feat: consume context window + percentage-based compact handoff

1. Real context window: GET /models now returns modelInfo[model].contextWindow.
   The Composer uses this instead of the hardcoded MAX_CONTEXT = 1,000,000.
   Falls back to 1M when modelInfo is absent or the model has no contextWindow.

2. Percentage-based auto-compact: the compact-threshold endpoint is renamed
   to compact-percent. The CompactionView now shows a percent input (0-100,
   default 85, 0 = manual) instead of a token count input. Types renamed:
   CompactThresholdResponse → CompactPercentResponse,
   SetCompactThresholdRequest → SetCompactPercentRequest.
   Note: the field name in the backend types is still 'threshold' (not
   'percent') — the FE maps between them.

Re-mirrored .dispatch/transport-contract.reference.md.

686 tests green. 0 svelte-check errors + warnings.

5 files changed, 111 insertions, 98 deletions

diff --git a/src/app/App.svelte b/src/app/App.svelte
index 57fe16f..ae09bd5 100644
--- a/src/app/App.svelte
+++ b/src/app/App.svelte
@@ -17,7 +17,7 @@
 		ReasoningEffortSelector,
 		type CompactNowResult,
 		type ReasoningEffortSaveResult,
-		type SaveCompactThresholdResult,
+		type SaveCompactPercentResult,
 	} from "../features/chat";
 	import { manifest as conversationCacheManifest } from "../features/conversation-cache";
 	import { manifest as markdownManifest } from "../features/markdown";
@@ -249,13 +249,13 @@
 			: { ok: false, error: result.error };
 	}
 
-	async function saveCompactThreshold(
-		threshold: number,
-	): Promise<SaveCompactThresholdResult | null> {
-		const result = await store.setCompactThreshold(threshold);
+	async function saveCompactPercent(
+		percent: number,
+	): Promise<SaveCompactPercentResult | null> {
+		const result = await store.setCompactPercent(percent);
 		if (result === null) return null;
 		return result.ok
-			? { ok: true, threshold: result.threshold }
+			? { ok: true, percent: result.percent }
 			: { ok: false, error: result.error };
 	}
 
@@ -393,6 +393,7 @@
 			onQueue={handleQueue}
 			onStop={handleStop}
 			contextSize={store.activeChat.currentContextSize}
+			contextWindow={store.modelInfo[store.activeModel]?.contextWindow}
 			status={store.activeChat.error
 				? "error"
 				: store.activeChat.generating
@@ -482,13 +483,13 @@
 			{/if}
 		{/key}
 	{:else if kind === "compaction"}
-		<!-- Re-mount per conversation so the threshold + feedback can't bleed across tabs. -->
+		<!-- Re-mount per conversation so the percent + feedback can't bleed across tabs. -->
 		{#key store.currentConversationId}
 			<CompactionView
-				threshold={store.compactThreshold}
+				percent={store.compactPercent}
 				canCompact={store.activeConversationId !== null}
 				{compactNow}
-				saveThreshold={saveCompactThreshold}
+				savePercent={saveCompactPercent}
 			/>
 		{/key}
 	{:else if kind === "settings"}
diff --git a/src/app/store.svelte.ts b/src/app/store.svelte.ts
index bb08585..3f78a97 100644
--- a/src/app/store.svelte.ts
+++ b/src/app/store.svelte.ts
@@ -1,8 +1,8 @@
 import type {
 	ChatDeltaMessage,
 	ChatErrorMessage,
+	CompactPercentResponse,
 	CompactResponse,
-	CompactThresholdResponse,
 	ConversationCompactedMessage,
 	ConversationHistoryResponse,
 	ConversationListResponse,
@@ -11,10 +11,11 @@ import type {
 	ConversationStatusChangedMessage,
 	CwdResponse,
 	LspStatusResponse,
+	ModelMetadata,
 	ModelsResponse,
 	ReasoningEffort,
 	ReasoningEffortResponse,
-	SetCompactThresholdRequest,
+	SetCompactPercentRequest,
 	SetCwdRequest,
 	SetReasoningEffortRequest,
 	WarmRequest,
@@ -73,9 +74,9 @@ export type CompactResult =
 	| { readonly ok: true; readonly response: CompactResponse }
 	| { readonly ok: false; readonly error: string };
 
