request.method} ${request.path}${modelInfo} streaming...`);
+		}
+	}
+	onRequestComplete(request) {
+		this.activeRequests.delete(request.id);
+		const status = request.statusCode ?? 0;
+		const duration = formatDuration$1(request.durationMs ?? 0);
+		const tokens = request.model ? formatTokens$1(request.inputTokens, request.outputTokens) : "";
+		const modelInfo = request.model ? ` ${request.model}` : "";
+		const isError = request.status === "error" || status >= 400;
+		const prefix = isError ? "[FAIL]" : "[ OK ]";
+		const tokensPart = tokens ? ` ${tokens}` : "";
+		const content = `${prefix} ${request.method} ${request.path} ${status} ${duration}${tokensPart}${modelInfo}`;
+		if (isError) {
+			const errorInfo = request.error ? `: ${request.error}` : "";
+			consola.log(content + errorInfo);
+		} else consola.log(content);
+	}
+	destroy() {
+		this.activeRequests.clear();
+	}
 };
 
 //#endregion
-//#region src/lib/queue.ts
-var RequestQueue = class {
-	queue = [];
-	processing = false;
-	lastRequestTime = 0;
-	async enqueue(execute, rateLimitSeconds) {
-		return new Promise((resolve, reject) => {
-			this.queue.push({
-				execute,
-				resolve,
-				reject
-			});
-			if (this.queue.length > 1) {
-				const waitTime = Math.ceil((this.queue.length - 1) * rateLimitSeconds);
-				consola.info(`Request queued. Position: ${this.queue.length}, estimated wait: ${waitTime}s`);
-			}
-			this.processQueue(rateLimitSeconds);
+//#region src/lib/tui/fullscreen-renderer.tsx
+const tuiState = {
+	activeRequests: /* @__PURE__ */ new Map(),
+	completedRequests: [],
+	errorRequests: []
+};
+const listeners = [];
+function notifyListeners() {
+	for (const listener of listeners) listener();
+}
+function formatDuration(ms) {
+	if (ms < 1e3) return `${ms}ms`;
+	return `${(ms / 1e3).toFixed(1)}s`;
+}
+function formatNumber(n) {
+	if (n >= 1e6) return `${(n / 1e6).toFixed(1)}M`;
+	if (n >= 1e3) return `${(n / 1e3).toFixed(1)}K`;
+	return String(n);
+}
+function formatTokens(input, output) {
+	if (input === void 0 || output === void 0) return "-";
+	return `${formatNumber(input)}/${formatNumber(output)}`;
+}
+function getElapsedTime(startTime) {
+	return formatDuration(Date.now() - startTime);
+}
+function TabHeader({ currentTab, counts }) {
+	const tabs = [
+		{
+			key: "active",
+			label: "Active",
+			count: counts.active
+		},
+		{
+			key: "completed",
+			label: "Completed",
+			count: counts.completed
+		},
+		{
+			key: "errors",
+			label: "Errors",
+			count: counts.errors
+		}
+	];
+	return /* @__PURE__ */ jsxs(Box, {
+		borderStyle: "single",
+		paddingX: 1,
+		children: [tabs.map((tab, idx) => /* @__PURE__ */ jsxs(React.Fragment, { children: [idx > 0 && /* @__PURE__ */ jsx(Text, { children: " │ " }), /* @__PURE__ */ jsxs(Text, {
+			bold: currentTab === tab.key,
+			color: currentTab === tab.key ? "cyan" : void 0,
+			inverse: currentTab === tab.key,
+			children: [
+				" ",
+				"[",
+				idx + 1,
+				"] ",
+				tab.label,
+				" (",
+				tab.count,
+				")",
+				" "
+			]
+		})] }, tab.key)), /* @__PURE__ */ jsx(Text, {
+			dimColor: true,
+			children: " │ Press 1/2/3 to switch tabs, q to quit"
+		})]
+	});
+}
+function getStatusColor(status) {
+	if (status === "streaming") return "yellow";
+	if (status === "queued") return "gray";
+	return "blue";
+}
+function getStatusIcon(status) {
+	if (status === "streaming") return "⟳";
+	if (status === "queued") return "◷";
+	return "●";
+}
+function ActiveRequestRow({ request }) {
+	const [, setTick] = useState(0);
+	useEffect(() => {
+		const interval = setInterval(() => setTick((t) => t + 1), 1e3);
+		return () => clearInterval(interval);
+	}, []);
+	const statusColor = getStatusColor(request.status);
+	const statusIcon = getStatusIcon(request.status);
+	return /* @__PURE__ */ jsxs(Box, { children: [
+		/* @__PURE__ */ jsxs(Text, {
+			color: statusColor,
+			children: [statusIcon, " "]
+		}),
+		/* @__PURE__ */ jsx(Text, {
+			bold: true,
+			children: request.method
+		}),
+		/* @__PURE__ */ jsxs(Text, { children: [
+			" ",
+			request.path,
+			" "
+		] }),
+		/* @__PURE__ */ jsxs(Text, {
+			dimColor: true,
+			children: [getElapsedTime(request.startTime), " "]
+		}),
+		request.queuePosition !== void 0 && request.queuePosition > 0 && /* @__PURE__ */ jsxs(Text, {
+			color: "gray",
+			children: [
+				"[queue #",
+				request.queuePosition,
+				"] "
+			]
+		}),
+		/* @__PURE__ */ jsx(Text, {
+			color: "magenta",
+			children: request.model
+		})
+	] });
+}
+function CompletedRequestRow({ request }) {
+	const isError = request.status === "error" || (request.statusCode ?? 0) >= 400;
+	return /* @__PURE__ */ jsxs(Box, { children: [
+		/* @__PURE__ */ jsxs(Text, {
+			color: isError ? "red" : "green",
+			children: [isError ? "✗" : "✓", " "]
+		}),
+		/* @__PURE__ */ jsx(Text, {
+			bold: true,
+			children: request.method
+		}),
+		/* @__PURE__ */ jsxs(Text, { children: [
+			" ",
+			request.path,
+			" "
+		] }),
+		/* @__PURE__ */ jsxs(Text, {
+			color: isError ? "red" : "green",
+			children: [request.statusCode ?? "-", " "]
+		}),
+		/* @__PURE__ */ jsxs(Text, {
+			dimColor: true,
+			children: [formatDuration(request.durationMs ?? 0), " "]
+		}),
+		/* @__PURE__ */ jsxs(Text, { children: [formatTokens(request.inputTokens, request.outputTokens), " "] }),
+		/* @__PURE__ */ jsx(Text, {
+			color: "magenta",
+			children: request.model
+		})
+	] });
+}
+function ErrorRequestRow({ request }) {
+	return /* @__PURE__ */ jsxs(Box, {
+		flexDirection: "column",
+		children: [/* @__PURE__ */ jsxs(Box, { children: [
+			/* @__PURE__ */ jsx(Text, {
+				color: "red",
+				children: "✗ "
+			}),
+			/* @__PURE__ */ jsx(Text, {
+				bold: true,
+				children: request.method
+			}),
+			/* @__PURE__ */ jsxs(Text, { children: [
+				" ",
+				request.path,
+				" "
+			] }),
+			/* @__PURE__ */ jsxs(Text, {
+				color: "red",
+				children: [request.statusCode ?? "-", " "]
+			}),
+			/* @__PURE__ */ jsxs(Text, {
+				dimColor: true,
+				children: [formatDuration(request.durationMs ?? 0), " "]
+			}),
+			/* @__PURE__ */ jsx(Text, {
+				color: "magenta",
+				children: request.model
+			})
+		] }), request.error && /* @__PURE__ */ jsx(Box, {
+			marginLeft: 2,
+			children: /* @__PURE__ */ jsxs(Text, {
+				color: "red",
+				dimColor: true,
+				children: ["└─ ", request.error]
+			})
+		})]
+	});
+}
+function ContentPanel({ currentTab, activeList, completedList, errorList, contentHeight }) {
+	if (currentTab === "active") {
+		if (activeList.length === 0) return /* @__PURE__ */ jsx(Text, {
+			dimColor: true,
+			children: "No active requests"
 		});
+		return /* @__PURE__ */ jsx(Fragment, { children: activeList.slice(0, contentHeight).map((req) => /* @__PURE__ */ jsx(ActiveRequestRow, { request: req }, req.id)) });
 	}
-	async processQueue(rateLimitSeconds) {
-		if (this.processing) return;
-		this.processing = true;
-		while (this.queue.length > 0) {
-			const elapsedMs = Date.now() - this.lastRequestTime;
-			const requiredMs = rateLimitSeconds * 1e3;
-			if (this.lastRequestTime > 0 && elapsedMs < requiredMs) {
-				const waitMs = requiredMs - elapsedMs;
-				consola.debug(`Rate limit: waiting ${Math.ceil(waitMs / 1e3)}s`);
-				await new Promise((resolve) => setTimeout(resolve, waitMs));
-			}
-			const request = this.queue.shift();
-			if (!request) break;
-			this.lastRequestTime = Date.now();
-			try {
-				const result = await request.execute();
-				request.resolve(result);
-			} catch (error) {
-				request.reject(error);
-			}
+	if (currentTab === "completed") {
+		if (completedList.length === 0) return /* @__PURE__ */ jsx(Text, {
+			dimColor: true,
+			children: "No completed requests"
+		});
+		return /* @__PURE__ */ jsx(Fragment, { children: completedList.slice(-contentHeight).reverse().map((req) => /* @__PURE__ */ jsx(CompletedRequestRow, { request: req }, req.id)) });
+	}
+	if (errorList.length === 0) return /* @__PURE__ */ jsx(Text, {
+		dimColor: true,
+		children: "No errors"
+	});
+	return /* @__PURE__ */ jsx(Fragment, { children: errorList.slice(-contentHeight).reverse().map((req) => /* @__PURE__ */ jsx(ErrorRequestRow, { request: req }, req.id)) });
+}
+function TuiApp() {
+	const [currentTab, setCurrentTab] = useState("active");
+	const [, forceUpdate] = useState(0);
+	const { stdout } = useStdout();
+	useEffect(() => {
+		const listener = () => forceUpdate((n) => n + 1);
+		listeners.push(listener);
+		return () => {
+			const idx = listeners.indexOf(listener);
+			if (idx !== -1) listeners.splice(idx, 1);
+		};
+	}, []);
+	useInput((input, key) => {
+		switch (input) {
+			case "1":
+				setCurrentTab("active");
+				break;
+			case "2":
+				setCurrentTab("completed");
+				break;
+			case "3":
+				setCurrentTab("errors");
+				break;
+			default: if (input === "q" || key.ctrl && input === "c") process.exit(0);
 		}
-		this.processing = false;
+	});
+	const activeList = Array.from(tuiState.activeRequests.values());
+	const completedList = tuiState.completedRequests;
