Function maskPIIBatch

apps/sim/lib/guardrails/validate_pii.ts:222–268

(
  texts: string[],
  entityTypes: string[],
  language = 'en'
)

 * can apply their own fail-safe (scrub).
 */
export async function maskPIIBatch(
  texts: string[],
  entityTypes: string[],
  language = 'en'
): Promise<string[]> {
  if (texts.length === 0) return []

  const result = new Array<string>(texts.length)

  await mapWithConcurrency(chunkIndicesByBudget(texts), CHUNK_CONCURRENCY, async (indices) => {
    const chunkTexts = indices.map((i) => texts[i])
    const spansPerText = await analyzeBatch(chunkTexts, entityTypes, language)

    // A short/misaligned batch response would silently leave the unmatched
    // strings unmasked (fail-open). Throw so the caller applies its fail-safe
    // (scrub for logs, abort for in-flight stages) instead of leaking PII.
    if (spansPerText.length !== chunkTexts.length) {
      throw new Error(
        `Presidio analyze_batch returned ${spansPerText.length} result(s) for ${chunkTexts.length} input(s)`
      )
    }

    const toAnonymize: AnonymizeBatchItem[] = []
    const anonymizePositions: number[] = []
    indices.forEach((originalIndex, pos) => {
      const spans = spansPerText[pos] ?? []
      if (spans.length === 0) {
        result[originalIndex] = chunkTexts[pos]
        return
      }
      toAnonymize.push({ text: chunkTexts[pos], analyzer_results: spans })
      anonymizePositions.push(pos)
    })

    const masked = await anonymizeBatch(toAnonymize)
    if (masked.length !== toAnonymize.length) {
      throw new Error(
        `Presidio anonymize_batch returned ${masked.length} result(s) for ${toAnonymize.length} input(s)`
      )
    }
    anonymizePositions.forEach((pos, k) => {
      result[indices[pos]] = masked[k]
    })
  })

  return result
}

export { type PIIEntityType, SUPPORTED_PII_ENTITIES } from '@/lib/guardrails/pii-entities'