compliance rules into separate files. Small Gradio UI changes

app.py

@@ -5,45 +5,13 @@ from transformers import pipeline, DistilBertTokenizer, DistilBertForSequenceCla
 import numpy as np
 from PIL import Image
 import json
+from compliance_rules import ComplianceRules
 
 # Initialize OCR reader
 reader = easyocr.Reader(['en'])
 
-# Define compliance rules
-COMPLIANCE_RULES = {
-    "US_SEC": {
-        "required_disclaimers": [
-            "past performance",
-            "investment risks",
-            "regulatory statement"
-        ],
-        "prohibited_terms": [
-            "guaranteed returns",
-            "risk-free",
-            "sure thing"
-        ]
-    },
-    "UK_FCA": {
-        "required_disclaimers": [
-            "capital at risk",
-            "regulated by FCA"
-        ],
-        "prohibited_terms": [
-            "guaranteed profit",
-            "no risk"
-        ]
-    },
-    "EU": {
-        "required_disclaimers": [
-            "risk warning",
-            "regulatory information"
-        ],
-        "prohibited_terms": [
-            "assured returns",
-            "no losses"
-        ]
-    }
-}
+# Initialize compliance rules
+compliance_rules = ComplianceRules()
 
 def extract_text_from_image(image):
     """Extract text from image using EasyOCR"""
@@ -52,35 +20,62 @@ def extract_text_from_image(image):
 
 def check_compliance(text):
     """Check text for compliance across all regions"""
+    rules = compliance_rules.get_all_rules()
     report = {
         "compliant": True,
         "violations": [],
         "warnings": [],
         "channel_risks": {
-            "email": 0,
-            "social": 0,
-            "print": 0
+            "email": {"score": 0, "details": []},
+            "social": {"score": 0, "details": []},
+            "print": {"score": 0, "details": []}
         }
     }
     
-    # Check each region's rules
-    for region, rules in COMPLIANCE_RULES.items():
+    for region, region_rules in rules.items():
         # Check prohibited terms
-        for term in rules["prohibited_terms"]:
-            if term.lower() in text.lower():
+        for term_info in region_rules["prohibited_terms"]:
+            term = term_info["term"].lower()
+            if term in text.lower() or any(var.lower() in text.lower() for var in term_info["variations"]):
                 report["compliant"] = False
-                report["violations"].append(f"{region}: Prohibited term '{term}' found")
-                report["channel_risks"]["email"] += 2
-                report["channel_risks"]["social"] += 2
-                report["channel_risks"]["print"] += 1
+                violation = f"{region}: Prohibited term '{term}' found"
+                report["violations"].append({
+                    "region": region,
+                    "type": "prohibited_term",
+                    "term": term,
+                    "severity": term_info["severity"]
+                })
+                
+                # Update channel risks
+                for channel in report["channel_risks"]:
+                    risk_score = compliance_rules.calculate_risk_score([violation], [], region)
+                    report["channel_risks"][channel]["score"] += risk_score
+                    report["channel_risks"][channel]["details"].append(
+                        f"Prohibited term '{term}' increases {channel} risk"
+                    )
         
         # Check required disclaimers
-        for disclaimer in rules["required_disclaimers"]:
-            if disclaimer.lower() not in text.lower():
-                report["warnings"].append(f"{region}: Missing disclaimer about {disclaimer}")
-                report["channel_risks"]["email"] += 1
-                report["channel_risks"]["social"] += 1
-                report["channel_risks"]["print"] += 1
+        for disclaimer in region_rules["required_disclaimers"]:
+            disclaimer_found = any(
+                disc_text.lower() in text.lower() 
+                for disc_text in disclaimer["text"]
+            )
+            if not disclaimer_found:
+                warning = f"{region}: Missing {disclaimer['type']} disclaimer"
+                report["warnings"].append({
+                    "region": region,
+                    "type": "missing_disclaimer",
+                    "disclaimer_type": disclaimer["type"],
+                    "severity": disclaimer["severity"]
+                })
+                
+                # Update channel risks
+                for channel in report["channel_risks"]:
+                    risk_score = compliance_rules.calculate_risk_score([], [warning], region)
+                    report["channel_risks"][channel]["score"] += risk_score
+                    report["channel_risks"][channel]["details"].append(
+                        f"Missing {disclaimer['type']} disclaimer affects {channel} risk"
+                    )
     
     return report
 
@@ -99,31 +94,47 @@ def analyze_ad_copy(image):
     if compliance_report["violations"]:
         report_text += "Violations Found:\n"
         for violation in compliance_report["violations"]:
-            report_text += f"• {violation}\n"
+            report_text += f"• {violation['region']}: {violation['type']} - '{violation['term']}' (Severity: {violation['severity']})\n"
         report_text += "\n"
     
     if compliance_report["warnings"]:
         report_text += "Warnings:\n"
         for warning in compliance_report["warnings"]:
-            report_text += f"• {warning}\n"
+            report_text += f"• {warning['region']}: {warning['disclaimer_type']} (Severity: {warning['severity']})\n"
         report_text += "\n"
     
