fix(jsonl): add explicit UTF-8 BOM stripping

Strip BOM character before JSON parsing to ensure robust
cross-platform JSONL file handling.

src/web-server/jsonl-parser.ts

@@ -70,10 +70,12 @@ export interface ParserOptions {
  * Returns null for non-assistant entries or entries without usage data
  */
 export function parseUsageEntry(line: string, projectPath: string): RawUsageEntry | null {
-  if (!line.trim()) return null;
+  // Strip UTF-8 BOM if present (can occur on first line of some files)
+  const cleanLine = line.replace(/^\uFEFF/, '').trim();
+  if (!cleanLine) return null;
 
   try {
-    const entry = JSON.parse(line);
+    const entry = JSON.parse(cleanLine);
 
     // Only process assistant entries with usage data
     if (entry.type !== 'assistant') return null;

tests/unit/jsonl-parser.test.ts

@@ -133,6 +133,16 @@ describe('parseUsageEntry', () => {
     expect(parseUsageEntry('not json at all', '/test')).toBeNull();
   });
 
+  test('strips UTF-8 BOM from line before parsing', () => {
+    // UTF-8 BOM character (\uFEFF) can appear at start of files
+    const bomEntry = '\uFEFF' + VALID_ASSISTANT_ENTRY;
+    const result = parseUsageEntry(bomEntry, '/test');
+
+    expect(result).not.toBeNull();
+    expect(result!.model).toBe('claude-sonnet-4-5');
+    expect(result!.inputTokens).toBe(1000);
+  });
+
   test('includes project path in result', () => {
     const result = parseUsageEntry(VALID_ASSISTANT_ENTRY, '/custom/project/path');
     expect(result!.projectPath).toBe('/custom/project/path');