Onyscribe-Admin/vendor/platform/core/protocol.py

"""Parsing/normalisation of the Claude protocol output.

Tenant-independent: turns the raw LLM text into a structured shape that both
the PDF template and the e-mail rendering consume.
"""
from __future__ import annotations

import re

SECTION_HEADINGS = (
    "Anlass & Kontext",
    "Besprochene Themen",
    "Entscheidungen",
    "Action Items",
    "Offene Punkte",
    "Nächste Schritte",
)


def normalize_protocol_text(text: str) -> str:
    normalized_lines = []
    for raw_line in text.replace("\r\n", "\n").replace("\r", "\n").split("\n"):
        line = raw_line.rstrip()
        if line.startswith("===") and line.endswith("==="):
            normalized_lines.append(line.strip())
            continue
        stripped = line.strip()
        if not stripped:
            normalized_lines.append("")
            continue

        cleaned = re.sub(r"^\s{0,3}#{1,6}\s*", "", stripped)
        cleaned = cleaned.replace("**", "").replace("__", "")
        cleaned = re.sub(r"^\s*[-*]\s+", "• ", cleaned)
        cleaned = re.sub(r"\s+", " ", cleaned).strip()
        normalized_lines.append(cleaned)

    text = "\n".join(normalized_lines)
    text = re.sub(r"\n{3,}", "\n\n", text).strip()
    return text


def parse_protocol_sections(text: str) -> dict:
    """Zerlegt den Claude-Output in Metadaten-Blöcke und Abschnitte."""
    lines = text.replace("\r\n", "\n").split("\n")

    def collect_block(marker: str) -> list[str] | None:
        open_tag = f"==={marker}==="
        close_tag = f"===/{marker}==="
        try:
            start = lines.index(open_tag)
            end = lines.index(close_tag, start + 1)
        except ValueError:
            return None
        block = [ln.strip() for ln in lines[start + 1:end] if ln.strip()]
        return block

    projekt = collect_block("PROJEKT") or []
    kontakt = collect_block("KONTAKT") or []
    terminart_block = collect_block("TERMINART") or []
    terminart = terminart_block[0] if terminart_block else ""

    if len(kontakt) == 1 and kontakt[0].lower().strip("() ") in {"keiner", "keine", "none", ""}:
        kontakt = []

    body_lines: list[str] = []
    skip = False
    for ln in lines:
        stripped = ln.strip()
        if stripped.startswith("===") and stripped.endswith("==="):
            if stripped.startswith("===/"):
                skip = False
                continue
            skip = True
            continue
        if skip:
            continue
        body_lines.append(ln)

    sections: list[dict] = []
    current = None
    for raw in body_lines:
        line = raw.strip()
        if line in SECTION_HEADINGS:
            if current is not None:
                sections.append(current)
            current = {"heading": line, "lines": []}
            continue
        if current is None:
            continue
        if line.startswith("• "):
            current["lines"].append(("bullet", line[2:].strip()))
        elif line:
            current["lines"].append(("paragraph", line))
        else:
            current["lines"].append(("blank", ""))

    if current is not None:
        sections.append(current)

    return {
        "projekt": projekt,
        "kontakt": kontakt,
        "terminart": terminart,
        "sections": sections,
    }


def parse_address_lines(address: str) -> list[str]:
    """Splits a free-form address string into [street, PLZ city] lines."""
    address = address.strip()
    if not address:
        return []
    if "\n" in address:
        return [l.strip() for l in address.splitlines() if l.strip()]
    if "," in address:
        return [p.strip() for p in address.split(",") if p.strip()]
    m = re.search(r"(\b\d{5}\b.*)", address)
    if m:
        before = address[: m.start()].strip().rstrip(",").strip()
        plz_city = m.group(1).strip()
        return ([before] if before else []) + [plz_city]
    return [address]


def project_label(projekt_lines: list[str]) -> str:
    """Erste nicht-leere Zeile des PROJEKT-Blocks (für Dateinamen/Subject)."""
    for line in projekt_lines:
        if line and line.strip():
            return line.strip()
    return "Projekt"


def safe_filename(title: str) -> str:
    safe = re.sub(r"[^\w\-äöüÄÖÜß ]", "", title).strip()
    safe = re.sub(r"\s+", "_", safe)
    return safe or "Dokument"


def group_section_blocks(section: dict) -> list[dict]:
    """Turn the flat (kind, text) line list into HTML-friendly blocks:
    consecutive bullets become one {'type':'list','items':[...]}, paragraphs
    become {'type':'paragraph','text':...}."""
    blocks: list[dict] = []
    bullets: list[str] = []

    def flush_bullets():
        if bullets:
            blocks.append({"type": "list", "items": list(bullets)})
            bullets.clear()

    for kind, text in section.get("lines", []):
        if kind == "bullet":
            bullets.append(text)
        elif kind == "paragraph":
            flush_bullets()
            blocks.append({"type": "paragraph", "text": text})
        # blank lines are visual only — ignored in the structured shape
    flush_bullets()
    return blocks