-/** Outcome of `PUT /conversations/:id/compact-threshold`. */
-export type CompactThresholdResult =
-	| { readonly ok: true; readonly threshold: number }
+/** Outcome of `PUT /conversations/:id/compact-percent`. */
+export type CompactPercentResult =
+	| { readonly ok: true; readonly percent: number }
 	| { readonly ok: false; readonly error: string };
 
 /** Outcome of persisting a chat-limit setting (localStorage; FE-local). */
@@ -88,6 +89,8 @@ export interface AppStore {
 	readonly activeConversationId: string | null;
 	readonly activeChat: ChatStore;
 	readonly models: readonly string[];
+	/** Per-model metadata (contextWindow, etc.) from `GET /models`. */
+	readonly modelInfo: Readonly<Record<string, ModelMetadata>>;
 	readonly activeModel: string;
 	readonly catalog: ProtocolState["catalog"];
 	/** Every received surface spec, in catalog order — all auto-subscribed + expanded. */
@@ -152,17 +155,18 @@ export interface AppStore {
 	 */
 	stopGeneration(): void;
 	/**
-	 * The workspace conversation's auto-compact threshold (tokens). `0` = disabled
+	 * The workspace conversation's auto-compact percent (0-100). `0` = disabled
 	 * (manual only); a positive number = auto-compact triggers when the last
 	 * turn's input tokens exceed it. Seeded from the backend on focus change.
 	 */
-	readonly compactThreshold: number | null;
+	readonly compactPercent: number | null;
 	/**
-	 * Persist the workspace conversation's auto-compact threshold
-	 * (`PUT /conversations/:id/compact-threshold`). `0` disables; any positive
+	 * Persist the workspace conversation's auto-compact percent
+	 * (`PUT /conversations/:id/compact-percent`). `0` disables; 1-100 sets the
+	 * trigger percentage of the model's context window. Default (null) is 85.
 	 * number enables. Works for a draft too (its id survives promotion).
 	 */
-	setCompactThreshold(threshold: number): Promise<CompactThresholdResult | null>;
+	setCompactPercent(percent: number): Promise<CompactPercentResult | null>;
 	/**
 	 * Fetch the workspace conversation's language-server status (`GET /conversations/:id/lsp`).
 	 * The backend lazily spawns servers, so this may take a moment on the first call for a cwd.
@@ -233,6 +237,7 @@ function createMetricsSync(httpBase: string, fetchImpl: typeof fetch): MetricsSy
 export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 	let protocol = $state<ProtocolState>(protocolInitialState());
 	let models = $state<readonly string[]>([]);
+	let modelInfo = $state<Readonly<Record<string, ModelMetadata>>>({});
 	let activeModel = $state(DEFAULT_MODEL);
 
 	const wsLocation = typeof location !== "undefined" ? location : undefined;
@@ -358,23 +363,23 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 		}
 	}
 
-	// The workspace conversation's auto-compact threshold. Seeded from the
+	// The workspace conversation's auto-compact percent. Seeded from the
 	// backend on focus change; null = not yet fetched. 0 = disabled.
-	let compactThreshold = $state<number | null>(null);
+	let compactPercent = $state<number | null>(null);
 
-	/** Refetch the workspace conversation's compact threshold (works for a draft too). */
-	async function refreshCompactThreshold(): Promise<void> {
+	/** Refetch the workspace conversation's compact percent (works for a draft too). */
+	async function refreshCompactPercent(): Promise<void> {
 		const id = workspaceConversationId();
-		compactThreshold = null;
+		compactPercent = null;
 		try {
 			const res = await fetchImpl(
-				`${httpBase}/conversations/${encodeURIComponent(id)}/compact-threshold`,
+				`${httpBase}/conversations/${encodeURIComponent(id)}/compact-percent`,
 			);
 			if (!res.ok) return;
-			const data = (await res.json()) as CompactThresholdResponse;
-			if (workspaceConversationId() === id) compactThreshold = data.threshold;
+			const data = (await res.json()) as CompactPercentResponse;
+			if (workspaceConversationId() === id) compactPercent = data.threshold;
 		} catch {
-			// Non-fatal: a threshold fetch failure just leaves null.
+			// Non-fatal: a percent fetch failure just leaves null.
 		}
 	}
 
@@ -542,7 +547,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 		syncSubscriptions();
 		void refreshCwd();
 		void refreshReasoningEffort();
-		void refreshCompactThreshold();
+		void refreshCompactPercent();
 	}
 
