| 1 | #!/usr/bin/env python3 |
| 2 | """ |
| 3 | Shared helpers for image generation backends. |
| 4 | """ |
| 5 | |
| 6 | import sys |
| 7 | from pathlib import Path |
| 8 | |
| 9 | _SCRIPTS_DIR = Path(__file__).resolve().parents[1] |
| 10 | if str(_SCRIPTS_DIR) not in sys.path: |
| 11 | sys.path.insert(0, str(_SCRIPTS_DIR)) |
| 12 | |
| 13 | from console_encoding import configure_utf8_stdio # noqa: E402 |
| 14 | |
| 15 | configure_utf8_stdio() |
| 16 | |
| 17 | if __name__ == "__main__": |
| 18 | print(__doc__) |
| 19 | print("This is an internal helper module used by image_gen.py backends.") |
| 20 | raise SystemExit(0 if any(arg in {"-h", "--help", "help"} for arg in sys.argv[1:]) else 1) |
| 21 | |
| 22 | import base64 |
| 23 | import io |
| 24 | import os |
| 25 | import re |
| 26 | import time |
| 27 | |
| 28 | import requests |
| 29 | |
| 30 | try: |
| 31 | from PIL import Image as PILImage, ImageOps as PILImageOps |
| 32 | HAS_PIL = True |
| 33 | except ImportError: |
| 34 | HAS_PIL = False |
| 35 | |
| 36 | |
| 37 | MAX_RETRIES = 3 |
| 38 | RETRY_BASE_DELAY = 10 |
| 39 | RETRY_BACKOFF = 2 |
| 40 | |
| 41 | |
| 42 | def resolve_output_path(prompt: str, output_dir: str = None, |
| 43 | filename: str = None, ext: str = ".png") -> str: |
| 44 | """Compute the final output file path based on parameters.""" |
| 45 | if filename: |
| 46 | file_name = os.path.splitext(filename)[0] |
| 47 | else: |
| 48 | safe = "".join(c for c in prompt if c.isalnum() or c in (" ", "_")).rstrip() |
| 49 | safe = safe.replace(" ", "_").lower()[:30] |
| 50 | file_name = safe or "generated_image" |
| 51 | |
| 52 | full_name = f"{file_name}{ext}" |
| 53 | if output_dir: |
| 54 | os.makedirs(output_dir, exist_ok=True) |
| 55 | return os.path.join(output_dir, full_name) |
| 56 | return full_name |
| 57 | |
| 58 | |
| 59 | CONTENT_TYPE_TO_EXT = { |
| 60 | "image/png": ".png", |
| 61 | "image/jpeg": ".jpg", |
| 62 | "image/jpg": ".jpg", |
| 63 | "image/webp": ".webp", |
| 64 | "image/gif": ".gif", |
| 65 | "image/bmp": ".bmp", |
| 66 | "image/tiff": ".tiff", |
| 67 | } |
| 68 | |
| 69 | EXT_TO_PIL_FORMAT = { |
| 70 | ".png": "PNG", |
| 71 | ".jpg": "JPEG", |
| 72 | ".jpeg": "JPEG", |
| 73 | ".webp": "WEBP", |
| 74 | ".gif": "GIF", |
| 75 | ".bmp": "BMP", |
| 76 | ".tiff": "TIFF", |
| 77 | ".tif": "TIFF", |
| 78 | } |
| 79 | |
| 80 | |
| 81 | def detect_image_extension(image_bytes: bytes, content_type: str = None) -> str | None: |
| 82 | """Best-effort detection of the real image format.""" |
| 83 | if image_bytes.startswith(b"\x89PNG\r\n\x1a\n"): |
| 84 | return ".png" |
| 85 | if image_bytes.startswith(b"\xff\xd8\xff"): |
| 86 | return ".jpg" |
| 87 | if image_bytes.startswith(b"GIF87a") or image_bytes.startswith(b"GIF89a"): |
| 88 | return ".gif" |
| 89 | if image_bytes.startswith(b"RIFF") and image_bytes[8:12] == b"WEBP": |
| 90 | return ".webp" |
| 91 | if image_bytes.startswith(b"BM"): |
| 92 | return ".bmp" |
| 93 | if image_bytes.startswith((b"II*\x00", b"MM\x00*")): |
| 94 | return ".tiff" |
| 95 | if content_type: |
| 96 | clean_type = content_type.split(";", 1)[0].strip().lower() |
| 97 | if clean_type in CONTENT_TYPE_TO_EXT: |
| 98 | return CONTENT_TYPE_TO_EXT[clean_type] |
| 99 | return None |
| 100 | |
| 101 | |
| 102 | DATA_URI_HEADER = re.compile( |
| 103 | r"data:(?P<mime>image/[A-Za-z0-9.+-]+)(?P<params>;[^,]*)?,", |
| 104 | re.IGNORECASE, |
| 105 | ) |
| 106 | |
| 107 | |
