chat/reasoning_engine.py

# chat/reasoning_engine.py
"""
Advanced reasoning engine for deep methodological and clinical analysis
Provides evidence-based reasoning and quality assessment
"""

from typing import List, Dict, Any, Tuple
from llm.llm_provider import GrokLLM
from llm.prompt_templates import MedicalResearchPrompts, ResponseFormatter
import re


class ReasoningEngine:
    """
    Advanced reasoning engine for deep research analysis
    Focuses on methodological rigor, evidence quality, and clinical relevance
    """

    def __init__(self, llm=None):
        self.llm = llm or GrokLLM(model="model")  # Use shared LLM
        self.prompts = MedicalResearchPrompts()
        self.formatter = ResponseFormatter()

    def analyze_methodology(self, papers: List[Dict], query: str, domain: str) -> Dict[str, Any]:
        """
        Deep methodological analysis of research papers
        """
        print(f"🔬 Conducting methodological analysis on {len(papers)} papers")

        if len(papers) > 10:
            print(f"🔄 Focusing on top 10 most relevant papers for deep analysis")
            papers = papers[:10]

        try:
            # Generate methodological reasoning
            methodology_prompt = self.prompts.methodology_reasoning(papers, query, domain)

            response = self.llm.generate(
                methodology_prompt,
                system_message=self.prompts.SYSTEM_MESSAGES["methodology_expert"],
                temperature=0.1,
                max_tokens=2000
            )

            # Extract structured methodological insights
            method_insights = self._extract_methodological_insights(response)
            quality_assessment = self._assess_evidence_quality(papers)
            best_practices = self._identify_best_practices(papers)

            return {
                "methodological_analysis": response,
                "structured_insights": method_insights,
                "evidence_quality": quality_assessment,
                "best_practices": best_practices,
                "query": query,
                "domain": domain,
                "papers_analyzed": len(papers)
            }

        except Exception as e:
            print(f"❌ Methodological analysis error: {e}")
            return self._create_fallback_methodology_analysis(papers, query, domain)

    def analyze_clinical_implications(self, papers: List[Dict], domain: str) -> Dict[str, Any]:
        """
        Analyze clinical implications and translational potential
        """
        print(f"🏥 Analyzing clinical implications of {len(papers)} papers")

        try:
            # Generate clinical implications analysis
            clinical_prompt = self.prompts.clinical_implications(papers, domain)

            response = self.llm.generate(
                clinical_prompt,
                system_message=self.prompts.SYSTEM_MESSAGES["domain_specialist"].format(domain=domain),
                temperature=0.1,
                max_tokens=2000
            )

            # Extract structured clinical insights
            clinical_insights = self._extract_clinical_insights(response)
            translation_readiness = self._assess_translation_readiness(papers)
            implementation_challenges = self._identify_implementation_challenges(papers)

            return {
                "clinical_analysis": response,
                "clinical_insights": clinical_insights,
                "translation_readiness": translation_readiness,
                "implementation_challenges": implementation_challenges,
                "domain": domain,
                "papers_analyzed": len(papers)
            }

        except Exception as e:
            print(f"❌ Clinical implications analysis error: {e}")
            return self._create_fallback_clinical_analysis(papers, domain)

    def _extract_methodological_insights(self, analysis: str) -> List[Dict[str, Any]]:
        """Extract structured methodological insights"""
        insights = []

        # Patterns for different methodological aspects
        patterns = {
            "experimental_design": r"(?:experimental design|study design)[^.!?]*[.!?]",
            "statistical_methods": r"(?:statistical method|analysis approach|statistical test)[^.!?]*[.!?]",
            "validation_strategy": r"(?:validation|cross.validation|external validation)[^.!?]*[.!?]",
            "reproducibility": r"(?:reproducibility|replication|reproducible)[^.!?]*[.!?]",
            "limitations": r"(?:limitation|constraint|challenge)[^.!?]*[.!?]"
        }

        for aspect, pattern in patterns.items():
            matches = re.findall(pattern, analysis, re.IGNORECASE)
            for match in matches[:3]:  # Limit to top 3 per aspect
                insights.append({
                    "aspect": aspect.replace('_', ' ').title(),
                    "insight": match.strip(),
                    "confidence": self._assess_insight_confidence(match)
                })

        return insights

    def _assess_evidence_quality(self, papers: List[Dict]) -> Dict[str, Any]:
        """Assess overall evidence quality across papers"""
        quality_metrics = {
            "sample_sizes": [],
            "validation_mentioned": 0,
            "limitations_discussed": 0,
            "comparison_baselines": 0,
            "statistical_significance": 0
        }

        for paper in papers:
            abstract = paper.get('abstract', '').lower()