 	// Conversation lifecycle status (backend-owned, pushed via WS +
@@ -676,6 +681,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 		.then((data) => {
 			if (data === undefined) return;
 			models = data.models;
+			modelInfo = data.modelInfo ?? {};
 			if (data.models.length > 0 && !data.models.includes(activeModel)) {
 				const first = data.models[0];
 				if (first !== undefined) {
@@ -713,7 +719,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 	refreshActiveChat();
 	void refreshCwd();
 	void refreshReasoningEffort();
-	void refreshCompactThreshold();
+	void refreshCompactPercent();
 
 	// Fetch the authoritative open-conversation list from the backend (cross-
 	// device tab sync). Merges with the localStorage-restored tabs: opens new
@@ -733,6 +739,9 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 		get models(): readonly string[] {
 			return models;
 		},
+		get modelInfo(): Readonly<Record<string, ModelMetadata>> {
+			return modelInfo;
+		},
 		get activeModel(): string {
 			return activeModel;
 		},
@@ -759,8 +768,8 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 		get reasoningEffort(): ReasoningEffort | null {
 			return reasoningEffort;
 		},
-		get compactThreshold(): number | null {
-			return compactThreshold;
+		get compactPercent(): number | null {
+			return compactPercent;
 		},
 		get chatLimit(): number {
 			return chatLimit;
@@ -800,7 +809,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 				syncSubscriptions();
 				void refreshCwd();
 				void refreshReasoningEffort();
-				void refreshCompactThreshold();
+				void refreshCompactPercent();
 				// Now send on the promoted store
 				chatStores.get(conversationId)?.send(text);
 			} else {
@@ -837,7 +846,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 			syncSubscriptions();
 			void refreshCwd();
 			void refreshReasoningEffort();
-			void refreshCompactThreshold();
+			void refreshCompactPercent();
 		},
 
 		selectTab(conversationId: string): void {
@@ -850,7 +859,7 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 			syncSubscriptions();
 			void refreshCwd();
 			void refreshReasoningEffort();
-			void refreshCompactThreshold();
+			void refreshCompactPercent();
 		},
 
 		closeTab(conversationId: string): void {
@@ -988,12 +997,12 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 			}
 		},
 
-		async setCompactThreshold(threshold: number): Promise<CompactThresholdResult | null> {
+		async setCompactPercent(percent: number): Promise<CompactPercentResult | null> {
 			const id = workspaceConversationId();
-			const body: SetCompactThresholdRequest = { threshold };
+			const body: SetCompactPercentRequest = { threshold: percent };
 			try {
 				const res = await fetchImpl(
-					`${httpBase}/conversations/${encodeURIComponent(id)}/compact-threshold`,
+					`${httpBase}/conversations/${encodeURIComponent(id)}/compact-percent`,
 					{
 						method: "PUT",
 						headers: { "content-type": "application/json" },
@@ -1004,16 +1013,16 @@ export function createAppStore(opts?: CreateAppStoreOptions): AppStore {
 					const errBody = (await res.json().catch(() => null)) as { error?: string } | null;
 					return {
 						ok: false,
-						error: errBody?.error ?? `Set compact threshold failed (HTTP ${res.status})`,
+						error: errBody?.error ?? `Set compact percent failed (HTTP ${res.status})`,
 					};
 				}
-				const data = (await res.json()) as CompactThresholdResponse;
-				if (workspaceConversationId() === id) compactThreshold = data.threshold;
-				return { ok: true, threshold: data.threshold };
+				const data = (await res.json()) as CompactPercentResponse;
+				if (workspaceConversationId() === id) compactPercent = data.threshold;
+				return { ok: true, percent: data.threshold };
 			} catch (err) {
 				return {
 					ok: false,
-					error: err instanceof Error ? err.message : "Set compact threshold request failed",
+					error: err instanceof Error ? err.message : "Set compact percent request failed",
 				};
 			}
 		},
diff --git a/src/features/chat/index.ts b/src/features/chat/index.ts
index 9c65cd4..1596c53 100644
--- a/src/features/chat/index.ts
+++ b/src/features/chat/index.ts
@@ -17,7 +17,7 @@ export {
 export type { ChatStore, ChatStoreDependencies } from "./store.svelte";
 export { createChatStore } from "./store.svelte";
 export { default as ChatView } from "./ui/ChatView.svelte";
-export type { CompactNowResult, SaveCompactThresholdResult } from "./ui/CompactionView.svelte";
+export type { CompactNowResult, SaveCompactPercentResult } from "./ui/CompactionView.svelte";
 export { default as CompactionView } from "./ui/CompactionView.svelte";
 export { default as Composer } from "./ui/Composer.svelte";
 export { default as ModelSelector } from "./ui/ModelSelector.svelte";
diff --git a/src/features/chat/ui/CompactionView.svelte b/src/features/chat/ui/CompactionView.svelte
index ce2a0a0..7bec984 100644
--- a/src/features/chat/ui/CompactionView.svelte
+++ b/src/features/chat/ui/CompactionView.svelte
@@ -3,54 +3,54 @@
 		| { readonly ok: true; readonly messagesSummarized: number; readonly messagesKept: number }
 		| { readonly ok: false; readonly error: string };
 