+	const errorList = tuiState.errorRequests;
+	const counts = {
+		active: activeList.length,
+		completed: completedList.length,
+		errors: errorList.length
+	};
+	const terminalHeight = stdout.rows || 24;
+	const contentHeight = terminalHeight - 3 - 1 - 2;
+	return /* @__PURE__ */ jsxs(Box, {
+		flexDirection: "column",
+		height: terminalHeight,
+		children: [
+			/* @__PURE__ */ jsx(TabHeader, {
+				currentTab,
+				counts
+			}),
+			/* @__PURE__ */ jsx(Box, {
+				flexDirection: "column",
+				height: contentHeight,
+				borderStyle: "single",
+				paddingX: 1,
+				overflow: "hidden",
+				children: /* @__PURE__ */ jsx(ContentPanel, {
+					currentTab,
+					activeList,
+					completedList,
+					errorList,
+					contentHeight
+				})
+			}),
+			/* @__PURE__ */ jsx(Box, {
+				paddingX: 1,
+				children: /* @__PURE__ */ jsxs(Text, {
+					dimColor: true,
+					children: [
+						"copilot-api │ Active: ",
+						counts.active,
+						" │ Completed: ",
+						counts.completed,
+						" ",
+						"│ Errors: ",
+						counts.errors
+					]
+				})
+			})
+		]
+	});
+}
+/**
+* Fullscreen TUI renderer using Ink
+* Provides interactive terminal interface with tabs
+*/
+var FullscreenRenderer = class {
+	inkInstance = null;
+	maxHistory = 100;
+	constructor(options) {
+		if (options?.maxHistory !== void 0) this.maxHistory = options.maxHistory;
 	}
-	get length() {
-		return this.queue.length;
+	start() {
+		if (this.inkInstance) return;
+		this.inkInstance = render(/* @__PURE__ */ jsx(TuiApp, {}), {});
+	}
+	onRequestStart(request) {
+		tuiState.activeRequests.set(request.id, { ...request });
+		notifyListeners();
+	}
+	onRequestUpdate(id, update) {
+		const request = tuiState.activeRequests.get(id);
+		if (!request) return;
+		Object.assign(request, update);
+		notifyListeners();
+	}
+	onRequestComplete(request) {
+		tuiState.activeRequests.delete(request.id);
+		if (request.status === "error" || (request.statusCode ?? 0) >= 400) {
+			tuiState.errorRequests.push({ ...request });
+			while (tuiState.errorRequests.length > this.maxHistory) tuiState.errorRequests.shift();
+		}
+		tuiState.completedRequests.push({ ...request });
+		while (tuiState.completedRequests.length > this.maxHistory) tuiState.completedRequests.shift();
+		notifyListeners();
+	}
+	destroy() {
+		if (this.inkInstance) {
+			this.inkInstance.unmount();
+			this.inkInstance = null;
+		}
+		tuiState.activeRequests.clear();
+		tuiState.completedRequests = [];
+		tuiState.errorRequests = [];
 	}
 };
-const requestQueue = new RequestQueue();
+
+//#endregion
+//#region src/lib/tui/tracker.ts
+function generateId() {
+	return Date.now().toString(36) + Math.random().toString(36).slice(2, 6);
+}
+var RequestTracker = class {
+	requests = /* @__PURE__ */ new Map();
+	renderer = null;
+	completedQueue = [];
+	historySize = 5;
+	completedDisplayMs = 2e3;
+	setRenderer(renderer) {
+		this.renderer = renderer;
+	}
+	setOptions(options) {
+		if (options.historySize !== void 0) this.historySize = options.historySize;
+		if (options.completedDisplayMs !== void 0) this.completedDisplayMs = options.completedDisplayMs;
+	}
+	/**
+	* Start tracking a new request
+	* Returns the tracking ID
+	*/
+	startRequest(method, path$1, model) {
+		const id = generateId();
+		const request = {
+			id,
+			method,
+			path: path$1,
+			model,
+			startTime: Date.now(),
+			status: "executing"
+		};
+		this.requests.set(id, request);
+		this.renderer?.onRequestStart(request);
+		return id;
+	}
+	/**
+	* Update request status
+	*/
+	updateRequest(id, update) {
+		const request = this.requests.get(id);
+		if (!request) return;
+		if (update.status !== void 0) request.status = update.status;
+		if (update.statusCode !== void 0) request.statusCode = update.statusCode;
+		if (update.durationMs !== void 0) request.durationMs = update.durationMs;
+		if (update.inputTokens !== void 0) request.inputTokens = update.inputTokens;
+		if (update.outputTokens !== void 0) request.outputTokens = update.outputTokens;
+		if (update.error !== void 0) request.error = update.error;
+		if (update.queuePosition !== void 0) request.queuePosition = update.queuePosition;
+		this.renderer?.onRequestUpdate(id, update);
+	}
+	/**
+	* Mark request as completed
+	*/
+	completeRequest(id, statusCode, usage) {
+		const request = this.requests.get(id);
+		if (!request) return;
+		request.status = statusCode >= 200 && statusCode < 400 ? "completed" : "error";
+		request.statusCode = statusCode;
+		request.durationMs = Date.now() - request.startTime;
+		if (usage) {
+			request.inputTokens = usage.inputTokens;
+			request.outputTokens = usage.outputTokens;
+		}
+		this.renderer?.onRequestComplete(request);
+		this.requests.delete(id);
+		this.completedQueue.push(request);
+		while (this.completedQueue.length > this.historySize) this.completedQueue.shift();
+		setTimeout(() => {
+			const idx = this.completedQueue.indexOf(request);
+			if (idx !== -1) this.completedQueue.splice(idx, 1);
+		}, this.completedDisplayMs);
+	}
+	/**
+	* Mark request as failed with error
+	*/
+	failRequest(id, error) {
+		const request = this.requests.get(id);
+		if (!request) return;
+		request.status = "error";
+		request.error = error;
+		request.durationMs = Date.now() - request.startTime;
+		this.renderer?.onRequestComplete(request);
+		this.requests.delete(id);
+		this.completedQueue.push(request);
+		while (this.completedQueue.length > this.historySize) this.completedQueue.shift();
+	}
+	/**
+	* Get all active requests
+	*/
+	getActiveRequests() {
+		return Array.from(this.requests.values());
+	}
+	/**
+	* Get recently completed requests
+	*/
+	getCompletedRequests() {
+		return [...this.completedQueue];
+	}
+	/**
+	* Get request by ID
+	*/
+	getRequest(id) {
+		return this.requests.get(id);
+	}
+	/**
+	* Clear all tracked requests
+	*/
+	clear() {
+		this.requests.clear();
+		this.completedQueue = [];
+	}
+};
+const requestTracker = new RequestTracker();
+
+//#endregion
+//#region src/lib/tui/middleware.ts
 /**
-* Execute a request with rate limiting via queue.
-* Requests are queued and processed sequentially at the configured rate.
+* Custom logger middleware that tracks requests through the TUI system
+* Shows single-line output: METHOD /path 200 1.2s 1.5K/500 model-name
+*
+* For streaming responses (SSE), the handler is responsible for calling
+* completeRequest after the stream finishes.
 */
-async function executeWithRateLimit(state$1, execute) {
-	if (state$1.rateLimitSeconds === void 0) return execute();
-	return requestQueue.enqueue(execute, state$1.rateLimitSeconds);
+function tuiLogger() {
+	return async (c, next) => {
+		const method = c.req.method;
+		const path$1 = c.req.path;
+		const trackingId = requestTracker.startRequest(method, path$1, "");
+		c.set("trackingId", trackingId);
+		try {
+			await next();
+			if ((c.res.headers.get("content-type") ?? "").includes("text/event-stream")) return;
+			const status = c.res.status;
+			const inputTokens = c.res.headers.get("x-input-tokens");
+			const outputTokens = c.res.headers.get("x-output-tokens");
+			const model = c.res.headers.get("x-model");
+			if (model) {
+				const request = requestTracker.getRequest(trackingId);
+				if (request) request.model = model;
+			}
+			requestTracker.completeRequest(trackingId, status, inputTokens && outputTokens ? {
+				inputTokens: Number.parseInt(inputTokens, 10),
+				outputTokens: Number.parseInt(outputTokens, 10)
+			} : void 0);
+		} catch (error) {
+			requestTracker.failRequest(trackingId, error instanceof Error ? error.message : "Unknown error");
+			throw error;
+		}
+	};
+}
+
+//#endregion
+//#region src/lib/tui/index.ts
+/**
+* Initialize the TUI system
+* @param options.mode - "console" for simple log output (default), "fullscreen" for interactive TUI
+*/
+function initTui(options) {
+	const enabled = options?.enabled ?? process.stdout.isTTY;
+	const mode = options?.mode ?? "console";
+	if (enabled) if (mode === "fullscreen") {
+		const renderer = new FullscreenRenderer({ maxHistory: options?.historySize ?? 100 });
+		requestTracker.setRenderer(renderer);
+		renderer.start();
+	} else {
+		const renderer = new ConsoleRenderer();
+		requestTracker.setRenderer(renderer);
+	}
+	if (options?.historySize !== void 0 || options?.completedDisplayMs !== void 0) requestTracker.setOptions({
+		historySize: options.historySize,
+		completedDisplayMs: options.completedDisplayMs
+	});
 }
 
+//#endregion
+//#region src/lib/approval.ts
+const awaitApproval = async () => {
+	if (!await consola.prompt(`Accept incoming request?`, { type: "confirm" })) throw new HTTPError("Request rejected", 403, JSON.stringify({ message: "Request rejected" }));
+};
+
 //#endregion
 //#region src/lib/tokenizer.ts
 const ENCODING_MAP = {
@@ -1085,6 +1634,229 @@ const getTokenCount = async (payload, model) => {
 	};
 };
 
+//#endregion
+//#region src/lib/auto-compact.ts
+const DEFAULT_CONFIG = {
+	targetTokens: 1e5,
+	safetyMarginPercent: 10
+};
+/**
+* Check if payload needs compaction based on model limits.