            # Sample size extraction (simplified)
            sample_matches = re.findall(r'(\d+)\s*(?:patient|subject|sample|case)', abstract)
            if sample_matches:
                quality_metrics["sample_sizes"].extend([int(match) for match in sample_matches])

            # Methodological quality indicators
            if any(term in abstract for term in ['validation', 'cross-validation', 'external validation']):
                quality_metrics["validation_mentioned"] += 1

            if any(term in abstract for term in ['limitation', 'constraint', 'challenge']):
                quality_metrics["limitations_discussed"] += 1

            if any(term in abstract for term in ['compared to', 'baseline', 'versus']):
                quality_metrics["comparison_baselines"] += 1

            if any(term in abstract for term in ['p value', 'p<', 'statistical significance', 'confidence interval']):
                quality_metrics["statistical_significance"] += 1

        # Calculate overall quality score
        total_papers = len(papers)
        quality_score = (
                                (quality_metrics["validation_mentioned"] / total_papers * 0.3) +
                                (quality_metrics["limitations_discussed"] / total_papers * 0.2) +
                                (quality_metrics["comparison_baselines"] / total_papers * 0.25) +
                                (quality_metrics["statistical_significance"] / total_papers * 0.25)
                        ) * 100

        # Sample size analysis
        sample_sizes = quality_metrics["sample_sizes"]
        avg_sample_size = sum(sample_sizes) / len(sample_sizes) if sample_sizes else 0

        return {
            "overall_quality_score": round(quality_score, 1),
            "quality_level": "high" if quality_score > 70 else "medium" if quality_score > 40 else "low",
            "sample_size_analysis": {
                "average_sample_size": round(avg_sample_size),
                "papers_with_sample_info": len(sample_sizes),
                "sample_size_range": f"{min(sample_sizes)}-{max(sample_sizes)}" if sample_sizes else "N/A"
            },
            "methodological_indicators": quality_metrics
        }

    def _identify_best_practices(self, papers: List[Dict]) -> List[Dict[str, Any]]:
        """Identify emerging best practices from the literature"""
        best_practices = []

        practice_categories = {
            "validation": ["cross-validation", "external validation", "independent test set"],
            "transparency": ["open source", "code available", "model sharing"],
            "reproducibility": ["reproducible", "replication", "detailed methods"],
            "ethical_considerations": ["ethical", "bias assessment", "fairness"],
            "clinical_relevance": ["clinical utility", "patient outcomes", "clinical impact"]
        }

        for paper in papers:
            abstract = paper.get('abstract', '').lower()
            title = paper.get('title', '').lower()

            for category, keywords in practice_categories.items():
                if any(keyword in abstract or keyword in title for keyword in keywords):
                    # Check if this practice is already recorded
                    existing_practice = next((p for p in best_practices if p['practice'] == category), None)

                    if existing_practice:
                        existing_practice['supporting_papers'] += 1
                    else:
                        best_practices.append({
                            "practice": category.replace('_', ' ').title(),
                            "description": self._get_practice_description(category),
                            "supporting_papers": 1,
                            "examples": [paper.get('title', 'Unknown')]
                        })

