import requests
import json

def extract_text_from_image(image_path, api_key, model, task_type, max_tokens, temperature, top_p, repetition_penalty, pages=None):
    url = "https://api.opentyphoon.ai/v1/ocr"

    with open(image_path, 'rb') as file:
        files = {'file': file}
        data = {
            'model': model,
            'task_type': task_type,
            'max_tokens': str(max_tokens),
            'temperature': str(temperature),
            'top_p': str(top_p),
            'repetition_penalty': str(repetition_penalty)
        }

        if pages:
            data['pages'] = json.dumps(pages)

        headers = {
            'Authorization': f'Bearer {api_key}'
        }

        response = requests.post(url, files=files, data=data, headers=headers)

        if response.status_code == 200:
            result = response.json()

            # Extract text from successful results
            extracted_texts = []
            for page_result in result.get('results', []):
                if page_result.get('success') and page_result.get('message'):
                    content = page_result['message']['choices'][0]['message']['content']
                    try:
                        # Try to parse as JSON if it's structured output
                        parsed_content = json.loads(content)
                        text = parsed_content.get('natural_text', content)
                    except json.JSONDecodeError:
                        text = content
                    extracted_texts.append(text)
                elif not page_result.get('success'):
                    print(f"Error processing {page_result.get('filename', 'unknown')}: {page_result.get('error', 'Unknown error')}")

            return '\n'.join(extracted_texts)
        else:
            print(f"Error: {response.status_code}")
            print(response.text)
            return None

# Usage
api_key = "sk-SJLOZn2kGqDX2zuO9doBAwb1alfocOLYQSM9SkiU2HW3KpQC"
image_path = "path/to/your/image.jpg"  # or path/to/your/document.pdf
model = "typhoon-ocr"
task_type = "default"
max_tokens = 16384
temperature = 0.1
top_p = 0.6
repetition_penalty = 1.2
pages = None
extracted_text = extract_text_from_image(image_path, api_key, model, task_type, max_tokens, temperature, top_p, repetition_penalty, pages)
print(extracted_text)