+* Uses a safety margin to account for token counting differences.
+*/
+async function checkNeedsCompaction(payload, model, safetyMarginPercent = 10) {
+	const currentTokens = (await getTokenCount(payload, model)).input;
+	const rawLimit = model.capabilities.limits.max_prompt_tokens ?? 128e3;
+	const limit = Math.floor(rawLimit * (1 - safetyMarginPercent / 100));
+	return {
+		needed: currentTokens > limit,
+		currentTokens,
+		limit
+	};
+}
+/**
+* Calculate approximate token count for a single message.
+* This is a fast estimation for splitting decisions.
+*/
+function estimateMessageTokens(message) {
+	let text = "";
+	if (typeof message.content === "string") text = message.content;
+	else if (Array.isArray(message.content)) {
+		for (const part of message.content) if (part.type === "text") text += part.text;
+		else if ("image_url" in part) text += part.image_url.url;
+	}
+	if (message.tool_calls) text += JSON.stringify(message.tool_calls);
+	return Math.ceil(text.length / 4) + 10;
+}
+/**
+* Extract system messages from the beginning of the message list.
+*/
+function extractSystemMessages(messages) {
+	const systemMessages = [];
+	let i = 0;
+	while (i < messages.length) {
+		const msg = messages[i];
+		if (msg.role === "system" || msg.role === "developer") {
+			systemMessages.push(msg);
+			i++;
+		} else break;
+	}
+	return {
+		systemMessages,
+		remainingMessages: messages.slice(i)
+	};
+}
+/**
+* Find messages to keep from the end to stay under target tokens.
+* Returns the starting index of messages to preserve.
+*/
+function findPreserveIndex(messages, targetTokens, systemTokens) {
+	const availableTokens = targetTokens - systemTokens - 500;
+	let accumulatedTokens = 0;
+	for (let i = messages.length - 1; i >= 0; i--) {
+		const msgTokens = estimateMessageTokens(messages[i]);
+		if (accumulatedTokens + msgTokens > availableTokens) return i + 1;
+		accumulatedTokens += msgTokens;
+	}
+	return 0;
+}
+/**
+* Calculate estimated tokens for system messages.
+*/
+function estimateSystemTokens(systemMessages) {
+	return systemMessages.reduce((sum, msg) => sum + estimateMessageTokens(msg), 0);
+}
+/**
+* Create a truncation marker message.
+*/
+function createTruncationMarker(removedCount) {
+	return {
+		role: "user",
+		content: `[CONTEXT TRUNCATED: ${removedCount} earlier messages were removed to fit context limits. The conversation continues below.]`
+	};
+}
+/**
+* Perform auto-compaction on a payload that exceeds token limits.
+* This uses simple truncation - no LLM calls required.
+*/
+async function autoCompact(payload, model, config = {}) {
+	const cfg = {
+		...DEFAULT_CONFIG,
+		...config
+	};
+	const originalTokens = (await getTokenCount(payload, model)).input;
+	const rawLimit = model.capabilities.limits.max_prompt_tokens ?? 128e3;
+	const limit = Math.floor(rawLimit * (1 - cfg.safetyMarginPercent / 100));
+	if (originalTokens <= limit) return {
+		payload,
+		wasCompacted: false,
+		originalTokens,
+		compactedTokens: originalTokens,
+		removedMessageCount: 0
+	};
+	consola.info(`Auto-compact: ${originalTokens} tokens exceeds limit of ${limit}, truncating...`);
+	const { systemMessages, remainingMessages } = extractSystemMessages(payload.messages);
+	const systemTokens = estimateSystemTokens(systemMessages);
+	consola.debug(`Auto-compact: ${systemMessages.length} system messages (~${systemTokens} tokens)`);
+	const effectiveTarget = Math.min(cfg.targetTokens, limit);
+	const preserveIndex = findPreserveIndex(remainingMessages, effectiveTarget, systemTokens);
+	if (preserveIndex === 0) {
+		consola.warn("Auto-compact: Cannot truncate further without losing all conversation history");
+		return {
+			payload,
+			wasCompacted: false,
+			originalTokens,
+			compactedTokens: originalTokens,
+			removedMessageCount: 0
+		};
+	}
+	const removedMessages = remainingMessages.slice(0, preserveIndex);
+	const preservedMessages = remainingMessages.slice(preserveIndex);
+	consola.info(`Auto-compact: Removing ${removedMessages.length} messages, keeping ${preservedMessages.length}`);
+	const truncationMarker = createTruncationMarker(removedMessages.length);
+	const newPayload = {
+		...payload,
+		messages: [
+			...systemMessages,
+			truncationMarker,
+			...preservedMessages
+		]
+	};
+	const newTokenCount = await getTokenCount(newPayload, model);
+	consola.info(`Auto-compact: Reduced from ${originalTokens} to ${newTokenCount.input} tokens`);
+	if (newTokenCount.input > limit) {
+		consola.warn(`Auto-compact: Still over limit (${newTokenCount.input} > ${limit}), trying more aggressive truncation`);
+		const aggressiveTarget = Math.floor(effectiveTarget * .7);
+		if (aggressiveTarget < 2e4) {
+			consola.error("Auto-compact: Cannot reduce further, target too low");
+			return {
+				payload: newPayload,
+				wasCompacted: true,
+				originalTokens,
+				compactedTokens: newTokenCount.input,
+				removedMessageCount: removedMessages.length
+			};
+		}
+		return autoCompact(payload, model, {
+			...cfg,
+			targetTokens: aggressiveTarget
+		});
+	}
+	return {
+		payload: newPayload,
+		wasCompacted: true,
+		originalTokens,
+		compactedTokens: newTokenCount.input,
+		removedMessageCount: removedMessages.length
+	};
+}
+/**
+* Create a marker to append to responses indicating auto-compaction occurred.
+*/
+function createCompactionMarker(result) {
+	if (!result.wasCompacted) return "";
+	const reduction = result.originalTokens - result.compactedTokens;
+	const percentage = Math.round(reduction / result.originalTokens * 100);
+	return `\n\n---\n[Auto-compacted: ${result.removedMessageCount} messages removed, ${result.originalTokens} → ${result.compactedTokens} tokens (${percentage}% reduction)]`;
+}
+
+//#endregion
+//#region src/lib/queue.ts
+var RequestQueue = class {
+	queue = [];
+	processing = false;
+	lastRequestTime = 0;
+	async enqueue(execute, rateLimitSeconds) {
+		return new Promise((resolve, reject) => {
+			this.queue.push({
+				execute,
+				resolve,
+				reject
+			});
+			if (this.queue.length > 1) {
+				const waitTime = Math.ceil((this.queue.length - 1) * rateLimitSeconds);
+				consola.info(`Request queued. Position: ${this.queue.length}, estimated wait: ${waitTime}s`);
+			}
+			this.processQueue(rateLimitSeconds);
+		});
+	}
+	async processQueue(rateLimitSeconds) {
+		if (this.processing) return;
+		this.processing = true;
+		while (this.queue.length > 0) {
+			const elapsedMs = Date.now() - this.lastRequestTime;
+			const requiredMs = rateLimitSeconds * 1e3;
+			if (this.lastRequestTime > 0 && elapsedMs < requiredMs) {
+				const waitMs = requiredMs - elapsedMs;
+				consola.debug(`Rate limit: waiting ${Math.ceil(waitMs / 1e3)}s`);
+				await new Promise((resolve) => setTimeout(resolve, waitMs));
+			}
+			const request = this.queue.shift();
+			if (!request) break;
+			this.lastRequestTime = Date.now();
+			try {
+				const result = await request.execute();
+				request.resolve(result);
+			} catch (error) {
+				request.reject(error);
+			}
+		}
+		this.processing = false;
+	}
+	get length() {
+		return this.queue.length;
+	}
+};
+const requestQueue = new RequestQueue();
+/**
+* Execute a request with rate limiting via queue.
+* Requests are queued and processed sequentially at the configured rate.