        # Sort by number of supporting papers
        best_practices.sort(key=lambda x: x['supporting_papers'], reverse=True)

        return best_practices[:10]  # Return top 10 best practices

    def _get_practice_description(self, practice_category: str) -> str:
        """Get description for best practice categories"""
        descriptions = {
            "validation": "Rigorous validation methods including cross-validation and external testing",
            "transparency": "Open sharing of code, data, and methodologies",
            "reproducibility": "Detailed methods enabling study replication",
            "ethical_considerations": "Consideration of ethical implications and bias assessment",
            "clinical_relevance": "Focus on clinically meaningful outcomes and applications"
        }
        return descriptions.get(practice_category, "Emerging best practice in the field")

    def _extract_clinical_insights(self, analysis: str) -> List[Dict[str, Any]]:
        """Extract structured clinical insights"""
        insights = []

        # Patterns for clinical insights
        patterns = {
            "immediate_applications": r"(?:immediate application|ready for use|clinical implementation)[^.!?]*[.!?]",
            "future_directions": r"(?:future direction|further research|next steps)[^.!?]*[.!?]",
            "patient_impact": r"(?:patient impact|clinical benefit|patient outcomes)[^.!?]*[.!?]",
            "implementation_challenges": r"(?:implementation challenge|adoption barrier|clinical integration)[^.!?]*[.!?]"
        }

        for aspect, pattern in patterns.items():
            matches = re.findall(pattern, analysis, re.IGNORECASE)
            for match in matches[:2]:  # Limit to top 2 per aspect
                insights.append({
                    "aspect": aspect.replace('_', ' ').title(),
                    "insight": match.strip(),
                    "readiness": self._assess_clinical_readiness(match)
                })

        return insights

    def _assess_translation_readiness(self, papers: List[Dict]) -> Dict[str, Any]:
        """Assess translational readiness of research findings"""
        readiness_indicators = {
            "clinical_trial_mentions": 0,
            "patient_outcomes": 0,
            "real_world_data": 0,
            "implementation_studies": 0,
            "guideline_references": 0
        }

        for paper in papers:
            abstract = paper.get('abstract', '').lower()
            title = paper.get('title', '').lower()

            if any(term in abstract or term in title for term in ['clinical trial', 'randomized', 'controlled study']):
                readiness_indicators["clinical_trial_mentions"] += 1

            if any(term in abstract for term in ['patient outcome', 'survival', 'mortality', 'quality of life']):
                readiness_indicators["patient_outcomes"] += 1

            if any(term in abstract for term in ['real world', 'clinical practice', 'routine care']):
                readiness_indicators["real_world_data"] += 1

            if any(term in abstract for term in ['implementation', 'adoption', 'integration']):
                readiness_indicators["implementation_studies"] += 1

            if any(term in abstract for term in ['guideline', 'recommendation', 'standard of care']):
                readiness_indicators["guideline_references"] += 1

        total_papers = len(papers)
        readiness_score = sum(readiness_indicators.values()) / (total_papers * 5) * 100

        return {
            "readiness_score": round(readiness_score, 1),
            "readiness_level": "high" if readiness_score > 60 else "medium" if readiness_score > 30 else "low",
            "indicators": readiness_indicators
        }

    def _identify_implementation_challenges(self, papers: List[Dict]) -> List[str]:
        """Identify common implementation challenges"""
        challenges = []
        challenge_keywords = [
            'cost', 'expensive', 'infrastructure', 'expertise', 'training',
            'regulatory', 'fda', 'approval', 'validation', 'integration',
            'workflow', 'adoption', 'resistance', 'change management'
        ]

        for paper in papers:
            abstract = paper.get('abstract', '').lower()

            for keyword in challenge_keywords:
                if keyword in abstract and keyword not in challenges:
                    challenges.append(keyword)

        return challenges[:10]  # Return top 10 challenges

    def _assess_insight_confidence(self, insight: str) -> str:
        """Assess confidence level of an insight"""
        insight_lower = insight.lower()

        if any(term in insight_lower for term in ['clearly', 'definitely', 'strongly', 'convincingly']):
            return "high"
        elif any(term in insight_lower for term in ['suggest', 'indicate', 'likely', 'probably']):
            return "medium"
        else:
            return "low"

    def _assess_clinical_readiness(self, insight: str) -> str:
        """Assess clinical readiness level"""
        insight_lower = insight.lower()

