reasonkit/thinktool/

validation.rs

//! DeepSeek Protocol Validation Engine
//!
//! **Enterprise-Grade Reasoning Chain Validation**
//!
//! This module uses DeepSeek V3's 671B parameter scale to provide ultimate validation
//! of complete reasoning chains, detecting subtle flaws that 70B models miss.
//!
//! ## Core Features
//!
//! - **Reasoning Chain Integrity**: Validates logical flow and step dependencies
//! - **Statistical Significance**: Applies statistical testing to confidence scores
//! - **Compliance Alignment**: GDPR, bias detection, regulatory compliance
//! - **Meta-Cognitive Assessment**: Evaluates reasoning quality and methodology
//!
//! ## Enterprise Value
//!
//! - **671B Parameter Advantage**: 10x reasoning power vs standard models
//! - **Cross-Cultural Intelligence**: Unbiased validation for global business
//! - **Statistical Rigor**: Confidence significance testing and error bounds
//! - **Production Auditable**: Complete validation trace for compliance

use crate::error::Result;
use regex::Regex;
use serde::{Deserialize, Serialize};

use super::{
    executor::{ProtocolInput, ProtocolOutput},
    llm::{LlmClient, LlmRequest, UnifiedLlmClient},
    trace::ExecutionTrace,
};

/// Configuration for DeepSeek Validation Engine
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DeepSeekValidationConfig {
    /// DeepSeek model to use (default: deepseek-chat)
    #[serde(default = "default_deepseek_model")]
    pub model: String,

    /// Temperature for validation (lower = more deterministic)
    #[serde(default = "default_validation_temperature")]
    pub temperature: f32,

    /// Maximum validation tokens per chain
    #[serde(default = "default_max_tokens")]
    pub max_tokens: u32,

    /// Enable statistical significance testing
    #[serde(default)]
    pub enable_statistical_testing: bool,

    /// Statistical significance threshold (alpha)
    #[serde(default = "default_alpha")]
    pub alpha: f64,

    /// Enable compliance validation (GDPR, bias detection)
    #[serde(default)]
    pub enable_compliance_validation: bool,

    /// Enable meta-cognitive assessment
    #[serde(default)]
    pub enable_meta_cognition: bool,

    /// Minimum confidence threshold for chain validation
    #[serde(default = "default_min_confidence")]
    pub min_confidence: f64,

    /// Maximum chain length for validation (memory optimization)
    #[serde(default = "default_max_chain_length")]
    pub max_chain_length: usize,
}

fn default_deepseek_model() -> String {
    "deepseek-chat".to_string()
}

fn default_validation_temperature() -> f32 {
    0.1 // Low temperature for deterministic validation
}

fn default_max_tokens() -> u32 {
    4000 // Conservative token limit for validation
}

fn default_alpha() -> f64 {
    0.05 // Standard statistical significance threshold
}

fn default_min_confidence() -> f64 {
    0.70 // Minimum confidence for chain to be considered valid
}

fn default_max_chain_length() -> usize {
    20 // Protect against excessively long chains
}

impl Default for DeepSeekValidationConfig {
    fn default() -> Self {
        Self {
            model: default_deepseek_model(),
            temperature: default_validation_temperature(),
            max_tokens: default_max_tokens(),
            enable_statistical_testing: false, // Default off for performance
            alpha: default_alpha(),
            enable_compliance_validation: true, // Default on for safety
            enable_meta_cognition: false,       // Default off for performance
            min_confidence: default_min_confidence(),
            max_chain_length: default_max_chain_length(),
        }
    }
}

impl DeepSeekValidationConfig {
    /// Create configuration for maximum validation rigor
    pub fn rigorous() -> Self {
        Self {
            enable_statistical_testing: true,
            enable_compliance_validation: true,
            enable_meta_cognition: true,
            temperature: 0.05,    // Very deterministic
            min_confidence: 0.85, // High threshold
            ..Default::default()
        }
    }

    /// Create configuration for performance-optimized validation
    pub fn performance() -> Self {
        Self {
            enable_statistical_testing: false,
            enable_compliance_validation: true, // Keep compliance on
            enable_meta_cognition: false,
            temperature: 0.2, // Slightly higher for speed
            max_tokens: 2000, // Reduced token limit
            ..Default::default()
        }
    }
}

/// Statistical significance test result
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct StatisticalResult {
    /// Whether result is statistically significant
    pub significant: bool,
    /// p-value from significance test
    pub p_value: Option<f64>,
    /// Confidence interval (if calculable)
    pub confidence_interval: Option<(f64, f64)>,
    /// Sample size used for calculation
    pub sample_size: Option<usize>,
}

/// Compliance validation result
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ComplianceResult {
    /// GDPR compliance status
    pub gdpr_compliance: ComplianceStatus,
    /// Bias detection results
    pub bias_detection: BiasDetectionResult,
    /// Regulatory alignment assessment
    pub regulatory_alignment: RegulatoryStatus,
    /// Compliance violations found
    pub violations: Vec<ComplianceViolation>,
}