| 108 | def decode_data_uri(value: str) -> tuple[bytes, str | None]: |
| 109 | """ |
| 110 | Decode a base64 image data URI into raw bytes plus its declared content type. |
| 111 | |
| 112 | The declared type is returned so callers can hand it to `save_image_bytes` instead of |
| 113 | assuming the payload matches the output extension. |
| 114 | """ |
| 115 | header = DATA_URI_HEADER.match(value.strip()) |
| 116 | if not header: |
| 117 | raise ValueError("Expected a base64 image data URI (data:image/...;base64,...).") |
| 118 | |
| 119 | params = (header.group("params") or "").lower() |
| 120 | if "base64" not in params: |
| 121 | raise ValueError("Only base64-encoded image data URIs are supported.") |
| 122 | |
| 123 | payload = "".join(value.strip()[header.end():].split()) |
| 124 | payload += "=" * (-len(payload) % 4) |
| 125 | return base64.urlsafe_b64decode(payload), header.group("mime").lower() |
| 126 | |
| 127 | |
| 128 | def find_data_uri(content) -> str | None: |
| 129 | """ |
| 130 | Return the first base64 image data URI inside a chat completion `content` value. |
| 131 | |
| 132 | OpenAI-compatible gateways differ here: some return a dedicated image field, others |
| 133 | inline the image in the message text (often as ``) |
| 134 | or in a content-part list. |
| 135 | """ |
| 136 | if isinstance(content, str): |
| 137 | header = DATA_URI_HEADER.search(content) |
| 138 | if not header: |
| 139 | return None |
| 140 | payload = re.match(r"[A-Za-z0-9+/=_-]*", content[header.end():]).group(0) |
| 141 | return content[header.start():header.end()] + payload |
| 142 | |
| 143 | if isinstance(content, list): |
| 144 | for part in content: |
| 145 | if isinstance(part, dict): |
| 146 | nested = part.get("image_url") |
| 147 | if isinstance(nested, dict): |
| 148 | nested = nested.get("url") |
| 149 | found = find_data_uri(nested if nested else part.get("text")) |
| 150 | else: |
| 151 | found = find_data_uri(part) |
| 152 | if found: |
| 153 | return found |
| 154 | |
| 155 | return None |
| 156 | |
| 157 | |
| 158 | def _normalize_extension(ext: str) -> str: |
| 159 | """Normalize equivalent image extensions to a canonical form.""" |
| 160 | ext = ext.lower() |
| 161 | if ext == ".jpeg": |
| 162 | return ".jpg" |
| 163 | if ext == ".tif": |
| 164 | return ".tiff" |
| 165 | return ext |
| 166 | |
| 167 | |
| 168 | def save_image_bytes(image_bytes: bytes, path: str, content_type: str = None) -> str: |
| 169 | """ |
| 170 | Save image bytes to disk while keeping the file extension and the real bytes aligned. |
| 171 | |
| 172 | If the target extension differs from the actual bytes, transcode through Pillow when |
| 173 | available. Otherwise fail loudly instead of writing a misleading file. |
| 174 | """ |
| 175 | target_ext = _normalize_extension(os.path.splitext(path)[1]) |
| 176 | actual_ext = _normalize_extension(detect_image_extension(image_bytes, content_type) or "") |
| 177 | |
| 178 | if not target_ext: |
| 179 | raise ValueError(f"Output path must include an image extension: {path}") |
| 180 | |
| 181 | if actual_ext and target_ext == actual_ext: |
| 182 | with open(path, "wb") as f: |
| 183 | f.write(image_bytes) |
| 184 | print(f" File saved to: {path}") |
| 185 | report_resolution(path) |
| 186 | return path |
| 187 | |
| 188 | if not HAS_PIL: |
| 189 | actual_label = actual_ext or "unknown" |
| 190 | raise RuntimeError( |
| 191 | f"Image format mismatch for {path}: target extension is {target_ext}, " |