     report_text += "Channel Risk Assessment:\n"
-    for channel, risk in compliance_report["channel_risks"].items():
-        risk_level = "Low" if risk < 2 else "Medium" if risk < 4 else "High"
-        report_text += f"• {channel.capitalize()}: {risk_level} Risk\n"
+    for channel, risk_info in compliance_report["channel_risks"].items():
+        score = risk_info["score"]
+        risk_level = "Low" if score < 3 else "Medium" if score < 6 else "High"
+        report_text += f"• {channel.capitalize()}: {risk_level} Risk (Score: {score})\n"
+        if risk_info["details"]:
+            for detail in risk_info["details"]:
+                report_text += f"  - {detail}\n"
     
     return report_text
 
 # Create Gradio interface
 iface = gr.Interface(
     fn=analyze_ad_copy,
-    inputs=gr.Image(type="pil"),
+    inputs=[
+        gr.Image(
+            type="pil",
+            label="Upload Marketing Material",
+            height=300,  # Fixed height
+            width=400,   # Fixed width
+            image_mode="RGB",
+            scale=1,     # Prevents auto-scaling
+            source="upload",
+            tool="select"
+        )
+    ],
     outputs=gr.Textbox(label="Compliance Report", lines=10),
     title="Marketing Campaign Compliance Checker",
     description="Upload marketing material to check compliance with US (SEC), UK (FCA), and EU financial regulations.",
     examples=[],
-    theme=gr.themes.Base()
+    theme=gr.themes.Base(),
+    allow_flagging="never"
 )
 
 # Launch the app

compliance_rules/__init__.py

@@ -0,0 +1,56 @@
+"""Loader for compliance rules from all regulatory bodies"""
+
+from .sec_rules import SEC_RULES
+from .fca_rules import FCA_RULES
+from .eu_rules import EU_RULES
+
+class ComplianceRules:
+    def __init__(self):
+        self.rules = {
+            "US_SEC": SEC_RULES,
+            "UK_FCA": FCA_RULES,
+            "EU": EU_RULES
+        }
+    
+    def get_all_rules(self):
+        """Return all compliance rules"""
+        return self.rules
+    
+    def get_rules_by_region(self, region):
+        """Get rules for a specific region"""
+        return self.rules.get(region, {})
+    
+    def get_combined_prohibited_terms(self):
+        """Get all prohibited terms across regions"""
+        prohibited_terms = set()
+        for region_rules in self.rules.values():
+            for term_dict in region_rules["prohibited_terms"]:
+                prohibited_terms.add(term_dict["term"])
+                prohibited_terms.update(term_dict["variations"])
+        return list(prohibited_terms)
+    
+    def get_channel_requirements(self, channel):
+        """Get requirements for a specific channel across all regions"""
+        requirements = {}
+        for region, rules in self.rules.items():
+            if "channel_specific_rules" in rules and channel in rules["channel_specific_rules"]:
+                requirements[region] = rules["channel_specific_rules"][channel]
+        return requirements
+    
+    def calculate_risk_score(self, violations, warnings, region):
+        """Calculate risk score based on violations and warnings"""
+        if region not in self.rules:
+            return 0
+            
+        risk_scoring = self.rules[region]["risk_scoring"]
+        score = 0
+        
+        for violation in violations:
+            if "disclaimer" in violation.lower():
+                score += risk_scoring["missing_disclaimer"]
+            elif "prohibited" in violation.lower():
+                score += risk_scoring["prohibited_term"]
+            else:
+                score += risk_scoring["misleading_statement"]
+                
+        return score

compliance_rules/eu_rules.py

@@ -0,0 +1,83 @@
+"""EU compliance rules for marketing materials"""
+
+EU_RULES = {
+    "required_disclaimers": [
+        {
+            "type": "risk_warning",
+            "text": ["investment involves risk",
+                    "you could lose your invested capital",
+                    "past performance is not a reliable guide to future performance"],
+            "severity": "high",
+            "languages": ["en"]  # expand for other EU languages
+        },
+        {
+            "type": "regulatory_information",
+            "text": ["regulated by", "authorized under EU regulations"],
+            "severity": "high",
+            "languages": ["en"]
+        },
+        {
+            "type": "costs_disclosure",
+            "text": ["fees and charges apply",
+                    "view our fee schedule"],
+            "severity": "medium",
+            "languages": ["en"]
+        }
+    ],
+    "prohibited_terms": [
+        {
+            "term": "assured returns",
+            "variations": ["guaranteed returns", "secure profit", "guaranteed profit"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "no losses",
+            "variations": ["cannot lose", "risk free", "safe investment"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "guaranteed performance",
+            "variations": ["assured performance", "secured returns"],
+            "severity": "high",
+            "context_check": True
+        }
+    ],
+    "channel_specific_rules": {
+        "email": {
+            "required": ["unsubscribe option", "company information"],
+            "character_limit": None,
+            "gdpr_compliance": True
+        },
+        "social": {
+            "required": ["#ad", "#promotion"],
+            "risk_warning_placement": "visible without clicking"
+        },
+        "print": {
+            "required": ["full risk warning", "company details"],
+            "font_size_minimum": "9pt",
+            "prominence": "clearly legible"
+        }
+    },
+    "risk_scoring": {
+        "missing_disclaimer": 4,
+        "prohibited_term": 5,
+        "misleading_statement": 4,
+        "risk_thresholds": {
+            "low": 3,
+            "medium": 6,
+            "high": 9
+        }
+    },
+    "mifid_requirements": {
+        "fair_presentation": {
+            "required": ["balanced view", "prominent risk warnings"],
+            "prohibited": ["emphasize benefits without risks"]
+        },
+        "target_market": {
+            "required": ["clear target market identification"],
+            "prohibited": ["mass marketing of professional products"]
+        }
+    }
+}

