feat: improved context usage display and update model configurations

- Updated status line to display token usage with an unknown context marker (" 5K/? ") when the model context window is unavailable.
- Updated `fugu` model specifications in `models.json` and catalog constants with corrected pricing, increased context windows, and disabled stream idle timeouts.
- Corrected OpenAI usage accounting by excluding redundant orchestration input tokens in `openai-shared` logic.

packages/ai/src/providers/openai-shared.ts

@@ -2419,7 +2419,7 @@ export function populateResponsesUsageFromResponse(
 	const orchestrationInputCachedTokens = details?.orchestration_input_cached_tokens ?? 0;
 	const orchestrationOutputTokens = outputDetails?.orchestration_output_tokens ?? 0;
 	const accounting = calculateOpenAIUsageAccounting({
-		promptTokens: (usage.input_tokens ?? 0) + orchestrationInputTokens + orchestrationInputCachedTokens,
+		promptTokens: (usage.input_tokens ?? 0) + orchestrationInputTokens,
 		outputTokens: (usage.output_tokens ?? 0) + orchestrationOutputTokens,
 		cachedTokens: (details?.cached_tokens ?? usage.prompt_cache_hit_tokens ?? 0) + orchestrationInputCachedTokens,
 		reasoningTokens: outputDetails?.reasoning_tokens ?? 0,

packages/ai/test/usage-attribution.test.ts

@@ -1,7 +1,10 @@
 import { describe, expect, it } from "bun:test";
 import { applyAnthropicUsageExtras } from "@oh-my-pi/pi-ai/providers/anthropic";
 import { parseChunkUsage } from "@oh-my-pi/pi-ai/providers/openai-completions";
-import { calculateOpenAIUsageAccounting, populateResponsesUsageFromResponse } from "@oh-my-pi/pi-ai/providers/openai-shared";
+import {
+	calculateOpenAIUsageAccounting,
+	populateResponsesUsageFromResponse,
+} from "@oh-my-pi/pi-ai/providers/openai-shared";
 import type { AssistantMessage, Model, Usage } from "@oh-my-pi/pi-ai/types";
 import { buildModel } from "@oh-my-pi/pi-catalog/build";
 
@@ -273,7 +276,7 @@ describe("openai-responses usage attribution", () => {
 		populateResponsesUsageFromResponse(output, {
 			input_tokens: 120,
 			output_tokens: 80,
-			total_tokens: 200,
+			total_tokens: 270,
 			input_tokens_details: {
 				cached_tokens: 10,
 				orchestration_input_tokens: 30,

packages/catalog/src/models.json

@@ -60466,12 +60466,12 @@
 			],
 			"cost": {
 				"input": 0.66,
-				"output": 3.5,
-				"cacheRead": 0.33,
+				"output": 3.41,
+				"cacheRead": 0.144,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 65535,
+			"maxTokens": 262144,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -64132,12 +64132,12 @@
 			],
 			"cost": {
 				"input": 0.66,
-				"output": 3.5,
-				"cacheRead": 0.33,
+				"output": 3.41,
+				"cacheRead": 0.144,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 65535,
+			"maxTokens": 262144,
 			"thinking": {
 				"mode": "effort",
 				"efforts": [
@@ -69259,7 +69259,7 @@
 				"cacheRead": 0,
 				"cacheWrite": 0
 			},
-			"contextWindow": null,
+			"contextWindow": 1000000,
 			"maxTokens": null,
 			"thinking": {
 				"mode": "effort",
@@ -69273,7 +69273,7 @@
 			},
 			"compat": {
 				"includeEncryptedReasoning": false,
-				"streamIdleTimeoutMs": 300000
+				"streamIdleTimeoutMs": 0
 			}
 		},
 		"fugu-ultra": {
@@ -69292,7 +69292,7 @@
 				"cacheRead": 0.5,
 				"cacheWrite": 0
 			},
-			"contextWindow": null,
+			"contextWindow": 1000000,
 			"maxTokens": null,
 			"thinking": {
 				"mode": "effort",
@@ -69306,7 +69306,7 @@
 			},
 			"compat": {
 				"includeEncryptedReasoning": false,
-				"streamIdleTimeoutMs": 300000
+				"streamIdleTimeoutMs": 0
 			}
 		},
 		"fugu-ultra-20260615": {
@@ -69325,7 +69325,7 @@
 				"cacheRead": 0.5,
 				"cacheWrite": 0
 			},
-			"contextWindow": null,
+			"contextWindow": 1000000,
 			"maxTokens": null,
 			"thinking": {
 				"mode": "effort",
@@ -69339,7 +69339,7 @@
 			},
 			"compat": {
 				"includeEncryptedReasoning": false,
-				"streamIdleTimeoutMs": 300000
+				"streamIdleTimeoutMs": 0
 			}
 		}
 	},

