danny-avila/LibreChat · warning · Error

old_text did not match the file content.

Error message

old_text did not match the file content.

What it means

Thrown by applyTextEdits when findReplacementMatch returns status 'none' for an edit's old_text — i.e. none of the four matching strategies (exact, line-trimmed, whitespace-normalized, indentation-flexible) could locate the text in the current file content. This is the file-authoring tool's 'old_text not found' failure, surfaced to the caller as a tool error via errorResult.

Source

Thrown at packages/api/src/agents/handlers.ts:1103

  }
  const whitespaceNormalized = findWhitespaceNormalizedMatch(content, needle);
  if (whitespaceNormalized.status !== 'none') {
    return whitespaceNormalized;
  }
  return findLineWindowMatch(content, needle, 'indentation-flexible');
}

function applyTextEdits(
  content: string,
  edits: TextEdit[],
): { content: string; strategies: string[] } {
  let working = content;
  const strategies: string[] = [];

  for (const edit of edits) {
    const match = findReplacementMatch(working, edit.old_text);
    if (match.status === 'none') {
      throw new Error('old_text did not match the file content.');
    }
    if (match.status === 'ambiguous') {
      throw new Error(
        `old_text matched ${match.count} locations with ${match.strategy}; make it unique before retrying.`,
      );
    }
    working =
      working.slice(0, match.index) + edit.new_text + working.slice(match.index + match.length);
    strategies.push(match.strategy);
  }

  return { content: working, strategies };
}

function formatRange(start: number, count: number): string {
  return count === 1 ? String(start) : `${start},${count}`;
}

View on GitHub (pinned to 5ff282f900)

Solutions

1. Re-read the current file content and copy old_text verbatim from it (include surrounding unique context lines).
2. Check for line-ending and tab/space mismatches; match the file's actual style.
3. Confirm the correct filePath and that no other process mutated the file since the last read.
4. Shorten old_text to a distinctive unique fragment if normalization is dropping it.

Example fix

// before
{ old_text: "function run() {\n  return 1\r\n}", new_text: "function run() {\n  return 2\n}" } // CRLF mismatch

// after
// re-read file, copy exact bytes including its real line endings
{ old_text: "function run() {\n  return 1\n}", new_text: "function run() {\n  return 2\n}" }

function findExactOccurrences(content: string, needle: string): number[] {
  const out: number[] = [];
  let from = 0;
  while (true) {
    const i = content.indexOf(needle, from);
    if (i === -1) break;
    out.push(i);
    from = i + Math.max(1, needle.length);
  }
  return out;
}
// before calling applyTextEdits, verify old_text is present:
for (const e of edits) {
  const hits = findExactOccurrences(currentContent, e.old_text);
  if (hits.length === 0) throw new Error(`old_text not found: ${e.old_text.slice(0, 60)}...`);
}

function isOldTextPresent(content: string, oldText: string): boolean {
  return content.includes(oldText);
}

try {
  const { content, strategies } = applyTextEdits(current.content, edits);
} catch (e) {
  if (e instanceof Error && /old_text did not match/.test(e.message)) {
    // re-read the file, recompute old_text from current bytes, and retry once
  } else throw e;
}

When it happens

Trigger: old_text that was never in the file; old_text copied from a different version of the file than the one currently on disk; whitespace/indentation differences larger than the normalization tolerates; old_text pointing at the wrong file path.

Common situations: The file was edited concurrently between read and write; the model hallucinated content; line-ending (CRLF vs LF) or tab/space differences exceed normalization; the file was reloaded with different formatting (a linter ran).

Related errors

• old_text matched ${match.count} locations with ${match.strat

AI-assisted analysis of danny-avila/LibreChat@5ff282f900 (2026-08-12). Data as JSON: /api/errors/79fd003d78b3e527. Report an issue: GitHub.