-	export type SaveCompactThresholdResult =
-		| { readonly ok: true; readonly threshold: number }
+	export type SaveCompactPercentResult =
+		| { readonly ok: true; readonly percent: number }
 		| { readonly ok: false; readonly error: string };
 
 	let {
-		threshold,
+		percent,
 		canCompact,
 		compactNow,
-		saveThreshold,
+		savePercent,
 	}: {
-		/** The conversation's auto-compact threshold, or null when not yet fetched. 0 = disabled. */
-		threshold: number | null;
+		/** The conversation's auto-compact percent (0-100), or null when not yet fetched. 0 = disabled. */
+		percent: number | null;
 		/** Whether a real conversation is focused (a draft has nothing to compact). */
 		canCompact: boolean;
 		compactNow: () => Promise<CompactNowResult | null>;
-		saveThreshold: (threshold: number) => Promise<SaveCompactThresholdResult | null>;
+		savePercent: (percent: number) => Promise<SaveCompactPercentResult | null>;
 	} = $props();
 
-	const DEFAULT_THRESHOLD = 350000;
+	const DEFAULT_PERCENT = 85;
 
 	let compacting = $state(false);
 	let compactError = $state<string | null>(null);
 	let compactResult = $state<{ summarized: number; kept: number } | null>(null);
 
-	let thresholdInput = $state("");
-	let savingThreshold = $state(false);
-	let thresholdError = $state<string | null>(null);
-	let thresholdSaved = $state(false);
+	let percentInput = $state("");
+	let savingPercent = $state(false);
+	let percentError = $state<string | null>(null);
+	let percentSaved = $state(false);
 
 	// Sync the input from the prop when it changes (focus switch / initial load).
-	let lastThreshold = $state<number | null>(null);
+	let lastPercent = $state<number | null>(null);
 	$effect(() => {
-		if (threshold !== lastThreshold) {
-			lastThreshold = threshold;
-			thresholdInput = threshold !== null ? String(threshold) : "";
-			thresholdError = null;
-			thresholdSaved = false;
+		if (percent !== lastPercent) {
+			lastPercent = percent;
+			percentInput = percent !== null ? String(percent) : "";
+			percentError = null;
+			percentSaved = false;
 		}
 	});
 
-	const thresholdLabel = $derived(
-		threshold == null
+	const percentLabel = $derived(
+		percent == null
 			? "Loading…"
-			: threshold === 0
+			: percent === 0
 				? "Disabled (manual only)"
-				: threshold === DEFAULT_THRESHOLD
-					? `${threshold.toLocaleString("en-US")} (default)`
-					: threshold.toLocaleString("en-US"),
+				: percent === DEFAULT_PERCENT
+					? `${percent}% (default)`
+					: `${percent}%`,
 	);
 
 	async function handleCompact() {
@@ -68,22 +68,22 @@
 		}
 	}
 
-	async function handleSaveThreshold() {
-		const value = Number.parseInt(thresholdInput, 10);
-		if (Number.isNaN(value) || value < 0) {
-			thresholdError = "Must be a non-negative number";
+	async function handleSavePercent() {
+		const value = Number.parseInt(percentInput, 10);
+		if (Number.isNaN(value) || value < 0 || value > 100) {
+			percentError = "Must be 0-100";
 			return;
 		}
-		savingThreshold = true;
-		thresholdError = null;
-		thresholdSaved = false;
-		const result = await saveThreshold(value);
-		savingThreshold = false;
+		savingPercent = true;
+		percentError = null;
+		percentSaved = false;
+		const result = await savePercent(value);
+		savingPercent = false;
 		if (result === null) return;
 		if (result.ok) {
-			thresholdSaved = true;
+			percentSaved = true;
 		} else {
-			thresholdError = result.error;
+			percentError = result.error;
 		}
 	}
 </script>
@@ -120,33 +120,34 @@
 		{/if}
 	</section>
 
