File size: 3,636 Bytes
b64d2eb | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | """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))
|