| 192 | f"but the actual image bytes are {actual_label}. " |
| 193 | "Install Pillow to enable automatic format conversion." |
| 194 | ) |
| 195 | |
| 196 | target_format = EXT_TO_PIL_FORMAT.get(target_ext) |
| 197 | if not target_format: |
| 198 | raise ValueError(f"Unsupported output image extension: {target_ext}") |
| 199 | |
| 200 | with PILImage.open(io.BytesIO(image_bytes)) as source: |
| 201 | image = PILImageOps.exif_transpose(source) |
| 202 | try: |
| 203 | if target_format == "JPEG": |
| 204 | has_alpha = ( |
| 205 | image.mode in ("RGBA", "LA") |
| 206 | or "transparency" in getattr(image, "info", {}) |
| 207 | ) |
| 208 | if has_alpha: |
| 209 | rgba = image.convert("RGBA") |
| 210 | alpha = rgba.getchannel("A") |
| 211 | rgb = rgba.convert("RGB") |
| 212 | converted = PILImage.new("RGB", image.size, (255, 255, 255)) |
| 213 | converted.paste(rgb, mask=alpha) |
| 214 | rgb.close() |
| 215 | alpha.close() |
| 216 | rgba.close() |
| 217 | if image is not source: |
| 218 | image.close() |
| 219 | image = converted |
| 220 | elif image.mode != "RGB": |
| 221 | converted = image.convert("RGB") |
| 222 | if image is not source: |
| 223 | image.close() |
| 224 | image = converted |
| 225 | image.save(path, format=target_format) |
| 226 | finally: |
| 227 | if image is not source: |
| 228 | image.close() |
| 229 | |
| 230 | if actual_ext and actual_ext != target_ext: |
| 231 | print(f" Converted: {actual_ext} -> {target_ext}") |
| 232 | print(f" File saved to: {path}") |
| 233 | report_resolution(path) |
| 234 | return path |
| 235 | |
| 236 | |
| 237 | def validate_image_file(path: str) -> str: |
| 238 | """Require an existing regular file that Pillow can read as an image.""" |
| 239 | image_path = Path(path) |
| 240 | if not image_path.exists(): |
| 241 | raise RuntimeError(f"Image output path does not exist: {path}") |
| 242 | if not image_path.is_file(): |
| 243 | raise RuntimeError(f"Image output path is not a file: {path}") |
| 244 | if not HAS_PIL: |
| 245 | raise RuntimeError( |
| 246 | "Pillow is required to verify generated images. " |
| 247 | "Install it with: pip install Pillow" |
| 248 | ) |
| 249 | |
| 250 | try: |
| 251 | with PILImage.open(image_path) as image: |
| 252 | image.verify() |
| 253 | except (OSError, ValueError, SyntaxError) as exc: |
| 254 | raise RuntimeError(f"Image output is not readable: {path}: {exc}") from exc |
| 255 | return str(image_path) |
| 256 | |
| 257 | |
| 258 | def report_resolution(path: str) -> None: |
| 259 | """Try to report image resolution using PIL.""" |
| 260 | if HAS_PIL: |
| 261 | try: |
| 262 | img = PILImage.open(path) |
| 263 | print(f" Resolution: {img.size[0]}x{img.size[1]}") |
| 264 | except Exception: |
| 265 | pass |
| 266 | |
| 267 | |
| 268 | def normalize_image_size(image_size: str) -> str: |
| 269 | """Normalize image size input to standard format.""" |
| 270 | s = image_size.strip() |
| 271 | upper = s.upper() |
| 272 | if upper in ("1K", "2K", "4K"): |
| 273 | return upper |
| 274 | if upper in ("512PX", "512"): |
| 275 | return "512px" |
| 276 | return s |
| 277 | |
| 278 | |
| 279 | def is_rate_limit_error(exc: Exception) -> bool: |
| 280 | """Check whether the exception appears to be rate limiting.""" |
| 281 | err_str = str(exc).lower() |
| 282 | status_code = getattr(exc, "status_code", None) |
| 283 | error_code = getattr(exc, "code", None) |
| 284 | response = getattr(exc, "response", None) |
