"""Transcribe MP3 attachments (HF Inference, local whisper, or sidecar txt)."""
from __future__ import annotations

import os
import re
from pathlib import Path
from typing import Optional

TRANSCRIPT_DIR = Path(__file__).resolve().parent.parent / "files" / "transcripts"


def _sidecar_transcript(path: Path) -> Optional[str]:
    sidecars = [
        TRANSCRIPT_DIR / f"{path.stem}.txt",
        path.with_suffix(".txt"),
    ]
    for sc in sidecars:
        if sc.is_file():
            return sc.read_text(encoding="utf-8").strip()
    return None


def transcribe(path: Path) -> Optional[str]:
    text = _sidecar_transcript(path)
    if text:
        return text

    token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACEHUB_API_TOKEN")
    if token:
        try:
            from huggingface_hub import InferenceClient

            client = InferenceClient(token=token)
            with open(path, "rb") as f:
                out = client.automatic_speech_recognition(f, model="openai/whisper-large-v3")
            if isinstance(out, dict):
                return (out.get("text") or "").strip()
            return str(out).strip()
        except Exception:
            pass

    try:
        import whisper

        model = whisper.load_model("tiny")
        result = model.transcribe(str(path))
        return (result.get("text") or "").strip()
    except Exception:
        return None


def strawberry_pie_ingredients(transcript: str) -> str:
    # Pull filling ingredients; alphabetize; no measurements
    # Known pattern from recipe audio
    candidates = []
    patterns = [
        r"ripe strawberries",
        r"granulated sugar",
        r"freshly squeezed lemon juice",
        r"cornstarch",
        r"pure vanilla extract",
        r"\bsalt\b",
        r"butter",
    ]
    lower = transcript.lower()
    for p in patterns:
        if re.search(p, lower):
            # normalize name from pattern
            name = p.replace(r"\b", "").replace("\\", "")
            candidates.append(name)
    # Prefer explicit ordered extraction from combine clause
    m = re.search(
        r"combine ([^.]+?)(?:\.|Cook)",
        transcript,
        flags=re.I,
    )
    items = []
    if m:
        chunk = m.group(1)
        # split on commas and and
        parts = re.split(r",| and ", chunk)
        items = [p.strip().lower() for p in parts if p.strip()]
    # vanilla separately
    if "vanilla" in lower:
        for phrase in ("pure vanilla extract", "vanilla extract", "vanilla"):
            if phrase in lower and phrase not in items:
                items.append("pure vanilla extract" if "pure vanilla" in lower else phrase)
                break
    # Deduplicate preserving canonical names
    canon = []
    for it in items:
        it = it.strip(" .")
        if it and it not in canon:
            canon.append(it)
    if not canon:
        canon = sorted(set(candidates))
    return ", ".join(sorted(canon))


def calculus_pages(transcript: str) -> str:
    pages = set()
    # Matches: "page 245", "pages 132, 133, and 134", "On page 132, 133 and 134"
    for m in re.finditer(
        r"pages?\s+((?:\d+(?:\s*,\s*|\s+and\s+|\s+)*)+\d+|\d+)",
        transcript,
        flags=re.I,
    ):
        pages.update(int(x) for x in re.findall(r"\d+", m.group(1)))
    # Fallback: any 3-digit number near "page"
    if len(pages) < 3:
        for m in re.finditer(r"page[^.]{0,40}?(\d{3})", transcript, flags=re.I):
            pages.add(int(m.group(1)))
        pages.update(int(x) for x in re.findall(r"\b(1[3-9]\d|2\d{2})\b", transcript))
    return ", ".join(str(p) for p in sorted(pages))