fix(ccusage): dedupe copied Codex token events

Deduplicate Codex token usage events with a session-independent fingerprint so branched or repeated session files do not count copied history more than once.

Add regression coverage for copied branch history and validate against local Codex logs, where the current parser produced thousands of duplicate token events.

Author: ryoppippi

apps/ccusage/src/adapter/codex/parser.ts

@@ -122,6 +122,32 @@ function convertToEventUsage(
 	};
 }
 
+function createTokenUsageEventKey(event: TokenUsageEvent): string {
+	return JSON.stringify([
+		event.timestamp,
+		event.model ?? '',
+		event.inputTokens,
+		event.cachedInputTokens,
+		event.outputTokens,
+		event.reasoningOutputTokens,
+		event.totalTokens,
+	]);
+}
+
+function deduplicateTokenUsageEvents(events: TokenUsageEvent[]): TokenUsageEvent[] {
+	const seen = new Set<string>();
+	const deduplicated: TokenUsageEvent[] = [];
+	for (const event of events) {
+		const key = createTokenUsageEventKey(event);
+		if (seen.has(key)) {
+			continue;
+		}
+		seen.add(key);
+		deduplicated.push(event);
+	}
+	return deduplicated;
+}
+
 function asRecord(value: unknown): Record<string, unknown> | null {
 	return value != null && typeof value === 'object' && !Array.isArray(value)
 		? (value as Record<string, unknown>)
@@ -411,7 +437,9 @@ export async function loadTokenUsageEvents(): Promise<TokenUsageEvent[]> {
 	const directoryEvents = await Promise.all(
 		getCodexSessionsPaths().map(loadTokenUsageEventsFromDirectory),
 	);
-	return directoryEvents.flat().sort((a, b) => compareStrings(a.timestamp, b.timestamp));
+	return deduplicateTokenUsageEvents(directoryEvents.flat()).sort((a, b) =>
+		compareStrings(a.timestamp, b.timestamp),
+	);
 }
 
 async function runCodexWorker(data: CodexWorkerData): Promise<void> {
@@ -734,6 +762,78 @@ if (import.meta.vitest != null) {
 				{ sessionId: 'b', model: 'gpt-5.2', inputTokens: 20 },
 			]);
 		});
+
+		it('deduplicates copied branch history across session files', async () => {
+			const copiedHistory = [
+				JSON.stringify({
+					timestamp: '2026-05-12T08:00:00.000Z',
+					type: 'turn_context',
+					payload: {
+						model: 'gpt-5.2',
+					},
+				}),
+				JSON.stringify({
+					timestamp: '2026-05-12T08:01:00.000Z',
+					type: 'event_msg',
+					payload: {
+						type: 'token_count',
+						info: {
+							total_token_usage: {
+								input_tokens: 1_000,
+								cached_input_tokens: 100,
+								output_tokens: 200,
+								reasoning_output_tokens: 20,
+								total_tokens: 1_200,
+							},
+						},
+					},
+				}),
+			].join('\n');
+
+			await using fixture = await createFixture({
+				sessions: {
+					'project-parent.jsonl': copiedHistory,
+					'project-branch.jsonl': [
+						copiedHistory,
+						JSON.stringify({
+							timestamp: '2026-05-12T08:02:00.000Z',
+							type: 'event_msg',
+							payload: {
+								type: 'token_count',
+								info: {
+									total_token_usage: {
+										input_tokens: 1_600,
+										cached_input_tokens: 300,
+										output_tokens: 450,
+										reasoning_output_tokens: 40,
+										total_tokens: 2_050,
+									},
+								},
+							},
+						}),
+					].join('\n'),
+				},
+			});
+			vi.stubEnv('CODEX_HOME', fixture.path);
+
+			const events = await loadTokenUsageEvents();
+			expect(events).toMatchObject([
+				{
+					inputTokens: 1_000,
+					cachedInputTokens: 100,
+					outputTokens: 200,
+					totalTokens: 1_200,
+				},
+				{
+					sessionId: 'project-branch',
+					inputTokens: 600,
+					cachedInputTokens: 200,
+					outputTokens: 250,
+					totalTokens: 850,
+				},
+			]);
+			expect(events).toHaveLength(2);
+		});
 	});
 
 	describe('getCodexWorkerThreadCount', () => {