packages/catalog/src/provider-models/openai-compat.ts

@@ -2527,6 +2527,7 @@ export function moonshotModelManagerOptions(
 const SAKANA_DEFAULT_BASE_URL = "https://api.sakana.ai/v1";
 const SAKANA_FREE_ROUTER_COST = { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 } as const;
 const SAKANA_FUGU_ULTRA_COST = { input: 5, output: 30, cacheRead: 0.5, cacheWrite: 0 } as const;
+const SAKANA_FUGU_ULTRA_CONTEXT_WINDOW = 1_000_000;
 const SAKANA_FUGU_THINKING: ThinkingConfig = {
 	mode: "effort",
 	efforts: [Effort.High, Effort.XHigh],
@@ -2534,7 +2535,7 @@ const SAKANA_FUGU_THINKING: ThinkingConfig = {
 };
 const SAKANA_RESPONSES_COMPAT: ModelSpec<"openai-responses">["compat"] = {
 	includeEncryptedReasoning: false,
-	streamIdleTimeoutMs: 300_000,
+	streamIdleTimeoutMs: 0,
 };
 
 function normalizeSakanaBaseUrl(baseUrl: string | undefined): string {
@@ -2551,6 +2552,7 @@ function createSakanaFuguStaticModel(
 	id: string,
 	name: string,
 	cost: ModelSpec<"openai-responses">["cost"],
+	contextWindow: number | null,
 ): ModelSpec<"openai-responses"> {
 	return {
 		id,
@@ -2561,7 +2563,7 @@ function createSakanaFuguStaticModel(
 		reasoning: true,
 		input: ["text"],
 		cost: { ...cost },
-		contextWindow: null,
+		contextWindow,
 		maxTokens: null,
 		thinking: { ...SAKANA_FUGU_THINKING },
 		compat: { ...SAKANA_RESPONSES_COMPAT },
@@ -2569,9 +2571,14 @@ function createSakanaFuguStaticModel(
 }
 
 export const SAKANA_FUGU_STATIC_MODELS: readonly ModelSpec<"openai-responses">[] = [
-	createSakanaFuguStaticModel("fugu", "Fugu", SAKANA_FREE_ROUTER_COST),
-	createSakanaFuguStaticModel("fugu-ultra", "Fugu Ultra", SAKANA_FUGU_ULTRA_COST),
-	createSakanaFuguStaticModel("fugu-ultra-20260615", "Fugu Ultra 20260615", SAKANA_FUGU_ULTRA_COST),
+	createSakanaFuguStaticModel("fugu", "Fugu", SAKANA_FREE_ROUTER_COST, SAKANA_FUGU_ULTRA_CONTEXT_WINDOW),
+	createSakanaFuguStaticModel("fugu-ultra", "Fugu Ultra", SAKANA_FUGU_ULTRA_COST, SAKANA_FUGU_ULTRA_CONTEXT_WINDOW),
+	createSakanaFuguStaticModel(
+		"fugu-ultra-20260615",
+		"Fugu Ultra 20260615",
+		SAKANA_FUGU_ULTRA_COST,
+		SAKANA_FUGU_ULTRA_CONTEXT_WINDOW,
+	),
 ];
 
 const SAKANA_FUGU_STATIC_MODEL_BY_ID = new Map(SAKANA_FUGU_STATIC_MODELS.map(model => [model.id, model] as const));

packages/catalog/test/sakana-provider.test.ts

@@ -51,11 +51,15 @@ describe("Sakana AI provider support", () => {
 
 		const bundled = getBundledModels("sakana");
 		expect(bundled.map(model => model.id).sort()).toEqual(["fugu", "fugu-ultra", "fugu-ultra-20260615"]);
+		expect(bundled.find(model => model.id === "fugu")?.contextWindow).toBe(1_000_000);
+		expect(bundled.find(model => model.id === "fugu-ultra")?.contextWindow).toBe(1_000_000);
+		expect(bundled.find(model => model.id === "fugu-ultra-20260615")?.contextWindow).toBe(1_000_000);
 		for (const model of bundled) {
 			expect(model.api).toBe("openai-responses");
 			expect(model.thinking?.efforts).toEqual([Effort.High, Effort.XHigh]);
 			expect(model.thinking?.effortMap?.[Effort.XHigh]).toBe("max");
 			expect((model.compat as ResolvedOpenAIResponsesCompat).includeEncryptedReasoning).toBe(false);
+			expect((model.compat as ResolvedOpenAIResponsesCompat).streamIdleTimeoutMs).toBe(0);
 		}
 
