Update app.py
Browse files
app.py
CHANGED
|
@@ -3,28 +3,30 @@ from newspaper import Article
|
|
| 3 |
from modules.online_search import search_online
|
| 4 |
from modules.validation import calculate_truthfulness_score
|
| 5 |
from modules.knowledge_graph import search_kg
|
| 6 |
-
from modules.generate_explanation import generate_explanation
|
| 7 |
from dotenv import load_dotenv
|
| 8 |
import os
|
| 9 |
from concurrent.futures import ThreadPoolExecutor
|
|
|
|
| 10 |
|
| 11 |
-
# Load environment variables
|
| 12 |
load_dotenv()
|
| 13 |
|
| 14 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
KG_INDEX_PATH = "KG/news_category_index.faiss"
|
| 16 |
KG_DATASET_PATH = "KG/News_Category_Dataset_v3.json"
|
| 17 |
SEARCH_API_KEY = os.getenv("SEARCH_API_KEY")
|
| 18 |
SEARCH_BASE_URL = os.getenv("SEARCH_BASE_URL")
|
| 19 |
SEARCH_MODEL = os.getenv("SEARCH_MODEL")
|
| 20 |
|
| 21 |
-
# Initialize ThreadPoolExecutor
|
| 22 |
-
executor = ThreadPoolExecutor(max_workers=3) # Increased workers to accommodate explanation task
|
| 23 |
|
| 24 |
-
# Function to process
|
| 25 |
def evaluate_news(news_input):
|
| 26 |
-
|
| 27 |
-
yield "**Processing... Please wait while we analyze the information.** β³"
|
| 28 |
|
| 29 |
# Handle URL input
|
| 30 |
if news_input.startswith("http"):
|
|
@@ -37,94 +39,120 @@ def evaluate_news(news_input):
|
|
| 37 |
yield f"**Error processing the URL:** {str(e)}"
|
| 38 |
return
|
| 39 |
else:
|
| 40 |
-
# Direct text input
|
| 41 |
news_text = news_input
|
| 42 |
|
| 43 |
try:
|
| 44 |
-
# Run
|
| 45 |
future_kg = executor.submit(search_kg, news_text, KG_INDEX_PATH, KG_DATASET_PATH)
|
| 46 |
future_online = executor.submit(search_online, news_text, SEARCH_API_KEY, SEARCH_BASE_URL, SEARCH_MODEL)
|
| 47 |
|
| 48 |
-
# Wait for
|
| 49 |
kg_content = future_kg.result()
|
| 50 |
online_search_results = future_online.result()
|
| 51 |
|
| 52 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
context = online_search_results['message_content'] + '\n' + kg_content + '\n' + 'Device set to use cpu'
|
| 54 |
-
# print(context) # Debug log
|
| 55 |
|
| 56 |
-
#
|
| 57 |
truth_score = calculate_truthfulness_score(info=news_text, context=context)
|
|
|
|
| 58 |
|
| 59 |
-
# Determine
|
| 60 |
if truth_score > 0.7:
|
| 61 |
-
status = "likely true"
|
| 62 |
-
recommendation = "You can reasonably trust this information, but further verification is always recommended for critical decisions."
|
| 63 |
elif truth_score > 0.4:
|
| 64 |
-
status = "uncertain"
|
| 65 |
-
recommendation = "This information might be partially true, but additional investigation is required before accepting it as fact."
|
| 66 |
else:
|
| 67 |
-
status = "unlikely to be true"
|
| 68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
-
# Display initial
|
| 71 |
-
result = f"**News**: \"{news_text[:300]}...\"\n\n"
|
| 72 |
-
result += f"**Truthfulness Score**: {truth_score:.2f} (**{status.capitalize()}**)\n\n"
|
| 73 |
-
result += f"**Analysis**: {recommendation}\n\n"
|
| 74 |
-
yield result # Immediately display score and recommendation
|
| 75 |
|
| 76 |
# Generate explanation asynchronously
|
| 77 |
future_explanation = executor.submit(generate_explanation, news_text, context, truth_score)
|
|
|
|
| 78 |
|
| 79 |
-
# Add explanation and sources once available
|
| 80 |
-
explanation = future_explanation.result() # Wait for explanation result
|
| 81 |
if explanation:
|
| 82 |
-
result += f"**Explanation
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
for i, source in enumerate(sources[:5]):
|
| 90 |
-
result += f"{i + 1}. {source}\n"
|
| 91 |
-
result += "\n*Please make sure to do your own research for more confirmation and to cross-check the information.*"
|
| 92 |
|
| 93 |
-
|
| 94 |
|
| 95 |
except Exception as e:
|
| 96 |
-
yield f"**Error
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
|
| 98 |
|
| 99 |
# Gradio Interface
|
| 100 |
-
with gr.Blocks(
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
gr.Markdown("### **About EchoTruth**")
|
| 128 |
-
gr.Markdown("EchoTruth uses AI to help users verify news authenticity in real-time.
|
| 129 |
-
|
| 130 |
-
demo.launch()
|
|
|
|
| 3 |
from modules.online_search import search_online
|
| 4 |
from modules.validation import calculate_truthfulness_score
|
| 5 |
from modules.knowledge_graph import search_kg
|
| 6 |
+
from modules.generate_explanation import generate_explanation
|
| 7 |
from dotenv import load_dotenv
|
| 8 |
import os
|
| 9 |
from concurrent.futures import ThreadPoolExecutor
|
| 10 |
+
from modules.record import DatabaseComponent # Import DatabaseComponent
|
| 11 |
|
| 12 |
+
# Load environment variables
|
| 13 |
load_dotenv()
|
| 14 |
|
| 15 |
+
# Initialize database and executor
|
| 16 |
+
db = DatabaseComponent()
|
| 17 |
+
executor = ThreadPoolExecutor(max_workers=3)
|
| 18 |
+
|
| 19 |
+
# Constants for file paths and API keys
|
| 20 |
KG_INDEX_PATH = "KG/news_category_index.faiss"
|
| 21 |
KG_DATASET_PATH = "KG/News_Category_Dataset_v3.json"
|
| 22 |
SEARCH_API_KEY = os.getenv("SEARCH_API_KEY")
|
| 23 |
SEARCH_BASE_URL = os.getenv("SEARCH_BASE_URL")
|
| 24 |
SEARCH_MODEL = os.getenv("SEARCH_MODEL")
|
| 25 |
|
|
|
|
|
|
|
| 26 |
|
| 27 |
+
# Function to process and verify news
|
| 28 |
def evaluate_news(news_input):
|
| 29 |
+
yield "**Processing... Please wait.** β³"
|
|
|
|
| 30 |
|
| 31 |
# Handle URL input
|
| 32 |
if news_input.startswith("http"):
|
|
|
|
| 39 |
yield f"**Error processing the URL:** {str(e)}"
|
| 40 |
return
|
| 41 |
else:
|
|
|
|
| 42 |
news_text = news_input
|
| 43 |
|
| 44 |
try:
|
| 45 |
+
# Run search tasks concurrently
|
| 46 |
future_kg = executor.submit(search_kg, news_text, KG_INDEX_PATH, KG_DATASET_PATH)
|
| 47 |
future_online = executor.submit(search_online, news_text, SEARCH_API_KEY, SEARCH_BASE_URL, SEARCH_MODEL)
|
| 48 |
|
| 49 |
+
# Wait for results
|
| 50 |
kg_content = future_kg.result()
|
| 51 |
online_search_results = future_online.result()
|
| 52 |
|
| 53 |
+
# Extract citations from the search results
|
| 54 |
+
citations = online_search_results.get("citations", []) # List of sources
|
| 55 |
+
first_citation = citations[0] if citations else "N/A" # Store first citation in DB
|
| 56 |
+
|
| 57 |
+
# Combine context
|
| 58 |
context = online_search_results['message_content'] + '\n' + kg_content + '\n' + 'Device set to use cpu'
|
|
|
|
| 59 |
|
| 60 |
+
# Compute truth score
|
| 61 |
truth_score = calculate_truthfulness_score(info=news_text, context=context)
|
| 62 |
+
truth_percentage = truth_score * 100 # Convert to percentage
|
| 63 |
|
| 64 |
+
# Determine truth status
|
| 65 |
if truth_score > 0.7:
|
| 66 |
+
status = f"**{truth_percentage:.0f}% chances to be true** - This news is likely true."
|
|
|
|
| 67 |
elif truth_score > 0.4:
|
| 68 |
+
status = f"**{truth_percentage:.0f}% chances to be true** - This news is uncertain. Please verify further."
|
|
|
|
| 69 |
else:
|
| 70 |
+
status = f"**{truth_percentage:.0f}% chances to be true** - This news is unlikely to be true. Proceed with caution."
|
| 71 |
+
|
| 72 |
+
# Save result in database
|
| 73 |
+
db.save_news_verification(news_text[:100], truth_score, first_citation)
|
| 74 |
+
|
| 75 |
+
# Initial result
|
| 76 |
+
result = f"**News:** \"{news_text[:300]}...\"\n\n"
|
| 77 |
+
result += f"**Truthfulness Score:** {truth_score:.2f} ({status})\n\n"
|
| 78 |
|
| 79 |
+
yield result # Display initial results
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
|
| 81 |
# Generate explanation asynchronously
|
| 82 |
future_explanation = executor.submit(generate_explanation, news_text, context, truth_score)
|
| 83 |
+
explanation = future_explanation.result()
|
| 84 |
|
|
|
|
|
|
|
| 85 |
if explanation:
|
| 86 |
+
result += f"**Explanation:** {explanation}\n\n"
|
| 87 |
|
| 88 |
+
# Display sources
|
| 89 |
+
if citations:
|
| 90 |
+
result += "\n**Sources & References:**\n"
|
| 91 |
+
for i, source in enumerate(citations[:5]): # Show up to 5 sources
|
| 92 |
+
result += f"{i + 1}. [{source}]({source})\n"
|
|
|
|
|
|
|
|
|
|
| 93 |
|
| 94 |
+
yield result # Final output with sources
|
| 95 |
|
| 96 |
except Exception as e:
|
| 97 |
+
yield f"**Error:** {str(e)}"
|
| 98 |
+
|
| 99 |
+
|
| 100 |
+
# Function to fetch dashboard data
|
| 101 |
+
def fetch_dashboard_data():
|
| 102 |
+
total_news = db.get_total_news_count()
|
| 103 |
+
last_10_news = db.get_last_10_news()
|
| 104 |
+
|
| 105 |
+
# Generate table-style layout for recent verifications
|
| 106 |
+
dashboard_info = f"**Total News Verified:** {total_news}\n\n"
|
| 107 |
+
|
| 108 |
+
if last_10_news:
|
| 109 |
+
table = "| # | News Title | Score (%) | Date Verified | Citation |\n"
|
| 110 |
+
table += "|---|------------|-----------|--------------|----------|\n"
|
| 111 |
+
|
| 112 |
+
for i, news in enumerate(last_10_news, start=1):
|
| 113 |
+
truth_percentage = news['score'] * 100 # Convert to percentage
|
| 114 |
+
citation = f"[Source]({news['citation']})" if news['citation'] != "N/A" else "N/A"
|
| 115 |
+
table += f"| {i} | {news['title'][:50]}... | {truth_percentage:.0f}% | {news['timestamp']} | {citation} |\n"
|
| 116 |
+
|
| 117 |
+
dashboard_info += table
|
| 118 |
+
else:
|
| 119 |
+
dashboard_info += "_No records found._"
|
| 120 |
+
|
| 121 |
+
return dashboard_info
|
| 122 |
|
| 123 |
|
| 124 |
# Gradio Interface
|
| 125 |
+
with gr.Blocks(css="""
|
| 126 |
+
.gradio-container { font-family: 'Georgia', serif; font-size: 16px; }
|
| 127 |
+
h1, h2, h3 { font-family: 'Times New Roman', serif; }
|
| 128 |
+
table { width: 100%; border-collapse: collapse; }
|
| 129 |
+
th, td { padding: 10px; border: 1px solid #ddd; text-align: left; }
|
| 130 |
+
""") as demo:
|
| 131 |
+
with gr.Tabs() as tabs:
|
| 132 |
+
with gr.Tab("π Verify News"):
|
| 133 |
+
gr.Markdown("# π° EchoTruth: News Verification")
|
| 134 |
+
gr.Markdown("""
|
| 135 |
+
**How it Works:**
|
| 136 |
+
- Enter a news article **or** a URL.
|
| 137 |
+
- Click **Check Truthfulness**.
|
| 138 |
+
- Get a **truth score**, an explanation, and references.
|
| 139 |
+
""")
|
| 140 |
+
|
| 141 |
+
input_box = gr.Textbox(placeholder="Paste news text or URL...", label="News Input", lines=5)
|
| 142 |
+
submit_btn = gr.Button("Check Truthfulness")
|
| 143 |
+
output_box = gr.Markdown()
|
| 144 |
+
submit_btn.click(fn=evaluate_news, inputs=[input_box], outputs=[output_box])
|
| 145 |
+
|
| 146 |
+
with gr.Tab("π Dashboard") as dashboard_tab:
|
| 147 |
+
gr.Markdown("# π Verification Dashboard")
|
| 148 |
+
dashboard_output = gr.Markdown()
|
| 149 |
+
refresh_btn = gr.Button("π Refresh Data")
|
| 150 |
+
refresh_btn.click(fn=fetch_dashboard_data, inputs=[], outputs=[dashboard_output])
|
| 151 |
+
|
| 152 |
+
# Automatically refresh dashboard when the tab is opened
|
| 153 |
+
tabs.select(fn=fetch_dashboard_data, inputs=[], outputs=[dashboard_output])
|
| 154 |
|
| 155 |
gr.Markdown("### **About EchoTruth**")
|
| 156 |
+
gr.Markdown("EchoTruth uses AI to help users verify news authenticity in real-time.")
|
| 157 |
+
|
| 158 |
+
demo.launch()
|