        if any(term in insight_lower for term in ['immediate', 'ready', 'current', 'now']):
            return "immediate"
        elif any(term in insight_lower for term in ['near future', 'soon', 'emerging']):
            return "near_term"
        else:
            return "long_term"

    def _create_fallback_methodology_analysis(self, papers: List[Dict], query: str, domain: str) -> Dict[str, Any]:
        """Create basic methodology analysis when LLM fails"""
        quality_assessment = self._assess_evidence_quality(papers)
        best_practices = self._identify_best_practices(papers)

        basic_analysis = f"""
        Basic Methodological Analysis for: {query}
        Domain: {domain}
        Papers Analyzed: {len(papers)}

        Evidence Quality: {quality_assessment['quality_level']} ({quality_assessment['overall_quality_score']}/100)
        Best Practices Identified: {len(best_practices)}

        Note: Detailed methodological reasoning unavailable.
        """

        return {
            "methodological_analysis": basic_analysis,
            "structured_insights": [],
            "evidence_quality": quality_assessment,
            "best_practices": best_practices,
            "query": query,
            "domain": domain,
            "papers_analyzed": len(papers),
            "fallback_used": True
        }

    def _create_fallback_clinical_analysis(self, papers: List[Dict], domain: str) -> Dict[str, Any]:
        """Create basic clinical analysis when LLM fails"""
        translation_readiness = self._assess_translation_readiness(papers)
        implementation_challenges = self._identify_implementation_challenges(papers)

        basic_analysis = f"""
        Basic Clinical Implications Analysis for: {domain}
        Papers Analyzed: {len(papers)}

        Translation Readiness: {translation_readiness['readiness_level']} ({translation_readiness['readiness_score']}/100)
        Implementation Challenges: {len(implementation_challenges)}

        Note: Detailed clinical analysis unavailable.
        """

        return {
            "clinical_analysis": basic_analysis,
            "clinical_insights": [],
            "translation_readiness": translation_readiness,
            "implementation_challenges": implementation_challenges,
            "domain": domain,
            "papers_analyzed": len(papers),
            "fallback_used": True
        }


# Quick test
def test_reasoning_engine():
    """Test the reasoning engine"""
    print("🧪 Testing Reasoning Engine")
    print("=" * 50)

    test_papers = [
        {
            'title': 'Randomized Trial of AI Diagnostic Tool',
            'authors': ['Smith J', 'Johnson A'],
            'abstract': 'Randomized controlled trial of 1000 patients comparing AI diagnostic tool with radiologist interpretation. The AI system showed non-inferiority with 94% accuracy vs 92% for radiologists (p<0.05). Limitations include single-center design.',
            'source': 'NEJM',
            'domain': 'medical_imaging',
            'publication_date': '2024-01-15'
        },
        {
            'title': 'Deep Learning for Early Cancer Detection',
            'authors': ['Lee K', 'Chen R'],
            'abstract': 'Prospective study applying deep learning to screening mammography in 50,000 patients. The model achieved AUC of 0.95 for early cancer detection. External validation performed on independent dataset.',
            'source': 'JAMA',
            'domain': 'medical_imaging',
            'publication_date': '2024-02-20'
        }
    ]

    engine = ReasoningEngine()

    try:
        # Test methodological analysis
        method_analysis = engine.analyze_methodology(
            test_papers,
            "AI diagnostic accuracy in medical imaging",
            "medical_imaging"
        )

        print(f"✅ Methodological analysis completed")
        print(f"📊 Evidence quality: {method_analysis['evidence_quality']['quality_level']}")
        print(f"💡 Best practices identified: {len(method_analysis['best_practices'])}")

        # Test clinical implications
        clinical_analysis = engine.analyze_clinical_implications(test_papers, "medical_imaging")

        print(f"✅ Clinical analysis completed")
        print(f"🏥 Translation readiness: {clinical_analysis['translation_readiness']['readiness_level']}")
        print(f"🚧 Implementation challenges: {len(clinical_analysis['implementation_challenges'])}")

    except Exception as e:
        print(f"❌ Reasoning engine test failed: {e}")


if __name__ == "__main__":
    test_reasoning_engine()