Hugging Face's logo

Spaces:
Vigneshsaminathan
/
VQA
Sleeping

App Files Community
VQA
File size: 2,522 Bytes
2c5cb17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import base64
from google import genai
from config import GEMINI_API_KEY
import re

client = genai.Client(api_key=GEMINI_API_KEY)

def gemini_vision_answer(image, question, lang="en"):

    try:
        prompt = f"""Analyze this image carefully.

Question:
{question}

You MUST respond strictly in the language code: {lang}. All text in your output must be translated to '{lang}'. 
However, keep the structural English keys intact (Caption, Final Answer, Explanation).

Respond strictly following this exact structure without any deviations or markdown blocks.

Caption: <short caption>
Final Answer: <direct answer>
Explanation: <detailed reasoning>"""

        response = client.models.generate_content(
            model="gemini-3.1-flash-lite-preview",
            contents=[image, prompt]
        )

        text = response.text

        # Parse response using regex for better stability since LLMs can sometimes alter spacing
        caption_match = re.search(r'Caption:\s*(.*?)(?=Final Answer:|$)', text, re.IGNORECASE | re.DOTALL)
        answer_match = re.search(r'Final Answer:\s*(.*?)(?=Explanation:|$)', text, re.IGNORECASE | re.DOTALL)
        explanation_match = re.search(r'Explanation:\s*(.*)', text, re.IGNORECASE | re.DOTALL)

        caption = caption_match.group(1).strip() if caption_match else "Caption not generated correctly."
        answer = answer_match.group(1).strip() if answer_match else "Answer not generated correctly."
        explanation = explanation_match.group(1).strip() if explanation_match else "Explanation not generated correctly."

        # Filter out markdown bolds if they leaked from the LLM prompt completion
        caption = caption.replace("**", "")
        answer = answer.replace("**", "")

        return caption, answer, explanation

    except Exception as e:
        error_msg = str(e)
        if "429" in error_msg or "RESOURCE_EXHAUSTED" in error_msg:
            return (
                "Rate Limit Exceeded",
                "You have hit the free tier quota limit for Google Gemini.",
                "Please wait or check your Google AI Studio dashboard for quota resets."
            )
        print("GEMINI ERROR:", error_msg)
        return "Gemini Error", "Error", error_msg



def extract_section(text, section_name):
    try:
        start = text.index(section_name) + len(section_name)
        end = text.find("\n", start)
        if end == -1:
            end = len(text)
        return text[start:end].strip()
    except:
        return "Not found"