| 285 | error_name = type(exc).__name__.lower() |
| 286 | if ( |
| 287 | status_code == 429 |
| 288 | or error_code == 429 |
| 289 | or getattr(response, "status_code", None) == 429 |
| 290 | or error_name in {"ratelimiterror", "toomanyrequestserror"} |
| 291 | ): |
| 292 | return True |
| 293 | return ( |
| 294 | "429" in err_str |
| 295 | or "rate limit" in err_str |
| 296 | or "rate-limit" in err_str |
| 297 | or "rate_limit" in err_str |
| 298 | or "too many requests" in err_str |
| 299 | or "quota" in err_str |
| 300 | or "resource_exhausted" in err_str |
| 301 | or "resource exhausted" in err_str |
| 302 | or "throttl" in err_str |
| 303 | ) |
| 304 | |
| 305 | |
| 306 | def retry_delay(attempt: int, rate_limited: bool) -> int: |
| 307 | """Return the retry delay for a given attempt.""" |
| 308 | if rate_limited: |
| 309 | return RETRY_BASE_DELAY * (RETRY_BACKOFF ** attempt) |
| 310 | return 5 |
| 311 | |
| 312 | |
| 313 | def download_image(url: str, path: str, headers: dict = None, timeout: int = 180) -> str: |
| 314 | """Download an image URL and save it to disk.""" |
| 315 | response = requests.get(url, headers=headers or {}, timeout=timeout) |
| 316 | response.raise_for_status() |
| 317 | return save_image_bytes( |
| 318 | response.content, |
| 319 | path, |
| 320 | content_type=response.headers.get("Content-Type"), |
| 321 | ) |
| 322 | |
| 323 | |
| 324 | def require_api_key(*candidates: str, message: str) -> str: |
| 325 | """Return the first non-empty env var from candidates or raise.""" |
| 326 | for name in candidates: |
| 327 | value = os.environ.get(name) |
| 328 | if value: |
| 329 | return value |
| 330 | raise ValueError(message) |
| 331 | |
| 332 | |
| 333 | def http_error(response: requests.Response, label: str) -> RuntimeError: |
| 334 | """Convert an HTTP response into a readable RuntimeError.""" |
| 335 | body = response.text.strip() |
| 336 | if len(body) > 500: |
| 337 | body = body[:500] + "..." |
| 338 | return RuntimeError(f"{label} failed ({response.status_code}): {body}") |
| 339 | |
| 340 | |
| 341 | def poll_json( |
| 342 | url: str, |
| 343 | headers: dict[str, str], |
| 344 | *, |
| 345 | interval_seconds: float = 2.0, |
| 346 | timeout_seconds: int = 300, |
| 347 | status_label: str = "status", |
| 348 | ready_values: list[str] | None = None, |
| 349 | failed_values: list[str] | None = None, |
| 350 | ) -> dict: |
| 351 | """Poll a JSON endpoint until it reports a ready or failed status.""" |
| 352 | ready = {value.lower() for value in (ready_values or ["ready", "success", "succeeded"])} |
| 353 | failed = {value.lower() for value in (failed_values or ["error", "failed", "fail"])} |
| 354 | |
| 355 | start = time.time() |
| 356 | while True: |
| 357 | response = requests.get(url, headers=headers, timeout=180) |
| 358 | response.raise_for_status() |
| 359 | payload = response.json() |
| 360 | raw_status = str(payload.get(status_label, "")).strip() |
| 361 | status = raw_status.lower() |
| 362 | |
| 363 | if raw_status: |
| 364 | print(f" Status: {raw_status}") |
| 365 | |
| 366 | if status in ready: |
| 367 | return payload |
| 368 | |
| 369 | if status in failed: |
| 370 | raise RuntimeError(f"Remote generation failed: {payload}") |
| 371 | |
| 372 | if time.time() - start > timeout_seconds: |
| 373 | raise RuntimeError( |
| 374 | f"Timed out after {timeout_seconds}s while polling {url}" |
| 375 | ) |
| 376 | |
| 377 | time.sleep(interval_seconds) |
| 378 |