Upload app.py

app.py

@@ -0,0 +1,132 @@
+# -*- coding: utf-8 -*-
+"""app
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/1ui1WeuYj77BJ2N4KZNWcxSgIwfOsWxFu
+"""
+
+# -*- coding: utf-8 -*-
+"""Gannett Fact Checker"""
+
+import os
+import requests
+import openai
+import random
+import gradio as gr
+from openai import OpenAI
+
+# Fetch your API keys securely from environment variables
+openai_api_key = os.getenv("OPENAI_API_KEY")
+serpapi_api_key = os.getenv("SERPAPI_API_KEY")
+
+# Initialize the client and set the API key
+client = OpenAI(api_key=openai_api_key)
+
+# Function to get latest news using SerpAPI
+def get_latest_news(topic):
+    search_url = f"https://serpapi.com/search?api_key={serpapi_api_key}&q={topic}&tbm=nws"
+    response = requests.get(search_url).json()
+    articles = response.get('news_results', [])
+
+    results = []
+    for article in articles:
+        results.append({
+            'title': article.get('title'),
+            'link': article.get('link')
+        })
+
+    return results
+
+def add_confidence_score(is_fake, no_evidence, model_score):
+    if no_evidence:
+        confidence_score = round(random.uniform(0.0, 20.0), 2)  # Very low confidence for no evidence
+        meter_status = "No Evidence"
+    elif is_fake:
+        confidence_score = round(random.uniform(0.0, 30.0), 2)  # Low confidence for fake
+        meter_status = "Fake"
+    else:
+        # Use the model's score if not flagged as fake or lacking evidence
+        confidence_score = model_score if model_score is not None else round(random.uniform(75.0, 99.0), 2)
+        meter_status = "Real"
+
+    return confidence_score, meter_status
+
+def summarize_news_and_check_authenticity(news_articles, topic):
+    if not news_articles:  # If there are no articles, handle it here
+        return "No relevant articles were found for this topic.", 0.0, "No Evidence"
+
+    # Prepare the prompt with the news articles
+    articles_text = "\n".join([f"{article['title']}: {article['link']}" for article in news_articles])
+
+    prompt = (f"Here are some news articles:\n{articles_text}\n\n"
+              f"Based on the provided articles, evaluate whether they support or contradict the following topic: '{topic}'. "
+              f"Summarize the content of these articles and determine if the topic is accurate or speculative. "
+              f"Provide a confidence score in percentage (0 to 100%) reflecting how well the articles align with the topic. "
+              f"Explain your reasoning for the confidence score given, and adjust the score according to how many articles support or contradict the topic. "
+              f"A higher confidence score should be given if the majority of the articles align with the topic, and a lower confidence score if they do not.")
+
+    # Call the OpenAI API for completion
+    response = client.chat.completions.create(
+        model="gpt-4",
+        messages=[
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": prompt}
+        ]
+    )
+
+    result = response.choices[0].message.content
+
+    # Extract confidence score from the response (if the model provides it)
+    try:
+        # Expecting confidence score to be presented as a percentage
+        model_score = float(result.split("confidence score of ")[-1].split("%")[0].strip())
+    except (IndexError, ValueError):
+        model_score = None  # Default if the score isn't parsed correctly
+
+    # Determine if the content is flagged as fake or if no evidence is found
+    is_fake = "fake" in result.lower() or "not true" in result.lower() or "speculative" in result.lower()
+    no_evidence = "no evidence" in result.lower() or "no articles" in result.lower()
+
+    # Add confidence score and status based on the outcome
+    confidence_score, meter_status = add_confidence_score(is_fake, no_evidence, model_score)
+
+    # Create a clean, unformatted result
+    result_display = (
+        f"Summary and Authenticity Check:\n{result}"
+    )
+
+    return result_display, confidence_score, meter_status
+
+
+# Function to combine everything for Gradio UI
+def get_news_and_check_authenticity(topic):
+    news_results = get_latest_news(topic)
+
+    # Get summary and authenticity check
+    summary_and_authenticity, confidence_score, meter_status = summarize_news_and_check_authenticity(news_results, topic)
+
+    # Create a formatted string to show the results
+    news_display = "\n".join([f"Title: {article['title']} \nLink: {article['link']}" for article in news_results])
+    result_display = (
+        f"News Results:\n{news_display}\n\n"
+        f"{summary_and_authenticity}\n\n"
+        f"Confidence Score: {confidence_score:.2f}%\n"
+        f"Fact Check: {meter_status}"
+    )
+
+    return result_display
+
+# Create the Gradio interface
+interface = gr.Interface(
+    fn=get_news_and_check_authenticity,
+    inputs="text",  # Users will input the topic in a text box
+    outputs="text",  # The result will be displayed as text
+    title="Fact Checker",
+    description="Enter a topic to check the latest news and verify its authenticity."
+)
+
+# Launch the Gradio interface
+interface.launch(share=True)
+