From e7738832c219688734ba3c8538291dd7be21c98c Mon Sep 17 00:00:00 2001
From: Leo Kim <tsagi2045@gmail.com>
Date: Wed, 13 May 2026 16:58:59 +0900
Subject: [PATCH] fix(coding-agent): align status-line context% with /context
 command output
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Status-line's context_pct segment was computing tokens via
calculatePromptTokens(lastAssistantMessage.usage), which sums input +
cacheRead + cacheWrite from the Anthropic API usage object. The /context
slash command is computed by computeContextBreakdown, an offline estimate
over the live session state (systemPrompt + tools + skills + messages).

Both numbers are correct under their own definition, but they can
diverge by 2x+ on the same session when a turn rotates cache tiers
(e.g. 5m → 1h ephemeral re-cache) and cache_creation_input_tokens spikes.
Users read the two surfaces as one consistent dashboard and treat the
mismatch as a bug.

Repro: same session at the same moment reports 212K (21.2%) in /context
and 44.2%/1M in the status line — ~230K gap driven by per-turn
cache_creation on a system-prompt boundary.

This change makes status-line use the same computeContextBreakdown
source as /context so both surfaces stay consistent. The breakdown
result is cached with a 2s TTL inside the component so the per-frame
status-line render does not re-walk every message via
estimateMessagesTokens on long sessions. The Anthropic API per-turn
prompt size remains observable via existing token_in / cache_read /
cache_write / token_total segments.
---
 .../src/modes/components/status-line.ts       | 32 +++++++++++++------
 1 file changed, 22 insertions(+), 10 deletions(-)

diff --git a/packages/coding-agent/src/modes/components/status-line.ts b/packages/coding-agent/src/modes/components/status-line.ts
index 3d8aab6ac..7b94c2507 100644
--- a/packages/coding-agent/src/modes/components/status-line.ts
+++ b/packages/coding-agent/src/modes/components/status-line.ts
@@ -1,5 +1,4 @@
 import * as fs from "node:fs";
-import type { AssistantMessage } from "@oh-my-pi/pi-ai";
 import { type Component, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
 import { formatCount, getProjectDir } from "@oh-my-pi/pi-utils";
 import { $ } from "bun";
@@ -7,7 +6,7 @@ import { settings } from "../../config/settings";
 import type { StatusLinePreset, StatusLineSegmentId, StatusLineSeparatorStyle } from "../../config/settings-schema";
 import { theme } from "../../modes/theme/theme";
 import type { AgentSession } from "../../session/agent-session";
-import { calculatePromptTokens } from "../../session/compaction/compaction";
+import { computeContextBreakdown } from "../utils/context-usage";
 import * as git from "../../utils/git";
 import { getSessionAccentAnsi, getSessionAccentHex } from "../../utils/session-color";
 import { sanitizeStatusText } from "../shared";
@@ -73,6 +72,10 @@ export class StatusLineComponent implements Component {
 	#lastTokensPerSecond: number | null = null;
 	#lastTokensPerSecondTimestamp: number | null = null;
 
+	// Context breakdown caching (2s TTL — aligns with /context command output)
+	#cachedBreakdown: { usedTokens: number; contextWindow: number } | null = null;
+	#breakdownFetchedAt = 0;
+
 	constructor(private readonly session: AgentSession) {
 		this.#settings = {
 			preset: settings.get("statusLine.preset"),
@@ -301,6 +304,19 @@ export class StatusLineComponent implements Component {
 		return null;
 	}
 
+	#getCachedContextBreakdown(): { usedTokens: number; contextWindow: number } {
+		const now = Date.now();
+		if (!this.#cachedBreakdown || now - this.#breakdownFetchedAt > 2_000) {
+			const breakdown = computeContextBreakdown(this.session);
+			this.#cachedBreakdown = {
+				usedTokens: breakdown.usedTokens,
+				contextWindow: breakdown.contextWindow,
+			};
+			this.#breakdownFetchedAt = now;
+		}
+		return this.#cachedBreakdown;
+	}
+
 	#buildSegmentContext(width: number): SegmentContext {
 		const state = this.session.state;
 
@@ -318,14 +334,10 @@ export class StatusLineComponent implements Component {
 			tokensPerSecond: this.#getTokensPerSecond(),
 		};
 
-		// Get context percentage
-		const lastAssistantMessage = state.messages
-			.slice()
-			.reverse()
-			.find(m => m.role === "assistant" && m.stopReason !== "aborted") as AssistantMessage | undefined;
-
-		const contextTokens = lastAssistantMessage ? calculatePromptTokens(lastAssistantMessage.usage) : 0;
-		const contextWindow = state.model?.contextWindow || 0;
+		// Context usage — aligned with /context command so both surfaces report the same value
+		const breakdown = this.#getCachedContextBreakdown();
+		const contextTokens = breakdown.usedTokens;
+		const contextWindow = breakdown.contextWindow || state.model?.contextWindow || 0;
 		const contextPercent = contextWindow > 0 ? (contextTokens / contextWindow) * 100 : 0;
 
 		return {