+*/
+async function executeWithRateLimit(state$1, execute) {
+	if (state$1.rateLimitSeconds === void 0) return execute();
+	return requestQueue.enqueue(execute, state$1.rateLimitSeconds);
+}
+
 //#endregion
 //#region src/services/copilot/create-chat-completions.ts
 const createChatCompletions = async (payload) => {
@@ -1112,20 +1884,83 @@ const createChatCompletions = async (payload) => {
 //#region src/routes/chat-completions/handler.ts
 async function handleCompletion$1(c) {
 	const startTime = Date.now();
-	let payload = await c.req.json();
-	consola.debug("Request payload:", JSON.stringify(payload).slice(-400));
-	const historyId = recordRequest("openai", {
-		model: payload.model,
-		messages: convertOpenAIMessages(payload.messages),
-		stream: payload.stream ?? false,
-		tools: payload.tools?.map((t) => ({
-			name: t.function.name,
-			description: t.function.description
-		})),
-		max_tokens: payload.max_tokens ?? void 0,
-		temperature: payload.temperature ?? void 0
-	});
-	const selectedModel = state.models?.data.find((model) => model.id === payload.model);
+	const originalPayload = await c.req.json();
+	consola.debug("Request payload:", JSON.stringify(originalPayload).slice(-400));
+	const trackingId = c.get("trackingId");
+	updateTrackerModel$1(trackingId, originalPayload.model);
+	const ctx = {
+		historyId: recordRequest("openai", {
+			model: originalPayload.model,
+			messages: convertOpenAIMessages(originalPayload.messages),
+			stream: originalPayload.stream ?? false,
+			tools: originalPayload.tools?.map((t) => ({
+				name: t.function.name,
+				description: t.function.description
+			})),
+			max_tokens: originalPayload.max_tokens ?? void 0,
+			temperature: originalPayload.temperature ?? void 0
+		}),
+		trackingId,
+		startTime
+	};
+	const selectedModel = state.models?.data.find((model) => model.id === originalPayload.model);
+	await logTokenCount(originalPayload, selectedModel);
+	const { finalPayload, compactResult } = await buildFinalPayload$1(originalPayload, selectedModel);
+	if (compactResult) ctx.compactResult = compactResult;
+	const payload = isNullish(finalPayload.max_tokens) ? {
+		...finalPayload,
+		max_tokens: selectedModel?.capabilities.limits.max_output_tokens
+	} : finalPayload;
+	if (isNullish(originalPayload.max_tokens)) consola.debug("Set max_tokens to:", JSON.stringify(payload.max_tokens));
+	if (state.manualApprove) await awaitApproval();
+	try {
+		const response = await executeWithRateLimit(state, () => createChatCompletions(payload));
+		if (isNonStreaming$1(response)) return handleNonStreamingResponse$1(c, response, ctx);
+		consola.debug("Streaming response");
+		updateTrackerStatus$1(trackingId, "streaming");
+		return streamSSE(c, async (stream) => {
+			await handleStreamingResponse$1({
+				stream,
+				response,
+				payload,
+				ctx
+			});
+		});
+	} catch (error) {
+		recordErrorResponse$1(ctx, payload.model, error);
+		throw error;
+	}
+}
+async function buildFinalPayload$1(payload, model) {
+	if (!state.autoCompact || !model) {
+		if (state.autoCompact && !model) consola.warn(`Auto-compact: Model '${payload.model}' not found in cached models, skipping`);
+		return {
+			finalPayload: payload,
+			compactResult: null
+		};
+	}
+	try {
+		const check = await checkNeedsCompaction(payload, model);
+		consola.info(`Auto-compact check: ${check.currentTokens} tokens, limit ${check.limit}, needed: ${check.needed}`);
+		if (!check.needed) return {
+			finalPayload: payload,
+			compactResult: null
+		};
+		consola.info(`Auto-compact triggered: ${check.currentTokens} tokens > ${check.limit} limit`);
+		const compactResult = await autoCompact(payload, model);
+		return {
+			finalPayload: compactResult.payload,
+			compactResult
+		};
+	} catch (error) {
+		consola.warn("Auto-compact failed, proceeding with original payload:", error);
+		return {
+			finalPayload: payload,
+			compactResult: null
+		};
+	}
+}
+async function logTokenCount(payload, selectedModel) {
 	try {
 		if (selectedModel) {
 			const tokenCount = await getTokenCount(payload, selectedModel);
@@ -1134,146 +1969,236 @@ async function handleCompletion$1(c) {
 	} catch (error) {
 		consola.warn("Failed to calculate token count:", error);
 	}
-	if (state.manualApprove) await awaitApproval();
-	if (isNullish(payload.max_tokens)) {
-		payload = {
-			...payload,
-			max_tokens: selectedModel?.capabilities.limits.max_output_tokens
+}
+function updateTrackerModel$1(trackingId, model) {
+	if (!trackingId) return;
+	const request = requestTracker.getRequest(trackingId);
+	if (request) request.model = model;
+}
+function updateTrackerStatus$1(trackingId, status) {
+	if (!trackingId) return;
+	requestTracker.updateRequest(trackingId, { status });
+}
+function recordErrorResponse$1(ctx, model, error) {
+	recordResponse(ctx.historyId, {
+		success: false,
+		model,
+		usage: {
+			input_tokens: 0,
+			output_tokens: 0
+		},
+		error: error instanceof Error ? error.message : "Unknown error",
+		content: null
+	}, Date.now() - ctx.startTime);
+}
+function handleNonStreamingResponse$1(c, originalResponse, ctx) {
+	consola.debug("Non-streaming response:", JSON.stringify(originalResponse));
+	let response = originalResponse;
+	if (ctx.compactResult?.wasCompacted && response.choices[0]?.message.content) {
+		const marker = createCompactionMarker(ctx.compactResult);
+		response = {
+			...response,
+			choices: response.choices.map((choice$1, i) => i === 0 ? {
+				...choice$1,
+				message: {
+					...choice$1.message,
+					content: (choice$1.message.content ?? "") + marker
+				}
+			} : choice$1)
 		};
-		consola.debug("Set max_tokens to:", JSON.stringify(payload.max_tokens));
 	}
+	const choice = response.choices[0];
+	const usage = response.usage;
+	recordResponse(ctx.historyId, {
+		success: true,
+		model: response.model,
+		usage: {
+			input_tokens: usage?.prompt_tokens ?? 0,
+			output_tokens: usage?.completion_tokens ?? 0
+		},
+		stop_reason: choice.finish_reason,
+		content: buildResponseContent(choice),
+		toolCalls: extractToolCalls(choice)
+	}, Date.now() - ctx.startTime);
+	if (ctx.trackingId && usage) requestTracker.updateRequest(ctx.trackingId, {
+		inputTokens: usage.prompt_tokens,
+		outputTokens: usage.completion_tokens
+	});
+	return c.json(response);
+}
+function buildResponseContent(choice) {
+	return {
+		role: choice.message.role,
+		content: typeof choice.message.content === "string" ? choice.message.content : JSON.stringify(choice.message.content),
+		tool_calls: choice.message.tool_calls?.map((tc) => ({
+			id: tc.id,
+			type: tc.type,
+			function: {
+				name: tc.function.name,
+				arguments: tc.function.arguments
+			}
+		}))
+	};
+}
+function extractToolCalls(choice) {
+	return choice.message.tool_calls?.map((tc) => ({
+		id: tc.id,
+		name: tc.function.name,
+		input: tc.function.arguments
+	}));
+}
+function createStreamAccumulator() {
+	return {
+		model: "",
+		inputTokens: 0,
+		outputTokens: 0,
+		finishReason: "",
+		content: "",
+		toolCalls: [],
+		toolCallMap: /* @__PURE__ */ new Map()
+	};
+}
+async function handleStreamingResponse$1(opts) {
+	const { stream, response, payload, ctx } = opts;
+	const acc = createStreamAccumulator();
 	try {
-		const response = await executeWithRateLimit(state, () => createChatCompletions(payload));
-		if (isNonStreaming$1(response)) {
-			consola.debug("Non-streaming response:", JSON.stringify(response));
-			const choice = response.choices[0];
-			recordResponse(historyId, {
-				success: true,
-				model: response.model,
-				usage: {
-					input_tokens: response.usage?.prompt_tokens ?? 0,
-					output_tokens: response.usage?.completion_tokens ?? 0
-				},
-				stop_reason: choice?.finish_reason ?? void 0,
-				content: choice?.message ? {
-					role: choice.message.role,
-					content: typeof choice.message.content === "string" ? choice.message.content : JSON.stringify(choice.message.content),
-					tool_calls: choice.message.tool_calls?.map((tc) => ({
-						id: tc.id,
-						type: tc.type,
-						function: {
-							name: tc.function.name,
-							arguments: tc.function.arguments
-						}
-					}))
-				} : null,
-				toolCalls: choice?.message?.tool_calls?.map((tc) => ({
-					id: tc.id,
-					name: tc.function.name,
-					input: tc.function.arguments
-				}))
-			}, Date.now() - startTime);
-			return c.json(response);
+		for await (const chunk of response) {
+			consola.debug("Streaming chunk:", JSON.stringify(chunk));
+			parseStreamChunk(chunk, acc);
+			await stream.writeSSE(chunk);
 		}
-		consola.debug("Streaming response");
-		return streamSSE(c, async (stream) => {
-			let streamModel = "";
-			let streamInputTokens = 0;
-			let streamOutputTokens = 0;
-			let streamFinishReason = "";
-			let streamContent = "";
-			const streamToolCalls = [];
-			const toolCallAccumulators = /* @__PURE__ */ new Map();
-			try {
-				for await (const chunk of response) {
-					consola.debug("Streaming chunk:", JSON.stringify(chunk));
-					if (chunk.data && chunk.data !== "[DONE]") try {
-						const parsed = JSON.parse(chunk.data);
-						if (parsed.model && !streamModel) streamModel = parsed.model;
-						if (parsed.usage) {
-							streamInputTokens = parsed.usage.prompt_tokens;
-							streamOutputTokens = parsed.usage.completion_tokens;
