| 1 | #!/usr/bin/env python3 |
| 2 | """ |
| 3 | OpenRouter image generation backend. |
| 4 | |
| 5 | Configuration keys: |
| 6 | OPENROUTER_API_KEY (required) |
| 7 | OPENROUTER_BASE_URL (optional) |
| 8 | OPENROUTER_MODEL (optional) |
| 9 | """ |
| 10 | |
| 11 | import sys |
| 12 | from pathlib import Path |
| 13 | |
| 14 | _SCRIPTS_DIR = Path(__file__).resolve().parents[1] |
| 15 | if str(_SCRIPTS_DIR) not in sys.path: |
| 16 | sys.path.insert(0, str(_SCRIPTS_DIR)) |
| 17 | |
| 18 | from console_encoding import configure_utf8_stdio # noqa: E402 |
| 19 | |
| 20 | configure_utf8_stdio() |
| 21 | |
| 22 | if __name__ == "__main__": |
| 23 | print(__doc__) |
| 24 | print("Use via: python3 skills/ppt-master/scripts/image_gen.py \"prompt\" --backend openrouter") |
| 25 | raise SystemExit(0 if any(arg in {"-h", "--help", "help"} for arg in sys.argv[1:]) else 1) |
| 26 | |
| 27 | import os |
| 28 | import time |
| 29 | import threading |
| 30 | import requests |
| 31 | |
| 32 | from image_backends.backend_common import ( |
| 33 | MAX_RETRIES, |
| 34 | decode_data_uri, |
| 35 | find_data_uri, |
| 36 | is_rate_limit_error, |
| 37 | normalize_image_size, |
| 38 | resolve_output_path, |
| 39 | retry_delay, |
| 40 | save_image_bytes, |
| 41 | ) |
| 42 | |
| 43 | |
| 44 | # ╔══════════════════════════════════════════════════════════════════╗ |
| 45 | # ║ Constants ║ |
| 46 | # ╚══════════════════════════════════════════════════════════════════╝ |
| 47 | |
| 48 | VALID_ASPECT_RATIOS = [ |
| 49 | "1:1", "1:4", "1:8", |
| 50 | "2:3", "3:2", "3:4", "4:1", "4:3", |
| 51 | "4:5", "5:4", "8:1", "9:16", "16:9", "21:9" |
| 52 | ] |
| 53 | |
| 54 | VALID_IMAGE_SIZES = ["1K", "2K", "4K", "0.5K"] |
| 55 | |
| 56 | DEFAULT_MODEL = "google/gemini-3.1-flash-image-preview" |
| 57 | DEFAULT_ENDPOINT = "https://openrouter.ai/api/v1" |
| 58 | |
| 59 | # ╔══════════════════════════════════════════════════════════════════╗ |
| 60 | # ║ Image Generation ║ |
| 61 | # ╚══════════════════════════════════════════════════════════════════╝ |
| 62 | |
| 63 | def _resolve_url(base_url: str) -> str: |
| 64 | """Resolve the OpenRouter generation endpoint.""" |
| 65 | return base_url.rstrip("/") + "/chat/completions" |
| 66 | |
| 67 | def _message_image_uri(message: dict) -> str | None: |
| 68 | """ |
| 69 | Locate the generated image in a chat completion message. |
| 70 | |
| 71 | OpenRouter returns it in a dedicated `images` array; other OpenAI-compatible endpoints |
| 72 | reachable through OPENROUTER_BASE_URL inline it in the message content instead. |
| 73 | """ |
| 74 | images = message.get("images") |
| 75 | if images: |
| 76 | url = images[0].get("image_url") |
| 77 | if isinstance(url, dict): |
| 78 | url = url.get("url") |
| 79 | if url: |
| 80 | return url |
| 81 | |
| 82 | return find_data_uri(message.get("content")) |
| 83 | |
| 84 | |
| 85 | def _generate_image(api_key: str, prompt: str, |
| 86 | aspect_ratio: str = "1:1", image_size: str = "1K", |
| 87 | output_dir: str = None, filename: str = None, |
| 88 | model: str = DEFAULT_MODEL, base_url: str = DEFAULT_ENDPOINT) -> str: |
| 89 | """ |
| 90 | Image generation via OpenRouter's API. |
| 91 | """ |
| 92 | |
| 93 | url = _resolve_url(base_url) |
| 94 | |
| 95 | headers = { |
| 96 | "Authorization": f"Bearer {api_key}", |
| 97 | "Content-Type": "application/json" |
| 98 | } |
| 99 | |
| 100 | payload = { |
| 101 | "model": model, |
| 102 | "messages": [ |
| 103 | { |
| 104 | "role": "user", |
| 105 | "content": prompt |
| 106 | } |
| 107 | ], |
| 108 | "modalities": ["image", "text"], |
| 109 | "image_config": { |
| 110 | "aspect_ratio": aspect_ratio, |
| 111 | "image_size": image_size |
| 112 | } |
| 113 | } |
| 114 | |
| 115 | print(f"[OpenRouter]") |
| 116 | print(f" Model: {model}") |
| 117 | print(f" Prompt: {prompt[:120]}{'...' if len(prompt) > 120 else ''}") |
| 118 | print(f" Aspect Ratio: {aspect_ratio}") |
