| 1 | """Shared helpers for TTS backends.""" |
| 2 | |
| 3 | from __future__ import annotations |
| 4 | |
| 5 | import json |
| 6 | import os |
| 7 | import tempfile |
| 8 | from pathlib import Path |
| 9 | from urllib import error, request |
| 10 | |
| 11 | |
| 12 | def read_api_key(*env_names: str, label: str) -> str: |
| 13 | for env_name in env_names: |
| 14 | api_key = os.environ.get(env_name, "").strip() |
| 15 | if api_key: |
| 16 | return api_key |
| 17 | joined = " or ".join(env_names) |
| 18 | raise RuntimeError(f"Missing {label} API key. Set {joined}=<key>.") |
| 19 | |
| 20 | |
| 21 | def read_http_error(exc: error.HTTPError) -> str: |
| 22 | try: |
| 23 | body = exc.read().decode("utf-8", errors="replace") |
| 24 | except Exception: |
| 25 | body = "" |
| 26 | return f"HTTP {exc.code}: {body or exc.reason}" |
| 27 | |
| 28 | |
| 29 | def post_json(url: str, *, headers: dict[str, str], payload: dict, timeout: int = 120) -> dict: |
| 30 | body = json.dumps(payload, ensure_ascii=False).encode("utf-8") |
| 31 | req = request.Request( |
| 32 | url, |
| 33 | data=body, |
| 34 | headers={ |
| 35 | "Content-Type": "application/json", |
| 36 | **headers, |
| 37 | }, |
| 38 | method="POST", |
| 39 | ) |
| 40 | try: |
| 41 | with request.urlopen(req, timeout=timeout) as response: |
| 42 | return json.loads(response.read().decode("utf-8")) |
| 43 | except error.HTTPError as exc: |
| 44 | raise RuntimeError(read_http_error(exc)) from exc |
| 45 | except error.URLError as exc: |
| 46 | raise RuntimeError(f"HTTP request failed: {exc.reason}") from exc |
| 47 | |
| 48 | |
| 49 | def get_bytes(url: str, *, timeout: int = 180) -> bytes: |
| 50 | req = request.Request(url, method="GET") |
| 51 | try: |
| 52 | with request.urlopen(req, timeout=timeout) as response: |
| 53 | return response.read() |
| 54 | except error.HTTPError as exc: |
| 55 | raise RuntimeError(read_http_error(exc)) from exc |
| 56 | except error.URLError as exc: |
| 57 | raise RuntimeError(f"Audio download failed: {exc.reason}") from exc |
| 58 | |
| 59 | |
| 60 | def download_audio(url: str, output_path: Path) -> None: |
| 61 | publish_audio_bytes(get_bytes(url), output_path) |
| 62 | |
| 63 | |
| 64 | def temporary_path(target: Path, suffix: str) -> tuple[int, Path]: |
| 65 | """Create a temporary file beside its eventual target.""" |
| 66 | target.parent.mkdir(parents=True, exist_ok=True) |
| 67 | descriptor, raw_path = tempfile.mkstemp( |
| 68 | prefix=f".{target.name}.", |
| 69 | suffix=suffix, |
| 70 | dir=target.parent, |
| 71 | ) |
| 72 | return descriptor, Path(raw_path) |
| 73 | |
| 74 | |
| 75 | def publish_staged_pair( |
| 76 | staged_first: Path, |
| 77 | first_target: Path, |
| 78 | staged_second: Path, |
| 79 | second_target: Path, |
| 80 | ) -> None: |
| 81 | """Publish two staged files together, restoring prior targets on failure.""" |
| 82 | targets = (first_target, second_target) |
| 83 | if first_target.resolve() == second_target.resolve(): |
| 84 | raise ValueError("paired outputs must use different paths") |
| 85 | |
| 86 | backups: dict[Path, Path] = {} |
| 87 | published: set[Path] = set() |
| 88 | try: |
| 89 | for target in targets: |
| 90 | if not target.exists(): |
| 91 | continue |
| 92 | descriptor, backup = temporary_path(target, ".bak") |
| 93 | os.close(descriptor) |
| 94 | backup.unlink() |
| 95 | os.replace(target, backup) |
| 96 | backups[target] = backup |
| 97 | |
| 98 | for staged, target in ( |
| 99 | (staged_first, first_target), |
| 100 | (staged_second, second_target), |
| 101 | ): |
| 102 | os.replace(staged, target) |
