app/services/name_signal_extractor.py

"""Deterministic preferred-name/form-of-address extraction. "I am X" is
deliberately excluded: it matches too many non-name sentences ("I am
confused", "I am working on..."). Patterns use a lookahead rather than an
end-of-string anchor so a name followed by more sentence content ("My name
is Anshuman, and I'm studying robotics") still matches. See
docs/commit-memory-adaptation-architecture.md.
"""
from __future__ import annotations

import re

_NAME_SIGNAL_PATTERNS = (
    re.compile(
        r"\bmy name is\s+([A-Za-z][A-Za-z .'-]{0,48}?)(?=\s+(?:and|but)\b|[,.;!?\n]|$)",
        re.IGNORECASE,
    ),
    re.compile(
        r"\b(?:please\s+)?call me\s+([A-Za-z][A-Za-z .'-]{0,48}?)(?=\s+(?:and|but)\b|[,.;!?\n]|$)",
        re.IGNORECASE,
    ),
    re.compile(
        r"\byou can call me\s+([A-Za-z][A-Za-z .'-]{0,48}?)(?=\s+(?:and|but)\b|[,.;!?\n]|$)",
        re.IGNORECASE,
    ),
    re.compile(
        r"\bi go by\s+([A-Za-z][A-Za-z .'-]{0,48}?)(?=\s+(?:and|but)\b|[,.;!?\n]|$)",
        re.IGNORECASE,
    ),
)


def extract_preferred_name(text: str) -> str | None:
    for pattern in _NAME_SIGNAL_PATTERNS:
        match = pattern.search(text or "")
        if not match:
            continue
        name = re.sub(r"\s+", " ", match.group(1)).strip(" .'-")
        if 1 < len(name) <= 80:
            return name
    return None