-						}
-						const choice = parsed.choices[0];
-						if (choice?.delta?.content) streamContent += choice.delta.content;
-						if (choice?.delta?.tool_calls) for (const tc of choice.delta.tool_calls) {
-							const idx = tc.index;
-							if (!toolCallAccumulators.has(idx)) toolCallAccumulators.set(idx, {
-								id: tc.id || "",
-								name: tc.function?.name || "",
-								arguments: ""
-							});
-							const acc = toolCallAccumulators.get(idx);
-							if (acc) {
-								if (tc.id) acc.id = tc.id;
-								if (tc.function?.name) acc.name = tc.function.name;
-								if (tc.function?.arguments) acc.arguments += tc.function.arguments;
-							}
-						}
-						if (choice?.finish_reason) streamFinishReason = choice.finish_reason;
-					} catch {}
-					await stream.writeSSE(chunk);
-				}
-				for (const tc of toolCallAccumulators.values()) if (tc.id && tc.name) streamToolCalls.push({
-					id: tc.id,
-					name: tc.name,
-					arguments: tc.arguments
-				});
-				const toolCallsForContent = streamToolCalls.map((tc) => ({
-					id: tc.id,
-					type: "function",
-					function: {
-						name: tc.name,
-						arguments: tc.arguments
-					}
-				}));
-				recordResponse(historyId, {
-					success: true,
-					model: streamModel || payload.model,
-					usage: {
-						input_tokens: streamInputTokens,
-						output_tokens: streamOutputTokens
-					},
-					stop_reason: streamFinishReason || void 0,
-					content: {
-						role: "assistant",
-						content: streamContent || void 0,
-						tool_calls: toolCallsForContent.length > 0 ? toolCallsForContent : void 0
-					},
-					toolCalls: streamToolCalls.length > 0 ? streamToolCalls.map((tc) => ({
-						id: tc.id,
-						name: tc.name,
-						input: tc.arguments
-					})) : void 0
-				}, Date.now() - startTime);
-			} catch (error) {
-				recordResponse(historyId, {
-					success: false,
-					model: streamModel || payload.model,
-					usage: {
-						input_tokens: 0,
-						output_tokens: 0
-					},
-					error: error instanceof Error ? error.message : "Stream error",
-					content: null
-				}, Date.now() - startTime);
-				throw error;
-			}
-		});
+		if (ctx.compactResult?.wasCompacted) {
+			const marker = createCompactionMarker(ctx.compactResult);
+			const markerChunk = {
+				id: `compact-marker-${Date.now()}`,
+				object: "chat.completion.chunk",
+				created: Math.floor(Date.now() / 1e3),
+				model: acc.model || payload.model,
+				choices: [{
+					index: 0,
+					delta: { content: marker },
+					finish_reason: null,
+					logprobs: null
+				}]
+			};
+			await stream.writeSSE({
+				data: JSON.stringify(markerChunk),
+				event: "message"
+			});
+			acc.content += marker;
+		}
+		recordStreamSuccess(acc, payload.model, ctx);
+		completeTracking$1(ctx.trackingId, acc.inputTokens, acc.outputTokens);
 	} catch (error) {
-		recordResponse(historyId, {
-			success: false,
-			model: payload.model,
-			usage: {
-				input_tokens: 0,
-				output_tokens: 0
-			},
-			error: error instanceof Error ? error.message : "Unknown error",
-			content: null
-		}, Date.now() - startTime);
+		recordStreamError({
+			acc,
+			fallbackModel: payload.model,
+			ctx,
+			error
+		});
+		failTracking$1(ctx.trackingId, error);
 		throw error;
 	}
 }
+function parseStreamChunk(chunk, acc) {
+	if (!chunk.data || chunk.data === "[DONE]") return;
+	try {
+		const parsed = JSON.parse(chunk.data);
+		accumulateModel(parsed, acc);
+		accumulateUsage(parsed, acc);
+		accumulateChoice(parsed.choices[0], acc);
+	} catch {}
+}
+function accumulateModel(parsed, acc) {
+	if (parsed.model && !acc.model) acc.model = parsed.model;
+}
+function accumulateUsage(parsed, acc) {
+	if (parsed.usage) {
+		acc.inputTokens = parsed.usage.prompt_tokens;
+		acc.outputTokens = parsed.usage.completion_tokens;
+	}
+}
+function accumulateChoice(choice, acc) {
+	if (!choice) return;
+	if (choice.delta.content) acc.content += choice.delta.content;
+	if (choice.delta.tool_calls) accumulateToolCalls(choice.delta.tool_calls, acc);
+	if (choice.finish_reason) acc.finishReason = choice.finish_reason;
+}
+function accumulateToolCalls(toolCalls, acc) {
+	if (!toolCalls) return;
+	for (const tc of toolCalls) {
+		const idx = tc.index;
+		if (!acc.toolCallMap.has(idx)) acc.toolCallMap.set(idx, {
+			id: tc.id ?? "",
+			name: tc.function?.name ?? "",
+			arguments: ""
+		});
+		const item = acc.toolCallMap.get(idx);
+		if (item) {
+			if (tc.id) item.id = tc.id;
+			if (tc.function?.name) item.name = tc.function.name;
+			if (tc.function?.arguments) item.arguments += tc.function.arguments;
+		}
+	}
+}
+function recordStreamSuccess(acc, fallbackModel, ctx) {
+	for (const tc of acc.toolCallMap.values()) if (tc.id && tc.name) acc.toolCalls.push(tc);
+	const toolCalls = acc.toolCalls.map((tc) => ({
+		id: tc.id,
+		type: "function",
+		function: {
+			name: tc.name,
+			arguments: tc.arguments
+		}
+	}));
+	recordResponse(ctx.historyId, {
+		success: true,
+		model: acc.model || fallbackModel,
+		usage: {
+			input_tokens: acc.inputTokens,
+			output_tokens: acc.outputTokens
+		},
+		stop_reason: acc.finishReason || void 0,
+		content: {
+			role: "assistant",
+			content: acc.content,
+			tool_calls: toolCalls.length > 0 ? toolCalls : void 0
+		},
+		toolCalls: acc.toolCalls.length > 0 ? acc.toolCalls.map((tc) => ({
+			id: tc.id,
+			name: tc.name,
+			input: tc.arguments
+		})) : void 0
+	}, Date.now() - ctx.startTime);
+}
+function recordStreamError(opts) {
+	const { acc, fallbackModel, ctx, error } = opts;
+	recordResponse(ctx.historyId, {
+		success: false,
+		model: acc.model || fallbackModel,
+		usage: {
+			input_tokens: 0,
+			output_tokens: 0
+		},
+		error: error instanceof Error ? error.message : "Stream error",
+		content: null
+	}, Date.now() - ctx.startTime);
+}
+function completeTracking$1(trackingId, inputTokens, outputTokens) {
+	if (!trackingId) return;
+	requestTracker.updateRequest(trackingId, {
+		inputTokens,
+		outputTokens
+	});
+	requestTracker.completeRequest(trackingId, 200, {
+		inputTokens,
+		outputTokens
+	});
+}
+function failTracking$1(trackingId, error) {
+	if (!trackingId) return;
+	requestTracker.failRequest(trackingId, error instanceof Error ? error.message : "Stream error");
+}
 const isNonStreaming$1 = (response) => Object.hasOwn(response, "choices");
 function convertOpenAIMessages(messages) {
 	return messages.map((msg) => {
@@ -1461,6 +2386,78 @@ function getContentText(content) {
   return JSON.stringify(content, null, 2);
 }
 
+// Extract real user text, skipping system tags like <system-reminder>, <ide_opened_file>, etc.
+function extractRealUserText(content) {
+  if (!content) return '';
+  let text = '';
+  if (typeof content === 'string') {
+    text = content;
+  } else if (Array.isArray(content)) {
+    text = content
+      .filter(c => c.type === 'text' && c.text)
+      .map(c => c.text)
+      .join('\\n');
+  }
+  if (!text) return '';
+
+  // Remove system tags and their content
+  const systemTags = [
+    'system-reminder',
+    'ide_opened_file',
+    'ide_selection',
+    'ide_visible_files',
+    'ide_diagnostics',
+    'ide_cursor_position',
+    'user-prompt-submit-hook',
+    'antml:function_calls',
+    'antml:invoke',
+    'antml:parameter'
+  ];
+
+  let cleaned = text;
+  for (const tag of systemTags) {
+    // Remove <tag>...</tag> blocks (including multiline)
+    const regex = new RegExp('<' + tag + '[^>]*>[\\\\s\\\\S]*?</' + tag + '>', 'gi');
+    cleaned = cleaned.replace(regex, '');
+    // Remove self-closing <tag ... /> or <tag ...>content without closing
+    const selfClosingRegex = new RegExp('<' + tag + '[^>]*/>', 'gi');
+    cleaned = cleaned.replace(selfClosingRegex, '');
+  }
+
+  // Trim whitespace and return
+  return cleaned.trim();
+}
+
+// Get preview text from assistant message content
+function getAssistantPreview(content) {
+  if (!content) return '';
+  if (typeof content === 'string') {
+    const text = content.trim();
+    if (text.length > 0) {
+      return text.length > 80 ? text.slice(0, 80) + '...' : text;
+    }
+    return '';
+  }
+  if (Array.isArray(content)) {
+    // First try to get text content
+    const textParts = content.filter(c => c.type === 'text' && c.text).map(c => c.text);
+    if (textParts.length > 0) {
+      const text = textParts.join('\\n').trim();
+      if (text.length > 0) {
+        return text.length > 80 ? text.slice(0, 80) + '...' : text;
+      }
+    }
+    // If no text, show tool_use info
+    const toolUses = content.filter(c => c.type === 'tool_use');
+    if (toolUses.length === 1) {
+      return '[tool_use: ' + toolUses[0].name + ']';
+    } else if (toolUses.length > 1) {
+      return '[' + toolUses.length + ' tool_uses]';
+    }
+  }
+  return '';
+}
+
 function formatContentForDisplay(content) {
   if (!content) return { summary: '', raw: 'null' };
   if (typeof content === 'string') return { summary: content, raw: JSON.stringify(content) };
@@ -1516,6 +2513,7 @@ async function loadSessions() {
     for (const s of data.sessions) {
       const isActive = currentSessionId === s.id;
       const shortId = s.id.slice(0, 8);
+      const toolCount = s.toolsUsed ? s.toolsUsed.length : 0;
       html += \`
         <div class="session-item\${isActive ? ' active' : ''}" onclick="selectSession('\${s.id}')">