 		const provider = getOAuthProviders().find(item => item.id === "sakana");

packages/coding-agent/src/modes/components/footer.ts

@@ -142,7 +142,8 @@ export class FooterComponent implements Component {
 		// After compaction, tokens are unknown until the next LLM response.
 		const contextUsage = this.session.getContextUsage();
 		const contextWindow = contextUsage?.contextWindow ?? state.model?.contextWindow ?? 0;
-		const contextPercentValue = contextUsage?.percent ?? 0;
+		const contextTokens = contextUsage?.tokens ?? 0;
+		const contextPercentValue = contextWindow > 0 ? (contextUsage?.percent ?? 0) : null;
 
 		// Replace home directory with ~
 		let pwd = shortenPath(getProjectDir());
@@ -186,8 +187,8 @@ export class FooterComponent implements Component {
 		// Colorize context percentage based on usage
 		let contextPercentStr: string;
 		const autoIndicator = this.#autoCompactEnabled ? " (auto)" : "";
-		const contextPercentDisplay = `${formatContextUsage(contextPercentValue, contextWindow)}${autoIndicator}`;
-		if (contextUsage) {
+		const contextPercentDisplay = `${formatContextUsage(contextPercentValue, contextWindow, contextTokens)}${autoIndicator}`;
+		if (contextUsage && contextPercentValue !== null) {
 			const color = getContextUsageThemeColor(getContextUsageLevel(contextPercentValue, contextWindow));
 			contextPercentStr =
 				color === "statusLineContext" ? contextPercentDisplay : theme.fg(color, contextPercentDisplay);

packages/coding-agent/src/modes/components/status-line/component.ts

@@ -727,10 +727,12 @@ export class StatusLineComponent implements Component {
 
 		let contextWindow = state.model?.contextWindow ?? this.session.model?.contextWindow ?? 0;
 		let contextPercent: number | null = 0;
+		let contextTokens = 0;
 		if (includeContext) {
 			const breakdown = this.getCachedContextBreakdown();
+			contextTokens = breakdown.usedTokens;
 			contextWindow = breakdown.contextWindow || contextWindow;
-			contextPercent = contextWindow > 0 ? (breakdown.usedTokens / contextWindow) * 100 : 0;
+			contextPercent = contextWindow > 0 ? (breakdown.usedTokens / contextWindow) * 100 : null;
 		}
 
 		// Collab guest: context comes from the host's state frames — the local
@@ -738,6 +740,7 @@ export class StatusLineComponent implements Component {
 		const collabState = this.#collabStatus?.stateOverride;
 		if (collabState?.contextUsage) {
 			contextWindow = collabState.contextUsage.contextWindow || contextWindow;
+			contextTokens = collabState.contextUsage.tokens ?? contextTokens;
 			contextPercent = collabState.contextUsage.percent ?? contextPercent;
 		}
 
@@ -756,6 +759,7 @@ export class StatusLineComponent implements Component {
 			collab: this.#collabStatus,
 			usageStats,
 			contextPercent,
+			contextTokens,
 			contextWindow,
 			autoCompactEnabled: this.#autoCompactEnabled,
 			subagentCount: this.#subagentCount,

packages/coding-agent/src/modes/components/status-line/context-thresholds.ts

@@ -56,10 +56,18 @@ export function getContextUsageLevel(contextPercent: number, contextWindow: numb
 }
 
 /**
- * Format context usage as `<percent>%/<window>` (e.g. `5.1%/1M`), matching the
- * status line's context gauge so subagent and footer renderers stay in sync.
+ * Format context usage as `<percent>%/<window>` when the model window is known.
+ * Unknown windows render as `<tokens>/?`, because `0.0%/0` suggests a real
+ * empty context instead of missing provider metadata.
  */
-export function formatContextUsage(contextPercent: number | null | undefined, contextWindow: number): string {
+export function formatContextUsage(
+	contextPercent: number | null | undefined,
+	contextWindow: number,
+	usedTokens?: number,
+): string {
+	if (!Number.isFinite(contextWindow) || contextWindow <= 0) {
+		return `${formatNumber(usedTokens ?? 0)}/?`;
+	}
 	const pct = contextPercent === null || contextPercent === undefined ? "?" : `${contextPercent.toFixed(1)}%`;
 	return `${pct}/${formatNumber(contextWindow)}`;
 }

