Prompt As A Service Sample Code

prompt_as_service_sample_code.py

from datasets import load_dataset

# Load the dataset from Hugging Face
dataset = load_dataset("kurry/sp500_constituents_10q")

# Placeholder list for companies with known accounting issues
companies_with_issues = ["CompanyA", "CompanyB"]  # This needs to be populated

# Analyze each company's 10-Q using Claude 2
true_positives = 0
false_positives = 0

def analyze_with_claude_2(ten_q_content):
    url = "https://api.promptperfect.jina.ai/your-api-endpoint"
    headers = {"Content-Type": "application/json"}
    response = requests.post(url, headers=headers, json={"parameters": {"earnings_call_transcript": ten_q_content}})
    if response.status_code == 200:
        return response.json()  # Assuming the response contains the risks identified
    else:
        print(response.text)
        return []

    # Dictionary of known accounting issues for each company
known_issues_dict = {
    "CompanyA": ["issue1", "issue2", ...],
    "CompanyB": ["issue1", ...],
    # ... add other companies and their known issues
}

for company in companies_with_issues:
    # Extracting the report content using the 'report' field
    ten_q_content = dataset['train'][company]['report']
    
    # Analyze using Claude 2
    risks_identified = analyze_with_claude_2(ten_q_content)
    
    # Retrieve known issues for the current company
    known_issues = known_issues_dict.get(company, [])
    
    # Compare with known issues
    matching_issues = [issue for issue in risks_identified if issue in known_issues]
    if matching_issues:
        true_positives += 1
    else:
        false_positives += 1

# Calculate precision
precision = true_positives / (true_positives + false_positives)
print(f"Precision: {precision * 100:.2f}%")