           <div class="session-meta">
@@ -1526,6 +2524,7 @@ async function loadSessions() {
             <span style="color:var(--text-dim);font-family:monospace;font-size:10px;">\${shortId}</span>
             <span>\${s.requestCount} req</span>
             <span>\${formatNumber(s.totalInputTokens + s.totalOutputTokens)} tok</span>
+            \${toolCount > 0 ? '<span class="badge tool">' + toolCount + ' tools</span>' : ''}
             <span class="badge \${s.endpoint}">\${s.endpoint}</span>
           </div>
         </div>
@@ -1584,6 +2583,37 @@ async function loadEntries() {
       const tokens = e.response ? formatNumber(e.response.usage.input_tokens) + '/' + formatNumber(e.response.usage.output_tokens) : '-';
       const shortId = e.id.slice(0, 8);
 
+      // Get preview: show meaningful context about the request
+      let lastUserMsg = '';
+      const messages = e.request.messages;
+      const lastMsg = messages[messages.length - 1];
+
+      // If last message is tool_result, look at the previous assistant message for context
+      if (lastMsg && lastMsg.role === 'user') {
+        const content = lastMsg.content;
+        if (Array.isArray(content) && content.length > 0 && content[0].type === 'tool_result') {
+          // This is a tool_result response - look for previous assistant message
+          const prevMsg = messages.length >= 2 ? messages[messages.length - 2] : null;
+          if (prevMsg && prevMsg.role === 'assistant') {
+            lastUserMsg = getAssistantPreview(prevMsg.content);
+          }
+          // If no meaningful preview from assistant, show tool_result count
+          if (!lastUserMsg) {
+            const toolResults = content.filter(c => c.type === 'tool_result');
+            lastUserMsg = '[' + toolResults.length + ' tool_result' + (toolResults.length > 1 ? 's' : '') + ']';
+          }
+        } else {
+          // Regular user message, extract real text
+          const realText = extractRealUserText(lastMsg.content);
+          if (realText.length > 0) {
+            lastUserMsg = realText.slice(0, 80);
+            if (realText.length > 80) lastUserMsg += '...';
+          }
+        }
+      } else if (lastMsg && lastMsg.role === 'assistant') {
+        lastUserMsg = getAssistantPreview(lastMsg.content);
+      }
+
       html += \`
         <div class="entry-item\${isSelected ? ' selected' : ''}" onclick="showDetail('\${e.id}')">
           <div class="entry-header">
@@ -1596,6 +2626,7 @@ async function loadEntries() {
             <span class="entry-tokens">\${tokens}</span>
             <span class="entry-duration">\${formatDuration(e.durationMs)}</span>
           </div>
+          \${lastUserMsg ? '<div class="entry-preview">' + escapeHtml(lastUserMsg) + '</div>' : ''}
         </div>
       \`;
     }
@@ -1655,7 +2686,7 @@ async function showDetail(id) {
             <div class="info-item"><div class="info-label">Duration</div><div class="info-value">\${formatDuration(entry.durationMs)}</div></div>
             <div class="info-item"><div class="info-label">Stop Reason</div><div class="info-value">\${entry.response.stop_reason || '-'}</div></div>
           </div>
-          \${entry.response.error ? '<div style="color:var(--error);margin-top:8px;">Error: ' + entry.response.error + '</div>' : ''}
+          \${entry.response.error ? '<div class="error-detail"><div class="error-label">Error Details</div><pre class="error-content">' + escapeHtml(entry.response.error) + '</pre></div>' : ''}
         </div>
       \`;
     }
@@ -2023,6 +3054,14 @@ input::placeholder { color: var(--text-dim); }
 .entry-model { font-weight: 500; flex: 1; }
 .entry-tokens { font-size: 11px; color: var(--text-dim); }
 .entry-duration { font-size: 11px; color: var(--text-dim); min-width: 50px; text-align: right; }
+.entry-preview {
+  padding: 0 16px 8px 16px;
+  font-size: 11px;
+  color: var(--text-muted);
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
 
 /* Badges */
 .badge {
@@ -2038,6 +3077,7 @@ input::placeholder { color: var(--text-dim); }
 .badge.anthropic { background: rgba(163, 113, 247, 0.15); color: var(--purple); }
 .badge.openai { background: rgba(210, 153, 34, 0.15); color: var(--warning); }
 .badge.stream { background: rgba(57, 197, 207, 0.15); color: var(--cyan); }
+.badge.tool { background: rgba(88, 166, 255, 0.15); color: var(--primary); }
 
 /* Detail panel */
 .detail-panel {
@@ -2133,6 +3173,32 @@ input::placeholder { color: var(--text-dim); }
 .info-label { font-size: 11px; color: var(--text-muted); }
 .info-value { font-weight: 500; }
 
+/* Error detail display */
+.error-detail {
+  margin-top: 12px;
+  padding: 12px;
+  background: rgba(248, 81, 73, 0.1);
+  border: 1px solid rgba(248, 81, 73, 0.3);
+  border-radius: 6px;
+}
+.error-label {
+  font-size: 11px;
+  color: var(--error);
+  font-weight: 600;
+  margin-bottom: 8px;
+  text-transform: uppercase;
+}
+.error-content {
+  margin: 0;
+  font-family: 'SF Mono', Monaco, 'Courier New', monospace;
+  font-size: 12px;
+  color: var(--error);
+  white-space: pre-wrap;
+  word-break: break-word;
+  max-height: 300px;
+  overflow-y: auto;
+}
+
 /* Empty state */
 .empty-state {
   text-align: center;
@@ -2388,12 +3454,12 @@ function translateModelName(model) {
 		haiku: "claude-haiku-4.5"
 	};
 	if (shortNameMap[model]) return shortNameMap[model];
-	if (model.match(/^claude-sonnet-4-5-\d+$/)) return "claude-sonnet-4.5";
-	if (model.match(/^claude-sonnet-4-\d+$/)) return "claude-sonnet-4";
-	if (model.match(/^claude-opus-4-5-\d+$/)) return "claude-opus-4.5";
-	if (model.match(/^claude-opus-4-\d+$/)) return "claude-opus-4.5";
-	if (model.match(/^claude-haiku-4-5-\d+$/)) return "claude-haiku-4.5";
-	if (model.match(/^claude-haiku-3-5-\d+$/)) return "claude-haiku-4.5";
+	if (/^claude-sonnet-4-5-\d+$/.test(model)) return "claude-sonnet-4.5";
+	if (/^claude-sonnet-4-\d+$/.test(model)) return "claude-sonnet-4";
+	if (/^claude-opus-4-5-\d+$/.test(model)) return "claude-opus-4.5";
+	if (/^claude-opus-4-\d+$/.test(model)) return "claude-opus-4.5";
+	if (/^claude-haiku-4-5-\d+$/.test(model)) return "claude-haiku-4.5";
+	if (/^claude-haiku-3-5-\d+$/.test(model)) return "claude-haiku-4.5";
 	return model;
 }
 function translateAnthropicMessagesToOpenAI(anthropicMessages, system, toolNameMapping) {
@@ -2490,7 +3556,7 @@ function getTruncatedToolName(originalName, toolNameMapping) {
 	if (existingTruncated) return existingTruncated;
 	let hash = 0;
 	for (let i = 0; i < originalName.length; i++) {
-		const char = originalName.charCodeAt(i);
+		const char = originalName.codePointAt(i) ?? 0;
 		hash = (hash << 5) - hash + char;
 		hash = hash & hash;
 	}
@@ -2527,8 +3593,9 @@ function translateAnthropicToolChoiceToOpenAI(anthropicToolChoice, toolNameMappi
 		default: return;
 	}
 }
-function translateToAnthropic(response, toolNameMapping) {
-	if (response.choices.length === 0) return {
+/** Create empty response for edge case of no choices */
+function createEmptyResponse(response) {
+	return {
 		id: response.id,
 		type: "message",
 		role: "assistant",
@@ -2541,6 +3608,18 @@ function translateToAnthropic(response, toolNameMapping) {
 			output_tokens: response.usage?.completion_tokens ?? 0
 		}
 	};
+}
+/** Build usage object from response */
+function buildUsageObject(response) {
+	const cachedTokens = response.usage?.prompt_tokens_details?.cached_tokens;
+	return {
+		input_tokens: (response.usage?.prompt_tokens ?? 0) - (cachedTokens ?? 0),
+		output_tokens: response.usage?.completion_tokens ?? 0,
+		...cachedTokens !== void 0 && { cache_read_input_tokens: cachedTokens }
+	};
+}
+function translateToAnthropic(response, toolNameMapping) {
+	if (response.choices.length === 0) return createEmptyResponse(response);
 	const allTextBlocks = [];
 	const allToolUseBlocks = [];
 	let stopReason = null;
@@ -2560,11 +3639,7 @@ function translateToAnthropic(response, toolNameMapping) {
 		content: [...allTextBlocks, ...allToolUseBlocks],
 		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 }
-		}
+		usage: buildUsageObject(response)
 	};
 }
 function getAnthropicTextBlocks(messageContent) {
@@ -2776,175 +3851,365 @@ async function handleCompletion(c) {
 	const startTime = Date.now();
 	const anthropicPayload = await c.req.json();
 	consola.debug("Anthropic request payload:", JSON.stringify(anthropicPayload));
-	const historyId = recordRequest("anthropic", {
-		model: anthropicPayload.model,
-		messages: convertAnthropicMessages(anthropicPayload.messages),
-		stream: anthropicPayload.stream ?? false,
-		tools: anthropicPayload.tools?.map((t) => ({
-			name: t.name,
-			description: t.description
-		})),
-		max_tokens: anthropicPayload.max_tokens,
-		temperature: anthropicPayload.temperature,
-		system: extractSystemPrompt(anthropicPayload.system)
-	});
-	const { payload: openAIPayload, toolNameMapping } = translateToOpenAI(anthropicPayload);
-	consola.debug("Translated OpenAI request payload:", JSON.stringify(openAIPayload));
+	const trackingId = c.get("trackingId");
+	updateTrackerModel(trackingId, anthropicPayload.model);
+	const ctx = {
+		historyId: recordRequest("anthropic", {
+			model: anthropicPayload.model,
+			messages: convertAnthropicMessages(anthropicPayload.messages),
+			stream: anthropicPayload.stream ?? false,
+			tools: anthropicPayload.tools?.map((t) => ({
+				name: t.name,
+				description: t.description
+			})),
+			max_tokens: anthropicPayload.max_tokens,
+			temperature: anthropicPayload.temperature,
+			system: extractSystemPrompt(anthropicPayload.system)
+		}),
+		trackingId,
+		startTime
+	};
+	const { payload: translatedPayload, toolNameMapping } = translateToOpenAI(anthropicPayload);
+	consola.debug("Translated OpenAI request payload:", JSON.stringify(translatedPayload));
+	const selectedModel = state.models?.data.find((model) => model.id === translatedPayload.model);
+	const { finalPayload: openAIPayload, compactResult } = await buildFinalPayload(translatedPayload, selectedModel);
+	if (compactResult) ctx.compactResult = compactResult;
 	if (state.manualApprove) await awaitApproval();
 	try {
 		const response = await executeWithRateLimit(state, () => createChatCompletions(openAIPayload));
-		if (isNonStreaming(response)) {
-			consola.debug("Non-streaming response from Copilot:", JSON.stringify(response).slice(-400));
-			const anthropicResponse = translateToAnthropic(response, toolNameMapping);
-			consola.debug("Translated Anthropic response:", JSON.stringify(anthropicResponse));
-			recordResponse(historyId, {
-				success: true,
-				model: anthropicResponse.model,
-				usage: anthropicResponse.usage,
-				stop_reason: anthropicResponse.stop_reason ?? void 0,
-				content: {
-					role: "assistant",
-					content: anthropicResponse.content.map((block) => {
-						if (block.type === "text") return {
-							type: "text",
-							text: block.text
-						};
-						if (block.type === "tool_use") return {
-							type: "tool_use",
-							id: block.id,
-							name: block.name,
-							input: JSON.stringify(block.input)
-						};
-						return { type: block.type };
-					})
-				},
-				toolCalls: extractToolCallsFromContent(anthropicResponse.content)
-			}, Date.now() - startTime);
-			return c.json(anthropicResponse);
-		}
+		if (isNonStreaming(response)) return handleNonStreamingResponse({
+			c,
+			response,
+			toolNameMapping,
+			ctx
+		});
 		consola.debug("Streaming response from Copilot");
+		updateTrackerStatus(trackingId, "streaming");
 		return streamSSE(c, async (stream) => {
-			const streamState = {
-				messageStartSent: false,
-				contentBlockIndex: 0,
-				contentBlockOpen: false,
-				toolCalls: {}
-			};
-			let streamModel = "";
-			let streamInputTokens = 0;
-			let streamOutputTokens = 0;
-			let streamStopReason = "";
-			let streamContent = "";
-			const streamToolCalls = [];
-			let currentToolCall = null;
-			try {
-				for await (const rawEvent of response) {
-					consola.debug("Copilot raw stream event:", JSON.stringify(rawEvent));
-					if (rawEvent.data === "[DONE]") break;
-					if (!rawEvent.data) continue;
-					let chunk;
-					try {
-						chunk = JSON.parse(rawEvent.data);
-					} catch (parseError) {
-						consola.error("Failed to parse stream chunk:", parseError, rawEvent.data);
-						continue;
-					}
-					if (chunk.model && !streamModel) streamModel = chunk.model;
-					const events$1 = translateChunkToAnthropicEvents(chunk, streamState, toolNameMapping);
-					for (const event of events$1) {
-						consola.debug("Translated Anthropic event:", JSON.stringify(event));
-						switch (event.type) {
-							case "content_block_delta":
-								if ("text" in event.delta) streamContent += event.delta.text;
-								else if ("partial_json" in event.delta && currentToolCall) currentToolCall.input += event.delta.partial_json;
-								break;
-							case "content_block_start":
-								if (event.content_block.type === "tool_use") currentToolCall = {
-									id: event.content_block.id,
-									name: event.content_block.name,
-									input: ""
-								};
-								break;
-							case "content_block_stop":
-								if (currentToolCall) {
-									streamToolCalls.push(currentToolCall);
-									currentToolCall = null;
-								}
-								break;
-							case "message_delta":
-								if (event.delta.stop_reason) streamStopReason = event.delta.stop_reason;
-								if (event.usage) {
-									streamInputTokens = event.usage.input_tokens ?? 0;
-									streamOutputTokens = event.usage.output_tokens;
-								}
-								break;
-						}
-						await stream.writeSSE({
-							event: event.type,
-							data: JSON.stringify(event)
-						});
-					}
-				}
-				const contentBlocks = [];
-				if (streamContent) contentBlocks.push({
+			await handleStreamingResponse({
+				stream,
+				response,
+				toolNameMapping,
+				anthropicPayload,
+				ctx
+			});
+		});
+	} catch (error) {
+		recordErrorResponse(ctx, anthropicPayload.model, error);
+		throw error;
+	}
+}
+function updateTrackerModel(trackingId, model) {
+	if (!trackingId) return;
+	const request = requestTracker.getRequest(trackingId);
+	if (request) request.model = model;
+}
+async function buildFinalPayload(payload, model) {
+	if (!state.autoCompact || !model) {
+		if (state.autoCompact && !model) consola.warn(`Auto-compact: Model '${payload.model}' not found in cached models, skipping`);
+		return {
+			finalPayload: payload,
+			compactResult: null
+		};
+	}
+	try {
+		const check = await checkNeedsCompaction(payload, model);
+		consola.info(`Auto-compact check: ${check.currentTokens} tokens, limit ${check.limit}, needed: ${check.needed}`);
+		if (!check.needed) return {
+			finalPayload: payload,
+			compactResult: null
+		};
+		consola.info(`Auto-compact triggered: ${check.currentTokens} tokens > ${check.limit} limit`);
+		const compactResult = await autoCompact(payload, model);
+		return {
+			finalPayload: compactResult.payload,
+			compactResult
+		};
+	} catch (error) {
+		consola.warn("Auto-compact failed, proceeding with original payload:", error);
+		return {
+			finalPayload: payload,
+			compactResult: null
+		};
+	}
+}
+function updateTrackerStatus(trackingId, status) {
+	if (!trackingId) return;
+	requestTracker.updateRequest(trackingId, { status });
+}
+function recordErrorResponse(ctx, model, error) {
+	recordResponse(ctx.historyId, {
+		success: false,
+		model,
+		usage: {
+			input_tokens: 0,
+			output_tokens: 0
+		},
+		error: error instanceof Error ? error.message : "Unknown error",
+		content: null
+	}, Date.now() - ctx.startTime);
+}
+function handleNonStreamingResponse(opts) {
+	const { c, response, toolNameMapping, ctx } = opts;
+	consola.debug("Non-streaming response from Copilot:", JSON.stringify(response).slice(-400));
+	let anthropicResponse = translateToAnthropic(response, toolNameMapping);
+	consola.debug("Translated Anthropic response:", JSON.stringify(anthropicResponse));
+	if (ctx.compactResult?.wasCompacted) {
+		const marker = createCompactionMarker(ctx.compactResult);
+		anthropicResponse = appendMarkerToAnthropicResponse(anthropicResponse, marker);
+	}
+	recordResponse(ctx.historyId, {
+		success: true,
+		model: anthropicResponse.model,
+		usage: anthropicResponse.usage,
+		stop_reason: anthropicResponse.stop_reason ?? void 0,
+		content: {
+			role: "assistant",
+			content: anthropicResponse.content.map((block) => {
+				if (block.type === "text") return {
 					type: "text",
-					text: streamContent
-				});
-				for (const tc of streamToolCalls) contentBlocks.push({
+					text: block.text
+				};
+				if (block.type === "tool_use") return {
 					type: "tool_use",
-					...tc
-				});
-				recordResponse(historyId, {
-					success: true,
-					model: streamModel || anthropicPayload.model,
-					usage: {
-						input_tokens: streamInputTokens,
-						output_tokens: streamOutputTokens
-					},
-					stop_reason: streamStopReason || void 0,
-					content: contentBlocks.length > 0 ? {
-						role: "assistant",
-						content: contentBlocks
-					} : null,
-					toolCalls: streamToolCalls.length > 0 ? streamToolCalls.map((tc) => ({
-						id: tc.id,
-						name: tc.name,
-						input: tc.input
-					})) : void 0
-				}, Date.now() - startTime);
-			} catch (error) {
-				consola.error("Stream error:", error);
-				recordResponse(historyId, {
-					success: false,
-					model: streamModel || anthropicPayload.model,
-					usage: {
-						input_tokens: 0,
-						output_tokens: 0
-					},
-					error: error instanceof Error ? error.message : "Stream error",
-					content: null
-				}, Date.now() - startTime);
-				const errorEvent = translateErrorToAnthropicErrorEvent();
-				await stream.writeSSE({
-					event: errorEvent.type,
-					data: JSON.stringify(errorEvent)
-				});
-			}
+					id: block.id,
+					name: block.name,
+					input: JSON.stringify(block.input)
+				};
+				return { type: block.type };
+			})
+		},
+		toolCalls: extractToolCallsFromContent(anthropicResponse.content)
+	}, Date.now() - ctx.startTime);
+	if (ctx.trackingId) requestTracker.updateRequest(ctx.trackingId, {
+		inputTokens: anthropicResponse.usage.input_tokens,
+		outputTokens: anthropicResponse.usage.output_tokens
+	});
+	return c.json(anthropicResponse);
+}
+function appendMarkerToAnthropicResponse(response, marker) {
+	const content = [...response.content];
+	const lastTextIndex = content.findLastIndex((block) => block.type === "text");
+	if (lastTextIndex !== -1) {
+		const textBlock = content[lastTextIndex];
+		if (textBlock.type === "text") content[lastTextIndex] = {
+			...textBlock,
+			text: textBlock.text + marker
+		};
+	} else content.push({
+		type: "text",
+		text: marker
+	});
+	return {
+		...response,
+		content
+	};
+}
+function createAnthropicStreamAccumulator() {
+	return {
+		model: "",
+		inputTokens: 0,
+		outputTokens: 0,
+		stopReason: "",
+		content: "",
+		toolCalls: [],
+		currentToolCall: null
+	};
+}
+async function handleStreamingResponse(opts) {
+	const { stream, response, toolNameMapping, anthropicPayload, ctx } = opts;
+	const streamState = {
+		messageStartSent: false,
+		contentBlockIndex: 0,
+		contentBlockOpen: false,
+		toolCalls: {}
+	};
+	const acc = createAnthropicStreamAccumulator();
+	try {
+		await processStreamChunks({
+			stream,
+			response,
+			toolNameMapping,
+			streamState,
+			acc
 		});
+		if (ctx.compactResult?.wasCompacted) {
+			const marker = createCompactionMarker(ctx.compactResult);
+			await sendCompactionMarkerEvent(stream, streamState, marker);
+			acc.content += marker;
+		}
+		recordStreamingResponse(acc, anthropicPayload.model, ctx);
+		completeTracking(ctx.trackingId, acc.inputTokens, acc.outputTokens);
 	} catch (error) {
-		recordResponse(historyId, {
-			success: false,
-			model: anthropicPayload.model,
-			usage: {
-				input_tokens: 0,
-				output_tokens: 0
-			},
-			error: error instanceof Error ? error.message : "Unknown error",
-			content: null
-		}, Date.now() - startTime);
-		throw error;
+		consola.error("Stream error:", error);
+		recordStreamingError({
+			acc,
+			fallbackModel: anthropicPayload.model,
+			ctx,
+			error
+		});
+		failTracking(ctx.trackingId, error);
+		const errorEvent = translateErrorToAnthropicErrorEvent();
+		await stream.writeSSE({
+			event: errorEvent.type,
+			data: JSON.stringify(errorEvent)
+		});
+	}
+}
+async function sendCompactionMarkerEvent(stream, streamState, marker) {
+	const blockStartEvent = {
+		type: "content_block_start",
+		index: streamState.contentBlockIndex,
+		content_block: {
+			type: "text",
+			text: ""
+		}
+	};
+	await stream.writeSSE({
+		event: "content_block_start",
+		data: JSON.stringify(blockStartEvent)
+	});
+	const deltaEvent = {
+		type: "content_block_delta",
+		index: streamState.contentBlockIndex,
+		delta: {
+			type: "text_delta",
+			text: marker
+		}
+	};
+	await stream.writeSSE({
+		event: "content_block_delta",
+		data: JSON.stringify(deltaEvent)
+	});
+	const blockStopEvent = {
+		type: "content_block_stop",
+		index: streamState.contentBlockIndex
+	};
+	await stream.writeSSE({
+		event: "content_block_stop",
+		data: JSON.stringify(blockStopEvent)
+	});
+	streamState.contentBlockIndex++;
+}
+async function processStreamChunks(opts) {
+	const { stream, response, toolNameMapping, streamState, acc } = opts;
+	for await (const rawEvent of response) {
+		consola.debug("Copilot raw stream event:", JSON.stringify(rawEvent));
+		if (rawEvent.data === "[DONE]") break;
+		if (!rawEvent.data) continue;
+		let chunk;
+		try {
+			chunk = JSON.parse(rawEvent.data);
+		} catch (parseError) {
+			consola.error("Failed to parse stream chunk:", parseError, rawEvent.data);
+			continue;
+		}
+		if (chunk.model && !acc.model) acc.model = chunk.model;
+		const events$1 = translateChunkToAnthropicEvents(chunk, streamState, toolNameMapping);
+		for (const event of events$1) {
+			consola.debug("Translated Anthropic event:", JSON.stringify(event));
+			processAnthropicEvent(event, acc);
+			await stream.writeSSE({
+				event: event.type,
+				data: JSON.stringify(event)
+			});
+		}
 	}
 }
+function processAnthropicEvent(event, acc) {
+	switch (event.type) {
+		case "content_block_delta":
+			handleContentBlockDelta(event.delta, acc);
+			break;
+		case "content_block_start":
+			handleContentBlockStart(event.content_block, acc);
+			break;
+		case "content_block_stop":
+			handleContentBlockStop(acc);
+			break;
+		case "message_delta":
+			handleMessageDelta(event.delta, event.usage, acc);
+			break;
+		default: break;
+	}
+}
+function handleContentBlockDelta(delta, acc) {
+	if (delta.type === "text_delta") acc.content += delta.text;
+	else if (delta.type === "input_json_delta" && acc.currentToolCall) acc.currentToolCall.input += delta.partial_json;
+}
+function handleContentBlockStart(block, acc) {
+	if (block.type === "tool_use") acc.currentToolCall = {
+		id: block.id,
+		name: block.name,
+		input: ""
+	};
+}
+function handleContentBlockStop(acc) {
+	if (acc.currentToolCall) {
+		acc.toolCalls.push(acc.currentToolCall);
+		acc.currentToolCall = null;
+	}
+}
+function handleMessageDelta(delta, usage, acc) {
+	if (delta.stop_reason) acc.stopReason = delta.stop_reason;
+	if (usage) {
+		acc.inputTokens = usage.input_tokens ?? 0;
+		acc.outputTokens = usage.output_tokens;
+	}
+}
+function recordStreamingResponse(acc, fallbackModel, ctx) {
+	const contentBlocks = [];
+	if (acc.content) contentBlocks.push({
+		type: "text",
+		text: acc.content
+	});
+	for (const tc of acc.toolCalls) contentBlocks.push({
+		type: "tool_use",
+		...tc
+	});
+	recordResponse(ctx.historyId, {
+		success: true,
+		model: acc.model || fallbackModel,
+		usage: {
+			input_tokens: acc.inputTokens,
+			output_tokens: acc.outputTokens
+		},
+		stop_reason: acc.stopReason || void 0,
+		content: contentBlocks.length > 0 ? {
+			role: "assistant",
+			content: contentBlocks
+		} : null,
+		toolCalls: acc.toolCalls.length > 0 ? acc.toolCalls : void 0
+	}, Date.now() - ctx.startTime);
+}
+function recordStreamingError(opts) {
+	const { acc, fallbackModel, ctx, error } = opts;
+	recordResponse(ctx.historyId, {
+		success: false,
+		model: acc.model || fallbackModel,
+		usage: {
+			input_tokens: 0,
+			output_tokens: 0
+		},
+		error: error instanceof Error ? error.message : "Stream error",
+		content: null
+	}, Date.now() - ctx.startTime);
+}
+function completeTracking(trackingId, inputTokens, outputTokens) {
+	if (!trackingId) return;
+	requestTracker.updateRequest(trackingId, {
+		inputTokens,
+		outputTokens
+	});
+	requestTracker.completeRequest(trackingId, 200, {
+		inputTokens,
+		outputTokens
+	});
+}
+function failTracking(trackingId, error) {
+	if (!trackingId) return;
+	requestTracker.failRequest(trackingId, error instanceof Error ? error.message : "Stream error");
+}
 function convertAnthropicMessages(messages) {
 	return messages.map((msg) => {
 		if (typeof msg.content === "string") return {
@@ -3025,7 +4290,21 @@ modelRoutes.get("/", async (c) => {
 			created: 0,
 			created_at: (/* @__PURE__ */ new Date(0)).toISOString(),
 			owned_by: model.vendor,
-			display_name: model.name
+			display_name: model.name,
+			capabilities: {
+				family: model.capabilities.family,
+				type: model.capabilities.type,
+				tokenizer: model.capabilities.tokenizer,
+				limits: {
+					max_context_window_tokens: model.capabilities.limits.max_context_window_tokens,
+					max_output_tokens: model.capabilities.limits.max_output_tokens,
+					max_prompt_tokens: model.capabilities.limits.max_prompt_tokens
+				},
+				supports: {
+					tool_calls: model.capabilities.supports.tool_calls,
+					parallel_tool_calls: model.capabilities.supports.parallel_tool_calls
+				}
+			}
 		}));
 		return c.json({
 			object: "list",
@@ -3063,7 +4342,7 @@ usageRoute.get("/", async (c) => {
 //#endregion
 //#region src/server.ts
 const server = new Hono();
-server.use(logger());
+server.use(tuiLogger());
 server.use(cors());
 server.get("/", (c) => c.text("Server running"));
 server.get("/health", (c) => {
@@ -3103,8 +4382,17 @@ async function runServer(options) {
 	state.rateLimitSeconds = options.rateLimit;
 	state.rateLimitWait = options.rateLimitWait;
 	state.showToken = options.showToken;
+	state.autoCompact = options.autoCompact;
+	if (options.autoCompact) consola.info("Auto-compact enabled: will compress context when exceeding token limits");
 	initHistory(options.history, options.historyLimit);
-	if (options.history) consola.info(`History recording enabled (max ${options.historyLimit} entries)`);
+	if (options.history) {
+		const limitText = options.historyLimit === 0 ? "unlimited" : `max ${options.historyLimit}`;
+		consola.info(`History recording enabled (${limitText} entries)`);
+	}
+	initTui({
+		enabled: true,
+		mode: options.tui
+	});
 	await ensurePaths();
 	await cacheVSCodeVersion();
 	if (options.githubToken) {
@@ -3224,7 +4512,17 @@ const start = defineCommand({
 		"history-limit": {
 			type: "string",
 			default: "1000",
-			description: "Maximum number of history entries to keep in memory"
+			description: "Maximum number of history entries to keep in memory (0 = unlimited)"
+		},
+		tui: {
+			type: "string",
+			default: "console",
+			description: "TUI mode: 'console' for simple log output, 'fullscreen' for interactive terminal UI with tabs"
+		},
+		"auto-compact": {
+			type: "boolean",
+			default: false,
+			description: "Automatically compress conversation history when exceeding model token limits"
 		}
 	},
 	run({ args }) {
@@ -3243,7 +4541,9 @@ const start = defineCommand({
 			showToken: args["show-token"],
 			proxyEnv: args["proxy-env"],
 			history: args.history,
-			historyLimit: Number.parseInt(args["history-limit"], 10)
+			historyLimit: Number.parseInt(args["history-limit"], 10),
+			tui: args.tui,
+			autoCompact: args["auto-compact"]
 		});
 	}
 });