| 103 | published.add(target) |
| 104 | except Exception: |
| 105 | for target in published: |
| 106 | target.unlink(missing_ok=True) |
| 107 | for target, backup in backups.items(): |
| 108 | if backup.exists(): |
| 109 | os.replace(backup, target) |
| 110 | raise |
| 111 | finally: |
| 112 | staged_first.unlink(missing_ok=True) |
| 113 | staged_second.unlink(missing_ok=True) |
| 114 | for backup in backups.values(): |
| 115 | backup.unlink(missing_ok=True) |
| 116 | |
| 117 | |
| 118 | def publish_audio_bytes(audio: bytes, output_path: Path) -> None: |
| 119 | """Publish non-empty provider audio without exposing a partial target.""" |
| 120 | if not audio: |
| 121 | raise RuntimeError("TTS provider returned empty audio data") |
| 122 | |
| 123 | descriptor, staged_path = temporary_path(output_path, ".tmp") |
| 124 | try: |
| 125 | with os.fdopen(descriptor, "wb") as stream: |
| 126 | descriptor = -1 |
| 127 | stream.write(audio) |
| 128 | stream.flush() |
| 129 | os.fsync(stream.fileno()) |
| 130 | if staged_path.stat().st_size <= 0: |
| 131 | raise RuntimeError("TTS provider returned empty audio data") |
| 132 | os.replace(staged_path, output_path) |
| 133 | finally: |
| 134 | if descriptor >= 0: |
| 135 | os.close(descriptor) |
| 136 | staged_path.unlink(missing_ok=True) |
| 137 | |
| 138 | |
| 139 | def publish_audio_subtitle( |
| 140 | audio: bytes, |
| 141 | output_path: Path, |
| 142 | subtitle: str, |
| 143 | subtitle_path: Path, |
| 144 | ) -> None: |
| 145 | """Publish one audio/SRT pair without exposing mismatched targets.""" |
| 146 | if not audio: |
| 147 | raise RuntimeError("TTS provider returned empty audio data") |
| 148 | if not subtitle.strip(): |
| 149 | raise RuntimeError("TTS provider returned empty subtitle data") |
| 150 | |
| 151 | audio_descriptor = -1 |
| 152 | subtitle_descriptor = -1 |
| 153 | staged_audio: Path | None = None |
| 154 | staged_subtitle: Path | None = None |
| 155 | try: |
| 156 | audio_descriptor, staged_audio = temporary_path(output_path, ".tmp") |
| 157 | subtitle_descriptor, staged_subtitle = temporary_path(subtitle_path, ".tmp") |
| 158 | |
| 159 | audio_stream = os.fdopen(audio_descriptor, "wb") |
| 160 | audio_descriptor = -1 |
| 161 | with audio_stream: |
| 162 | audio_stream.write(audio) |
| 163 | audio_stream.flush() |
| 164 | os.fsync(audio_stream.fileno()) |
| 165 | |
| 166 | subtitle_stream = os.fdopen( |
| 167 | subtitle_descriptor, |
| 168 | "w", |
| 169 | encoding="utf-8", |
| 170 | newline="\n", |
| 171 | ) |
| 172 | subtitle_descriptor = -1 |
| 173 | with subtitle_stream: |
| 174 | subtitle_stream.write(subtitle) |
| 175 | subtitle_stream.flush() |
| 176 | os.fsync(subtitle_stream.fileno()) |
| 177 | |
| 178 | publish_staged_pair( |
| 179 | staged_audio, |
| 180 | output_path, |
| 181 | staged_subtitle, |
| 182 | subtitle_path, |
| 183 | ) |
| 184 | finally: |
| 185 | if audio_descriptor >= 0: |
| 186 | os.close(audio_descriptor) |
| 187 | if subtitle_descriptor >= 0: |
| 188 | os.close(subtitle_descriptor) |
| 189 | if staged_audio is not None: |
| 190 | staged_audio.unlink(missing_ok=True) |
| 191 | if staged_subtitle is not None: |
| 192 | staged_subtitle.unlink(missing_ok=True) |
| 193 | |
| 194 | |
| 195 | def extension_from_format(audio_format: str) -> str: |
| 196 | normalized = audio_format.strip().lower() |
| 197 | if normalized in {"mp3", "wav"}: |
| 198 | return f".{normalized}" |
| 199 | raise RuntimeError( |
| 200 | f"Unsupported audio format for PPT narration: {audio_format}. " |
| 201 | "Use mp3 or wav, or transcode provider output before embedding." |
| 202 | ) |
| 203 |