Spaces:
Paused
Paused
| import os | |
| import time | |
| from pathlib import Path | |
| from dotenv import load_dotenv | |
| from google import genai | |
| from google.genai import errors, types | |
| # ============================================================ | |
| # Config | |
| # ============================================================ | |
| load_dotenv() | |
| MODELS = [ | |
| "gemini-3.6-flash", | |
| "gemini-3.5-flash", | |
| "gemini-3.5-flash-lite", | |
| "gemini-flash-lite-latest", | |
| ] | |
| MAX_QUESTION_LENGTH = 300 | |
| SYSTEM_PROMPT = ( | |
| "You answer questions about a single photo. " | |
| "Base your answer only on what is visible in the image. " | |
| "If the answer can't be seen in the photo, say so. " | |
| "Answer in one to three short sentences." | |
| ) | |
| GENERATION_CONFIG = types.GenerateContentConfig( | |
| system_instruction=SYSTEM_PROMPT, | |
| temperature=0.2, | |
| thinking_config=types.ThinkingConfig(thinking_level="low"), | |
| ) | |
| HTTP_OPTIONS = types.HttpOptions( | |
| timeout=10_000, | |
| retry_options=types.HttpRetryOptions(attempts=1), | |
| ) | |
| client = genai.Client(http_options=HTTP_OPTIONS) if os.getenv("GEMINI_API_KEY") else None | |
| # ============================================================ | |
| # Q&A | |
| # ============================================================ | |
| def ask_about_image(image_path: str, question: str) -> str: | |
| question = question.strip() | |
| if not question: | |
| return "Type a question about the image first." | |
| if len(question) > MAX_QUESTION_LENGTH: | |
| return f"Please keep the question under {MAX_QUESTION_LENGTH} characters." | |
| if client is None: | |
| return "The Q&A model isn't configured: GEMINI_API_KEY is missing." | |
| image_part = types.Part.from_bytes( | |
| data=Path(image_path).read_bytes(), | |
| mime_type="image/jpeg", | |
| ) | |
| for model in MODELS: | |
| start = time.perf_counter() | |
| try: | |
| response = client.models.generate_content( | |
| model=model, | |
| contents=[image_part, question], | |
| config=GENERATION_CONFIG, | |
| ) | |
| print(f"[qa] {model}: OK in {time.perf_counter() - start:.1f}s") | |
| return response.text.strip() | |
| except errors.ServerError as e: | |
| print(f"[qa] {model}: {e.code} after {time.perf_counter() - start:.1f}s, trying next") | |
| continue | |
| except errors.ClientError as e: | |
| print(f"[qa] {model}: {e.code} after {time.perf_counter() - start:.1f}s") | |
| if e.code in (404, 429): | |
| continue | |
| return f"The request was rejected ({e.code}). Please try a different question." | |
| except Exception as e: | |
| print(f"[qa] {model}: {type(e).__name__} after {time.perf_counter() - start:.1f}s, trying next") | |
| continue | |
| return "The model is busy right now. Please try again in a moment." | |
| # ============================================================ | |
| # Manual test | |
| # ============================================================ | |
| if __name__ == "__main__": | |
| import sys | |
| path = "data/Flicker8k_Dataset/1000268201_693b08cb0e.jpg" | |
| question = " ".join(sys.argv[1:]) or "What is the girl wearing?" | |
| print("Q:", question) | |
| start = time.perf_counter() | |
| answer = ask_about_image(path, question) | |
| print("A:", answer) | |
| print(f"Total: {time.perf_counter() - start:.1f}s") |