sanitizeQuery

Sanitizes a search query by removing personally identifiable information (PII) and identifiers that should not be forwarded to an external search engine. Serves as a second layer of defense in addition to possible instructions in the prompt to the model to avoid including sensitive information in the query.

Strips:

• Serial numbers (S/N..., SN:..., s/n ...)

• Case references / Aktenzeichen (Az., Az:, Aktenzeichen)

• Generic alphanumeric IDs that look like codes (6+ chars with mixed letters/digits/dashes)

• "unless / except / not" clauses that typically reference specific people

• Email addresses

• Phone numbers (international and local formats)

• IBAN, SSN-style national ID numbers

• Date-of-birth patterns (DOB:, born on, Geb.)

• Street addresses (house number + street name, EN/DE)

• Trailing noise (whitespace, commas, dots)

Words wrapped in << WORD >> bypass all sieve rules and are kept verbatim.

Parameters