compliance_rules/fca_rules.py

@@ -0,0 +1,81 @@
+"""UK FCA compliance rules for marketing materials"""
+
+FCA_RULES = {
+    "required_disclaimers": [
+        {
+            "type": "capital_risk",
+            "text": ["capital at risk",
+                    "you may lose some or all of your investment",
+                    "your capital is at risk"],
+            "severity": "high"
+        },
+        {
+            "type": "regulatory_status",
+            "text": ["regulated by the Financial Conduct Authority",
+                    "authorised and regulated by the FCA",
+                    "FCA regulated"],
+            "severity": "high"
+        },
+        {
+            "type": "past_performance",
+            "text": ["past performance is not a reliable indicator of future results",
+                    "past performance does not guarantee future returns"],
+            "severity": "high"
+        }
+    ],
+    "prohibited_terms": [
+        {
+            "term": "guaranteed profit",
+            "variations": ["secure profit", "assured gains", "guaranteed returns"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "no risk",
+            "variations": ["risk free", "zero risk", "safe investment"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "secured returns",
+            "variations": ["protected returns", "guaranteed income"],
+            "severity": "high",
+            "context_check": True
+        }
+    ],
+    "channel_specific_rules": {
+        "email": {
+            "required": ["opt-out mechanism", "firm details"],
+            "character_limit": None
+        },
+        "social": {
+            "required": ["#ad", "#financial promotion"],
+            "risk_warning_placement": "prominent"
+        },
+        "print": {
+            "required": ["risk warning", "firm details"],
+            "font_size_minimum": "10pt",
+            "risk_warning_prominence": "no less prominent than main message"
+        }
+    },
+    "risk_scoring": {
+        "missing_disclaimer": 4,
+        "prohibited_term": 5,
+        "misleading_statement": 4,
+        "risk_thresholds": {
+            "low": 3,
+            "medium": 6,
+            "high": 9
+        }
+    },
+    "specific_requirements": {
+        "retail_investment": {
+            "required_elements": ["past performance warning", "balanced message"],
+            "restricted_terms": ["tax-free", "guaranteed"]
+        },
+        "pension_products": {
+            "required_elements": ["tax treatment warning", "age restrictions"],
+            "restricted_terms": ["pension liberation", "pension loan"]
+        }
+    }
+}

compliance_rules/sec_rules.py

@@ -0,0 +1,73 @@
+# US SEC compliance rules for marketing materials
+
+SEC_RULES = {
+    "required_disclaimers": [
+        {
+            "type": "past_performance",
+            "text": ["past performance is not indicative of future results",
+                    "past returns do not guarantee future performance"],
+            "severity": "high"
+        },
+        {
+            "type": "investment_risks",
+            "text": ["investment involves risk",
+                    "you may lose your principal",
+                    "investments may lose value"],
+            "severity": "high"
+        },
+        {
+            "type": "regulatory_statement",
+            "text": ["registered with the Securities and Exchange Commission",
+                    "SEC regulated"],
+            "severity": "medium"
+        }
+    ],
+    "prohibited_terms": [
+        {
+            "term": "guaranteed returns",
+            "variations": ["guarantee profits", "assured returns", "guaranteed investment"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "risk-free",
+            "variations": ["no risk", "zero risk", "riskless"],
+            "severity": "high",
+            "context_check": True
+        },
+        {
+            "term": "sure thing",
+            "variations": ["cant lose", "never lose", "always profits"],
+            "severity": "high",
+            "context_check": False
+        }
+    ],
+    "channel_specific_rules": {
+        "email": {
+            "required": ["unsubscribe option", "physical address"],
+            "character_limit": None
+        },
+        "social": {
+            "required": ["#ad", "disclosure"],
+            "character_limit": {
+                "twitter": 280,
+                "instagram": 2200,
+                "linkedin": 3000
+            }
+        },
+        "print": {
+            "required": ["full disclaimer", "company details"],
+            "font_size_minimum": "8pt"
+        }
+    },
+    "risk_scoring": {
+        "missing_disclaimer": 3,
+        "prohibited_term": 5,
+        "misleading_statement": 4,
+        "risk_thresholds": {
+            "low": 2,
+            "medium": 5,
+            "high": 8
+        }
+    }
+}