| 119 | print(f" Image Size: {image_size}") |
| 120 | print() |
| 121 | |
| 122 | start_time = time.time() |
| 123 | print(f" [..] Generating...", end="", flush=True) |
| 124 | |
| 125 | # Heartbeat thread |
| 126 | heartbeat_stop = threading.Event() |
| 127 | |
| 128 | def _heartbeat(): |
| 129 | while not heartbeat_stop.is_set(): |
| 130 | heartbeat_stop.wait(5) |
| 131 | if not heartbeat_stop.is_set(): |
| 132 | elapsed = time.time() - start_time |
| 133 | print(f" {elapsed:.0f}s...", end="", flush=True) |
| 134 | |
| 135 | hb_thread = threading.Thread(target=_heartbeat, daemon=True) |
| 136 | hb_thread.start() |
| 137 | |
| 138 | try: |
| 139 | result = requests.post(url, headers=headers, json=payload, timeout=300).json() |
| 140 | finally: |
| 141 | heartbeat_stop.set() |
| 142 | hb_thread.join(timeout=1) |
| 143 | |
| 144 | elapsed = time.time() - start_time |
| 145 | print(f"\n [DONE] Image generated ({elapsed:.1f}s)") |
| 146 | |
| 147 | if result.get("choices"): |
| 148 | message = result["choices"][0]["message"] |
| 149 | image_uri = _message_image_uri(message) |
| 150 | if image_uri: |
| 151 | image_data, content_type = decode_data_uri(image_uri) |
| 152 | path = resolve_output_path(prompt, output_dir, filename, ".png") |
| 153 | return save_image_bytes(image_data, path, content_type) |
| 154 | |
| 155 | raise RuntimeError("No image was generated. The server may have refused the request.") |
| 156 | |
| 157 | |
| 158 | # ╔══════════════════════════════════════════════════════════════════╗ |
| 159 | # ║ Public Entry Point ║ |
| 160 | # ╚══════════════════════════════════════════════════════════════════╝ |
| 161 | |
| 162 | def generate(prompt: str, |
| 163 | aspect_ratio: str = "1:1", image_size: str = "1K", |
| 164 | output_dir: str = None, filename: str = None, |
| 165 | model: str = None, max_retries: int = MAX_RETRIES) -> str: |
| 166 | """ |
| 167 | OpenRouter image generation with automatic retry. |
| 168 | |
| 169 | Reads credentials from the current process environment or a `.env` file: |
| 170 | OPENROUTER_API_KEY |
| 171 | OPENROUTER_BASE_URL |
| 172 | OPENROUTER_MODEL (optional override) |
| 173 | """ |
| 174 | api_key = os.environ.get("OPENROUTER_API_KEY") |
| 175 | base_url = os.environ.get("OPENROUTER_BASE_URL") |
| 176 | |
| 177 | if not api_key: |
| 178 | raise ValueError( |
| 179 | "No API key found. Set OPENROUTER_API_KEY in the current environment or a .env file." |
| 180 | ) |
| 181 | |
| 182 | if model is None: |
| 183 | model = os.environ.get("OPENROUTER_MODEL") or DEFAULT_MODEL |
| 184 | |
| 185 | image_size = normalize_image_size(image_size) |
| 186 | |
| 187 | if aspect_ratio not in VALID_ASPECT_RATIOS: |
| 188 | raise ValueError(f"Invalid aspect ratio '{aspect_ratio}'. Valid: {VALID_ASPECT_RATIOS}") |
| 189 | |
| 190 | if image_size not in VALID_IMAGE_SIZES: |
| 191 | raise ValueError(f"Invalid image size '{image_size}'. Valid: {VALID_IMAGE_SIZES}") |
| 192 | |
| 193 | last_error = None |
| 194 | for attempt in range(max_retries + 1): |
| 195 | try: |
| 196 | return _generate_image(api_key, prompt, |
| 197 | aspect_ratio, image_size, output_dir, |
| 198 | filename, model, base_url) |
| 199 | except Exception as e: |
| 200 | last_error = e |
| 201 | if attempt < max_retries and is_rate_limit_error(e): |
| 202 | delay = retry_delay(attempt, rate_limited=True) |
| 203 | print(f"\n [WARN] Rate limit hit (attempt {attempt + 1}/{max_retries + 1}). " |
| 204 | f"Waiting {delay}s before retry...") |
| 205 | time.sleep(delay) |
| 206 | elif attempt < max_retries: |
| 207 | delay = retry_delay(attempt, rate_limited=False) |
| 208 | print(f"\n [WARN] Error (attempt {attempt + 1}/{max_retries + 1}): {e}. " |
| 209 | f"Retrying in {delay}s...") |
| 210 | time.sleep(delay) |
| 211 | else: |
| 212 | break |
| 213 | |
| 214 | raise RuntimeError(f"Failed after {max_retries + 1} attempts. Last error: {last_error}") |
| 215 |