eye.py

"""
This module provides a function for parsing legal citations for factual accuracy and Legal Bluebook compliance.

The main function, `gpt4_parse_citations`, takes a list of legal citations as input and uses the GPT-4 API to parse the citations. It checks for factual accuracy by verifying if the citations contain the correct year of the case, the correct court, parties, reporter, volume, pages, etc. It also checks for compliance with the 21st edition of the Legal Bluebook.

If the GPT-4 API request fails, the function returns None.

Note: This module requires the `client` object to be initialized before calling the `gpt4_parse_citations` function.

Example usage:
    citations = ["Smith v. Johnson, 2022 WL 1234567 (Supreme Court)"]
    result = gpt4_parse_citations(citations)
    print(result)
"""

# Import required modules
from eyecite import get_citations, resolve_citations
from eyecite.clean import clean_text
from openai import OpenAI
import requests
import threading
import logging

# Set up basic logging
logging.basicConfig(level=logging.INFO)

# Initialize the OpenAI client
client = OpenAI(api_key="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")
def gpt4_parse_citations(citations):
    """
    Parses the provided legal citations for factual accuracy and Legal Bluebook compliance.

    Args:
        citations (list): A list of legal citations to be parsed.

    Returns:
        str: The parsed content of the first completion choice from the GPT-4 API, or None if the API request fails.
    """
    messages = [
        {"role": "system", "content": "You are a helpful, fact-checking, legal assistant. Your task is to check the provided legal citations for factual accuracy (e.g.,  does the the citation contain the correct <year> of case, the correct <court>, <parties>, <reporter>, <volume>, <pages>, etc.) and for 21 edition Legal Bluebook compliance. If you do not know the correct information, tell the user you are not sure."}
    ] + [{"role": "user", "content": citation} for citation in citations]

    try:
        completion = client.chat.completions.create(
            model="gpt-4-1106-preview",
            messages=messages
        )
        return completion.choices[0].message.content
    except Exception as e:
        logging.error(f"GPT-4 API request failed: {e}")
        return None

# Function to fetch case data from the Court Listener API
def fetch_case_data_from_court_listener(citation):
    """
    Fetches case data from the Court Listener API for a given citation.

    Args:
        citation (str): The legal citation to fetch case data for.

    Returns:
        dict: The case data retrieved from the Court Listener API, or None if the API request fails.
    """
    url = f"https://www.courtlistener.com/api/rest/v3/search/?type=o&q={citation}"
    headers = {'Authorization': 'Token 974993b55ad4144adf83d3fe942abc7210e6e10a'}
    response = requests.get(url, headers=headers)
    if response.status_code == 200:
        return response.json()
    else:
        return None

# Function to handle each citation batch
def process_citations_batch(citations_batch, results):
    """
    Process a batch of citations.

    Args:
        citations_batch (list): A list of citations to process.
        results (list): A list to store the results of processing.

    Returns:
        None
    """
    for citation in citations_batch:
        court_listener_data = fetch_case_data_from_court_listener(str(citation))
        gpt4_feedback = gpt4_parse_citations([str(citation)])
        results.append((str(citation), gpt4_feedback, court_listener_data))

# Citation checker function that uses both Eyecite, GPT-4, and Court Listener API
def check_citations(text):
    """
    Check citations in the given text.

    Args:
        text (str): The text to check for citations.

    Returns:
        list: A list of results from processing the citations. Each result is a tuple containing the citation text, GPT-4 feedback, and Court Listener data.
    """
    cleaning_steps = []
    cleaned_text = clean_text(text, steps=cleaning_steps)
    citations = get_citations(cleaned_text)
    normalized_citations = resolve_citations(citations)

    if not isinstance(normalized_citations, list):
        normalized_citations = list(normalized_citations)

    results = []
    threads = []

    batch_size = 5
    for i in range(0, len(normalized_citations), batch_size):
        batch = normalized_citations[i:i + batch_size]
        thread = threading.Thread(target=process_citations_batch, args=(batch, results))
        threads.append(thread)
        thread.start()

    for thread in threads:
        thread.join()

    return results

# Example usage
example_text = "As held in Roe v. Wade, 410 U.S. 113 (1973), privacy rights are fundamental. See also Planned Parenthood v. Casey, 505 U.S. 833 (1993)."
citation_results = check_citations(example_text)

# Display the results
for citation_text, gpt4_feedback, court_listener_data in citation_results:
    print(f"Citation: {citation_text}\nGPT-4 Feedback: {gpt4_feedback}\nCourt Listener Data: {court_listener_data}\n")