packages/coding-agent/src/modes/components/status-line/segments.ts

@@ -375,7 +375,7 @@ const contextPctSegment: StatusLineSegment = {
 		const window = ctx.contextWindow;
 
 		const autoIcon = ctx.autoCompactEnabled && theme.icon.auto ? ` ${theme.icon.auto}` : "";
-		const text = `${formatContextUsage(pct, window)}${autoIcon}`;
+		const text = `${formatContextUsage(pct, window, ctx.contextTokens)}${autoIcon}`;
 
 		const color = getContextUsageThemeColor(getContextUsageLevel(pct ?? 0, window));
 		const content = withIcon(theme.icon.context, theme.fg(color, text));

packages/coding-agent/src/modes/components/status-line/types.ts

@@ -73,6 +73,7 @@ export interface SegmentContext {
 	};
 	/** Context usage percent, or null when unknown (e.g. right after compaction). */
 	contextPercent: number | null;
+	contextTokens: number;
 	contextWindow: number;
 	autoCompactEnabled: boolean;
 	subagentCount: number;

packages/coding-agent/src/session/agent-session.ts

@@ -12168,8 +12168,8 @@ export class AgentSession {
 		pendingMessages?: AgentMessage[];
 	}): ContextUsageBreakdown | undefined {
 		const model = this.model;
-		const contextWindow = options?.contextWindow ?? model?.contextWindow ?? 0;
-		if (!Number.isFinite(contextWindow) || contextWindow <= 0) return undefined;
+		const rawContextWindow = options?.contextWindow ?? model?.contextWindow ?? 0;
+		const contextWindow = Number.isFinite(rawContextWindow) && rawContextWindow > 0 ? rawContextWindow : 0;
 
 		const { skillsTokens, toolsTokens, systemContextTokens, systemPromptTokens } = computeNonMessageBreakdown(this);
 		const categoryNonMessageTokens = skillsTokens + toolsTokens + systemContextTokens + systemPromptTokens;

packages/coding-agent/test/issue-953-repro.test.ts

@@ -32,6 +32,7 @@ function createCtx(usage: Partial<SegmentContext["usageStats"]>): SegmentContext
 			...usage,
 		},
 		contextPercent: 0,
+		contextTokens: 0,
 		contextWindow: 0,
 		autoCompactEnabled: false,
 		subagentCount: 0,

packages/coding-agent/test/status-line-context-cache.test.ts

@@ -248,4 +248,23 @@ describe("StatusLineComponent context breakdown", () => {
 		const plain = comp.getTopBorder(80).content.replaceAll(/\x1b\[[0-9;]*m/g, "");
 		expect(plain).toContain("0.5%/272K");
 	});
+
+	it("renders token usage with an unknown marker when the model window is unavailable", () => {
+		const { session } = makeSession({
+			messages: [userMessage("hi")],
+			contextWindow: 0,
+			usage: { tokens: 5000, contextWindow: 0, percent: 0 },
+		});
+		const comp = new StatusLineComponent(session);
+		comp.updateSettings({
+			preset: "custom",
+			leftSegments: ["context_pct"],
+			rightSegments: [],
+			separator: "powerline-thin",
+		});
+
+		const plain = comp.getTopBorder(80).content.replaceAll(/\x1b\[[0-9;]*m/g, "");
+		expect(plain).toContain("5K/?");
+		expect(plain).not.toContain("0.0%/0");
+	});
 });

packages/coding-agent/test/status-line-model.test.ts

@@ -32,6 +32,7 @@ function createModelContext(advisorActive: boolean): SegmentContext {
 			tokensPerSecond: null,
 		},
 		contextPercent: 0,
+		contextTokens: 0,
 		contextWindow: 0,
 		autoCompactEnabled: false,
 		subagentCount: 0,

packages/coding-agent/test/status-line-overflow.test.ts

@@ -56,6 +56,7 @@ function createCtx(overrides?: { pathMaxLength?: number; branch?: string | null
 			tokensPerSecond: null,
 		},
 		contextPercent: 0,
+		contextTokens: 0,
 		contextWindow: 0,
 		autoCompactEnabled: false,
 		subagentCount: 0,

packages/coding-agent/test/status-line-path.test.ts

@@ -42,6 +42,7 @@ function createPathContext(): SegmentContext {
 			tokensPerSecond: null,
 		},
 		contextPercent: 0,
+		contextTokens: 0,
 		contextWindow: 0,
 		autoCompactEnabled: false,
 		subagentCount: 0,