-	<!-- Auto-compact threshold -->
+	<!-- Auto-compact percent -->
 	<section class="flex flex-col gap-1">
-		<span class="text-xs font-semibold uppercase opacity-60">Auto-compact threshold</span>
+		<span class="text-xs font-semibold uppercase opacity-60">Auto-compact percent</span>
 		<div class="flex items-center gap-2">
 			<input
 				type="number"
-				class="input input-bordered input-sm w-32"
+				class="input input-bordered input-sm w-24"
 				min="0"
-				placeholder={DEFAULT_THRESHOLD.toLocaleString("en-US")}
-				value={thresholdInput}
-				disabled={savingThreshold}
-				onchange={handleSaveThreshold}
-				aria-label="Compact threshold (tokens)"
+				max="100"
+				placeholder={String(DEFAULT_PERCENT)}
+				value={percentInput}
+				disabled={savingPercent}
+				onchange={handleSavePercent}
+				aria-label="Compact percent (0-100)"
 			/>
-			<span class="text-xs opacity-60">tokens</span>
-			{#if savingThreshold}
+			<span class="text-xs opacity-60">%</span>
+			{#if savingPercent}
 				<span class="loading loading-spinner loading-xs"></span>
 			{/if}
 		</div>
 		<p class="text-xs opacity-50">
-			Current: {thresholdLabel}
+			Current: {percentLabel}
 			<br />
-			0 disables auto-compact. Default is {DEFAULT_THRESHOLD.toLocaleString("en-US")}.
+			0 disables auto-compact. Default is {DEFAULT_PERCENT}%.
 		</p>
-		{#if thresholdError}
-			<p class="text-xs text-error">{thresholdError}</p>
-		{:else if thresholdSaved}
+		{#if percentError}
+			<p class="text-xs text-error">{percentError}</p>
+		{:else if percentSaved}
 			<p class="text-xs text-success">Saved.</p>
 		{/if}
 	</section>
diff --git a/src/features/chat/ui/Composer.svelte b/src/features/chat/ui/Composer.svelte
index 7030153..fe9ea94 100644
--- a/src/features/chat/ui/Composer.svelte
+++ b/src/features/chat/ui/Composer.svelte
@@ -1,9 +1,7 @@
 <script lang="ts">
 	import { computeContextUsage, formatCompactTokens } from "../../../core/metrics";
 
-	// Placeholder context-window limit until the backend reports a real
-	// per-model max (see backend-handoff §3). Hardcoded to 1,000,000 tokens.
-	const MAX_CONTEXT = 1_000_000;
+	const FALLBACK_CONTEXT_WINDOW = 1_000_000;
 	const MAX_LINES = 7;
 
 	let {
@@ -11,6 +9,7 @@
 		onQueue,
 		onStop,
 		contextSize = undefined,
+		contextWindow = undefined,
 		status = "idle",
 	}: {
 		onSend: (text: string) => void;
@@ -26,6 +25,8 @@
 		// Current context occupancy (latest turn's contextSize), or `undefined`
 		// when unknown — the status bar then shows "— tokens", never 0%.
 		contextSize?: number | undefined;
+		/** Per-model context window (max tokens) from `GET /models` modelInfo. */
+		contextWindow?: number | undefined;
 		// Coarse agent status for the status-bar icon.
 		status?: "idle" | "running" | "error";
 	} = $props();
@@ -34,7 +35,8 @@
 	let inputEl: HTMLTextAreaElement | undefined;
 
 	const hasText = $derived(text.trim().length > 0);
-	const usage = $derived(computeContextUsage(contextSize, MAX_CONTEXT));
+	const effectiveMax = $derived(contextWindow ?? FALLBACK_CONTEXT_WINDOW);
+	const usage = $derived(computeContextUsage(contextSize, effectiveMax));
 	const hasUsage = $derived(contextSize !== undefined);
 
 	// One button, three modes: