Spaces:
Running
Running
File size: 18,328 Bytes
7d391cb 022f94e 7d391cb 132848a 7d391cb | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 | import os
from dotenv import load_dotenv
# Load for local testing only
if os.path.exists(".env"):
load_dotenv()
import streamlit as st
import pandas as pd
import plotly.express as px
import time
from datetime import datetime
# Initialize database connections properly
# Assuming database uses Supabase, if Supabase client init fails due to missing secrets, app will handle gracefully
import modules.database as db
from modules.etl import load_and_validate, engineer_features
from modules.detection import apply_detection
from modules.risk_profiling import build_customer_profiles, assign_kyc_tier
import modules.visualizations as viz
from modules.ai_agent import stream_compliance_report
from modules.pdf_report import build_pdf
st.set_page_config(page_title="AML Shield", page_icon="π‘οΈ", layout="wide", initial_sidebar_state="expanded")
# --- CSS Overrides ---
st.markdown("""
<style>
.stMetric { background-color: #1e1e2d; padding: 15px; border-radius: 5px; }
h1, h2, h3 { color: #f8f9fa; }
</style>
""", unsafe_allow_html=True)
# --- Sidebar ---
st.sidebar.title("π‘οΈ AML Shield Navigation")
tabs = ["Upload & Analyze", "Dashboard", "Customer Profiles", "AI Report", "Global Analytics", "About"]
page = st.sidebar.radio("Go to", tabs)
# --- Helper logic for analysis ---
def run_pipeline(file_obj, filename="Uploaded Data"):
progress_bar = st.progress(0)
status_text = st.empty()
# Step 1
status_text.text("[βββββ] Loading & validating data...")
df, msg = load_and_validate(file_obj)
if df is None:
st.error(msg)
progress_bar.empty()
status_text.empty()
return False
progress_bar.progress(25)
time.sleep(0.5)
# Step 2
status_text.text("[βββββ] Engineering features...")
df = engineer_features(df)
progress_bar.progress(50)
time.sleep(0.5)
# Step 3
status_text.text("[βββββ] Running anomaly detection...")
df = apply_detection(df)
progress_bar.progress(75)
time.sleep(0.5)
# Step 4
status_text.text("[βββββ] Building customer profiles...")
profile_df = build_customer_profiles(df)
profile_df = assign_kyc_tier(profile_df)
progress_bar.progress(100)
time.sleep(0.5)
status_text.empty()
progress_bar.empty()
# Summary Metrics
total_tx = len(df)
flagged = df['is_flagged'].sum()
high_risk = len(df[df['risk_level'] == 'High'])
med_risk = len(df[df['risk_level'] == 'Medium'])
avg_score = df['risk_score'].mean()
date_range = f"{df['timestamp'].dt.date.min()} to {df['timestamp'].dt.date.max()}"
# Structuring & Intl stats for report
struct_attempts = profile_df['structuring_attempts'].sum()
intl_high = len(df[(df['is_international'] == 1) & (df['amount'] > 25000)])
kyc_counts = profile_df['kyc_tier'].value_counts().to_dict()
# Top flagged rules
rules_flat = [rule for sublist in df['rule_flags'] if isinstance(sublist, list) for rule in sublist]
top_rules = pd.Series(rules_flat).value_counts().head(3).to_dict() if rules_flat else {}
top_customers = profile_df.sort_values('avg_risk_score', ascending=False)['customer_id'].head(3).tolist()
# Save Upload to DB
upload_id = db.save_upload(
filename=filename, total=total_tx, flagged=flagged,
high_risk=high_risk, medium_risk=med_risk,
avg_score=avg_score, date_range=date_range
)
if upload_id:
# Batch insert chunks
db.save_transactions(df, upload_id)
db.save_customer_profiles(profile_df, upload_id)
summary_data = {
"filename": filename,
"total_transactions": int(total_tx),
"flagged_count": int(flagged),
"high_risk_count": int(high_risk),
"medium_risk_count": int(med_risk),
"avg_risk_score": float(avg_score),
"date_range": date_range,
"structuring_attempts": int(struct_attempts),
"international_high_value_count": int(intl_high),
"kyc_tier_breakdown": kyc_counts,
"top_rules_triggered": top_rules,
"top_flagged_customers": top_customers
}
# Session State
st.session_state.df_raw = df.copy()
st.session_state.df_scored = df.copy()
st.session_state.profile_df = profile_df.copy()
st.session_state.upload_id = upload_id
st.session_state.summary_data = summary_data
st.session_state.ai_report = None
st.success(f"β
{total_tx} transactions analyzed | β οΈ {flagged} flagged | π΄ {high_risk} high risk | π Avg risk score: {avg_score:.1f}")
# --- Model Evaluation (if is_fraud exists) ---
if "is_fraud" in df.columns:
st.markdown("---")
st.subheader("π― Model Performance Evaluation")
# Normalize fraud labels
df["is_fraud_numeric"] = df["is_fraud"].astype(str).str.upper().map(
{"TRUE": 1, "FALSE": 0, "1": 1, "0": 0, "1.0": 1, "0.0": 0}
).fillna(0).astype(int)
from sklearn.metrics import classification_report, confusion_matrix
import seaborn as sns
import matplotlib.pyplot as plt
# ml_anomaly_flag comes from detection.py (it's 1 if scoring high)
# In our case, the 'is_flagged' column is the prediction
y_true = df["is_fraud_numeric"]
y_pred = df["is_flagged"]
report = classification_report(y_true, y_pred, output_dict=True)
c1, c2, c3 = st.columns(3)
c1.metric("Accuracy", f"{report['accuracy']:.2%}")
c2.metric("Precision (Fraud)", f"{report['1']['precision']:.2%}")
c3.metric("Recall (Fraud)", f"{report['1']['recall']:.2%}")
with st.expander("Detailed Classification Report"):
st.code(classification_report(y_true, y_pred))
st.markdown("---")
st.subheader("Top 5 Highest Risk Transactions Preview")
preview = df.sort_values('risk_score', ascending=False).head(5)
cols = ['transaction_id', 'customer_id', 'amount', 'transaction_type', 'risk_score', 'risk_level', 'rule_flags']
st.dataframe(preview[cols])
return True
# --- PAGE ROUTING ---
if page == "Upload & Analyze":
st.title("π‘οΈ AML Shield")
st.write("AI-Powered Anti-Money Laundering Transaction Intelligence Platform")
col1, col2 = st.columns(2)
with col1:
uploaded_file = st.file_uploader("Upload CSV Transactions", type=['csv'])
if uploaded_file is not None:
if st.button("Analyze Uploaded File"):
run_pipeline(uploaded_file, filename=uploaded_file.name)
with col2:
st.write("Or test with pre-generated synthetic data:")
if st.button("Use Sample Dataset"):
sample_path = "sample_data/sample_transactions.csv"
if os.path.exists(sample_path):
run_pipeline(sample_path, filename="sample_transactions.csv")
else:
st.error("Sample dataset not found. Please ensure it was generated.")
elif page == "Dashboard":
if 'df_scored' not in st.session_state:
st.warning("Please upload or load sample data first in the 'Upload & Analyze' tab.")
else:
df = st.session_state.df_scored.copy()
summ = st.session_state.summary_data
# Dashboard Filters in Sidebar
st.sidebar.markdown("---")
st.sidebar.subheader("Dashboard Filters")
risk_filter = st.sidebar.multiselect("Risk Level", options=['High', 'Medium', 'Low'], default=['High', 'Medium', 'Low'])
type_filter = st.sidebar.multiselect("Transaction Type", options=df['transaction_type'].unique(), default=df['transaction_type'].unique())
min_date = df['timestamp'].min().date()
max_date = df['timestamp'].max().date()
date_filter = st.sidebar.slider("Date Range", min_value=min_date, max_value=max_date, value=(min_date, max_date))
# Apply filters
df_filtered = df[
(df['risk_level'].isin(risk_filter)) &
(df['transaction_type'].isin(type_filter)) &
(df['timestamp'].dt.date >= date_filter[0]) &
(df['timestamp'].dt.date <= date_filter[1])
]
# KPIs
c1, c2, c3, c4 = st.columns(4)
c1.metric("Total Transactions", summ['total_transactions'])
flagged_pct = (summ['flagged_count'] / summ['total_transactions']) * 100 if summ['total_transactions'] > 0 else 0
c2.metric("Flagged", summ['flagged_count'], delta=f"{flagged_pct:.1f}%")
c3.metric("High Risk", summ['high_risk_count'])
c4.metric("Avg Risk Score", f"{summ['avg_risk_score']:.1f}")
# Charts Row 1
r1c1, r1c2 = st.columns(2)
with r1c1:
st.subheader("Risk Distribution")
st.plotly_chart(viz.risk_distribution_chart(df_filtered), use_container_width=True)
with r1c2:
st.subheader("Daily Flagged Transactions")
st.plotly_chart(viz.flagged_transactions_timeline(df_filtered), use_container_width=True)
# Charts Row 2
st.subheader("Amount vs Risk Score Scatter")
st.plotly_chart(viz.amount_vs_risk_scatter(df_filtered), use_container_width=True)
# Charts Row 3
r3c1, r3c2 = st.columns(2)
with r3c1:
st.subheader("Transaction Types (Flagged vs Clean)")
st.plotly_chart(viz.transaction_type_breakdown(df_filtered), use_container_width=True)
with r3c2:
st.subheader("Rule Trigger Frequency")
st.plotly_chart(viz.rule_trigger_frequency(df_filtered), use_container_width=True)
# Charts Row 4
st.subheader("Top Flagged Customers")
st.plotly_chart(viz.top_flagged_customers_chart(df_filtered), use_container_width=True)
# Table
st.subheader("Flagged Transactions Explorer")
flagged_df = df_filtered[df_filtered['is_flagged'] == 1].copy()
# Convert rule_flags list to string for display/CSV
flagged_df['rule_flags_str'] = flagged_df['rule_flags'].apply(lambda x: ", ".join(x) if isinstance(x, list) else str(x))
disp_cols = ['transaction_id', 'customer_id', 'amount', 'transaction_type', 'risk_score', 'risk_level', 'rule_flags_str']
st.dataframe(flagged_df[disp_cols])
csv_data = flagged_df[disp_cols].to_csv(index=False).encode('utf-8')
st.download_button("Download Flagged Transactions CSV", data=csv_data, file_name="flagged_transactions.csv", mime="text/csv")
elif page == "Customer Profiles":
if 'profile_df' not in st.session_state:
st.warning("Please upload data first to analyze customer profiles.")
else:
profile_df = st.session_state.profile_df.copy()
df = st.session_state.df_scored
st.title("Customer KYC Profiles")
col1, col2 = st.columns([1, 2])
with col1:
st.subheader("KYC Tier Distribution")
st.plotly_chart(viz.kyc_tier_distribution(profile_df), use_container_width=True)
with col2:
st.subheader("All Customer Profiles")
st.dataframe(profile_df)
st.markdown("---")
st.subheader("Customer Drill-down")
selected_cust = st.selectbox("Select Customer ID", options=profile_df['customer_id'].unique())
cust_profile = profile_df[profile_df['customer_id'] == selected_cust].iloc[0]
cust_tx = df[df['customer_id'] == selected_cust].sort_values('timestamp', ascending=False)
cust_flags = cust_tx[cust_tx['is_flagged'] == 1]
c1, c2, c3 = st.columns(3)
c1.metric("KYC Tier", cust_profile['kyc_tier'])
c2.metric("Total Volume", f"${cust_profile['total_volume']:,.2f}")
c3.metric("Avg Risk Score", f"{cust_profile['avg_risk_score']:.1f}")
st.write("### Transaction History")
st.dataframe(cust_tx[['transaction_id', 'timestamp', 'amount', 'transaction_type', 'risk_score', 'risk_level']])
st.write("### Repeated Suspicious Behavior")
if len(cust_flags) > 0:
st.dataframe(cust_flags[['transaction_id', 'amount', 'rule_flags']])
else:
st.write("None detected.")
elif page == "AI Report":
if 'summary_data' not in st.session_state:
st.warning("Please upload data first to generate an AI report.")
else:
st.title("π€ AI Compliance Report Generation")
summ = st.session_state.summary_data
st.info(f"**Dataset loaded:** {summ['filename']} | **Total Transactions:** {summ['total_transactions']} | **Flagged:** {summ['flagged_count']}")
if st.button("π€ Generate AI Compliance Report", type="primary"):
if not os.environ.get("BYTEZ_API_KEY"):
st.error("BYTEZ_API_KEY requires to be set to generate AI report.")
else:
with st.spinner("Connecting to AI analyst..."):
placeholder = st.empty()
report_text = stream_compliance_report(summ, placeholder)
if report_text and not report_text.startswith("Error"):
st.success("β
Report generated using meta-llama/Llama-3.1-8B-Instruct via Bytez")
st.session_state.ai_report = report_text
if st.session_state.upload_id:
db.save_ai_report(st.session_state.upload_id, report_text, "meta-llama/Llama-3.1-8B-Instruct")
if st.session_state.get('ai_report'):
st.markdown("---")
st.write("### Actions")
# PDF generation
flagged_df = st.session_state.df_scored[st.session_state.df_scored['is_flagged'] == 1].copy()
pdf_bytes = build_pdf(st.session_state.ai_report, summ, flagged_df)
date_str = datetime.now().strftime("%Y%m%d_%H%M")
st.download_button("π Download PDF Report", data=pdf_bytes, file_name=f"AML_Shield_Report_{date_str}.pdf", mime="application/pdf")
st.markdown("---")
st.markdown(st.session_state.ai_report)
elif page == "Global Analytics":
st.title("π Global Analytics")
with st.spinner("Fetching global stats from Supabase..."):
try:
stats = db.get_global_stats()
uploads = db.get_all_uploads()
uploads_df = pd.DataFrame(uploads)
except Exception as e:
st.error(f"Could not connect to Supabase: {e}")
stats = None
uploads_df = pd.DataFrame()
if stats:
c1, c2, c3, c4 = st.columns(4)
c1.metric("All-time Transactions", stats['total_transactions_ever'])
c2.metric("Total Uploads", stats['total_uploads'])
c3.metric("All-time Flagged", stats['total_flagged_ever'])
c4.metric("Global Avg Risk", f"{stats['avg_risk_score_global']:.1f}")
st.markdown("---")
col1, col2 = st.columns(2)
with col1:
st.subheader("Global Trend: Flagged per Upload")
if not uploads_df.empty:
if 'uploaded_at' in uploads_df.columns:
uploads_df['date'] = pd.to_datetime(uploads_df['uploaded_at']).dt.date
trend_df = uploads_df.groupby('date')['flagged_count'].sum().reset_index()
fig = px.line(trend_df, x='date', y='flagged_count', markers=True)
fig.update_traces(line_color=viz.COLOR_MED)
st.plotly_chart(viz.apply_theme(fig), use_container_width=True)
else:
st.write("No historical data available.")
with col2:
st.subheader("Most Common Rule Triggered")
st.info(stats.get('most_common_rule_triggered', 'N/A'))
st.write("*(Approximation based on available metric patterns)*")
st.subheader("Uploads History")
if not uploads_df.empty:
st.dataframe(uploads_df[['filename', 'uploaded_at', 'total_transactions', 'flagged_count', 'high_risk_count', 'avg_risk_score']])
elif page == "About":
st.title("βΉοΈ About AML Shield")
st.write("""
### AI-Powered Anti-Money Laundering Transaction Intelligence Platform
AML Shield is built to demonstrate production-grade AML compliance analytics skills for financial services roles.
#### How it works:
1. **Upload CSV** β ETL validation & pre-processing.
2. **Rule-based AML flags** β applied to all inputs.
3. **Isolation Forest ML** β anomaly detection logic.
4. **Risk scoring (0-100)** β deterministic algorithm based on flags+ML.
5. **KYC customer profiling** β KMeans clustering into tiers.
6. **LangChain + Bytez** β streams a formal regulatory compliance report utilizing meta-llama/Llama-3.1-8B-Instruct.
7. **ReportLab** β renders professional downloadable PDF.
8. **Supabase** β All data natively persisted.
""")
with st.expander("AML Rules Explained"):
st.write("""
- **Structuring**: Transactions intentionally sizing just beneath the $10,000 CTR reporting requirement ($9000 - $9999).
- **Rapid Fire Transactions**: Accounts showing an abnormally high transaction velocity.
- **Large Cash Out**: Immediate cash liquidations above $50,000.
- **Dormant Account Spike**: High amounts triggered by newly created or previously dormant accounts (< 30 days).
- **International High Value**: Large wire transfers sent outside of the domestic region.
- **Suspicious Round Amount**: High net round payments generally uncharacteristic of organic spending.
""")
st.write("""
**Regulatory Frameworks Considered:**
- BSA (Bank Secrecy Act)
- FinCEN SAR (Suspicious Activity Report) requirements
- FATF Recommendation 16 (Wire transfers)
""")
st.markdown("---")
st.write("**Tech Stack:** Streamlit | Pandas | Scikit-learn | Plotly | ReportLab | LangChain | Bytez | Supabase")
st.write("**Deployments:** Live on Hugging Face Spaces")
|