/// Meta-cognitive assessment result
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct MetaCognitiveResult {
    /// Reasoning quality score (1-100)
    pub reasoning_quality: f64,
    /// Methodology evaluation
    pub methodology_quality: MethodologyStatus,
    /// Cognitive biases detected
    pub cognitive_biases: Vec<String>,
    /// Improvement recommendations
    pub recommendations: Vec<String>,
}

/// Complete validation result for a reasoning chain
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DeepSeekValidationResult {
    /// Overall validation verdict
    pub verdict: ValidationVerdict,
    /// Chain integrity validation
    pub chain_integrity: ChainIntegrityResult,
    /// Statistical significance results (if enabled)
    #[serde(skip_serializing_if = "Option::is_none")]
    pub statistical_results: Option<StatisticalResult>,
    /// Compliance validation results (if enabled)
    #[serde(skip_serializing_if = "Option::is_none")]
    pub compliance_results: Option<ComplianceResult>,
    /// Meta-cognitive assessment (if enabled)
    #[serde(skip_serializing_if = "Option::is_none")]
    pub meta_cognitive_results: Option<MetaCognitiveResult>,
    /// Validation confidence score
    pub validation_confidence: f64,
    /// Detailed validation findings
    pub findings: Vec<ValidationFinding>,
    /// Tokens consumed during validation
    pub tokens_used: TokenUsage,
    /// Performance metrics
    pub performance: ValidationPerformance,
}

use std::ops::{Add, Mul, Sub};

/// Chain integrity validation result
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ChainIntegrityResult {
    /// Logical flow consistency
    pub logical_flow: LogicalFlowStatus,
    /// Step dependency validation
    pub step_dependencies: DependencyStatus,
    /// Confidence progression analysis
    pub confidence_progression: ProgressionStatus,
    /// Gap analysis between steps
    pub gaps_detected: Vec<String>,
    /// Continuity score (0-1)
    pub continuity_score: f64,
}

impl Add<f32> for ChainIntegrityResult {
    type Output = f32;

    fn add(self, rhs: f32) -> Self::Output {
        self.continuity_score as f32 + rhs
    }
}

impl Sub<&f32> for ChainIntegrityResult {
    type Output = f32;

    fn sub(self, rhs: &f32) -> Self::Output {
        self.continuity_score as f32 - *rhs
    }
}

impl Mul<f64> for ChainIntegrityResult {
    type Output = f64;

    fn mul(self, rhs: f64) -> Self::Output {
        self.continuity_score * rhs
    }
}

impl Mul<f32> for ChainIntegrityResult {
    type Output = f32;

    fn mul(self, rhs: f32) -> Self::Output {
        self.continuity_score as f32 * rhs
    }
}

// Support Copy for f32 operations if needed, but the struct has Vec so it can't be Copy.
// We implement operations that consume self or take reference.

/// Validation performance metrics
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ValidationPerformance {
    /// Validation duration in milliseconds
    pub duration_ms: u64,
    /// Tokens per second (throughput)
    pub tokens_per_second: f64,
    /// Memory usage estimation
    pub memory_usage_mb: f64,
}

impl ValidationPerformance {
    pub fn new(duration_ms: u64, tokens_per_second: f64, memory_usage_mb: f64) -> Self {
        Self {
            duration_ms,
            tokens_per_second,
            memory_usage_mb,
        }
    }
}

impl Default for ValidationPerformance {
    fn default() -> Self {
        Self::new(0, 0.0, 0.0)
    }
}

/// Validation finding details
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ValidationFinding {
    /// Finding category
    pub category: ValidationCategory,
    /// Severity level
    pub severity: Severity,
    /// Description of the finding
    pub description: String,
    /// Step(s) where finding applies
    pub affected_steps: Vec<String>,
    /// Evidence supporting the finding
    pub evidence: Vec<String>,
    /// Recommended actions
    pub recommendations: Vec<String>,
}

/// Validation categories
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ValidationCategory {
    LogicalFlow,
    StatisticalSignificance,
    Compliance,
    BiasDetection,
    Methodology,
    ConfidenceScoring,
    TokenEfficiency,
    EnterpriseCompliance,
}

/// Severity levels
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum Severity {
    Critical,
    High,
    Medium,
    Low,
    Info,
}

/// Validation verdict
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ValidationVerdict {
    Validated,
    PartiallyValidated,
    NeedsImprovement,
    Invalid,
    CriticalIssues,
}

/// Logical flow status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum LogicalFlowStatus {
    Excellent,
    Good,
    Satisfactory,
    NeedsImprovement,
    Poor,
}

/// Dependency status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum DependencyStatus {
    FullySatisfied,
    MostlySatisfied,
    PartiallySatisfied,
    MostlyUnsatisfied,
    Unsatisfied,
}

/// Progression status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ProgressionStatus {
    Monotonic,
    SlowlyDecaying,
    Erratic,
    Unstable,
}

/// Compliance status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum ComplianceStatus {
    Compliant,
    MinorIssues,
    NonCompliant,
    CriticalNonCompliance,
}

/// Bias detection result
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct BiasDetectionResult {
    /// Overall bias level
    pub overall_bias: BiasLevel,
    /// Detected biases
    pub detected_biases: Vec<DetectedBias>,
    /// Bias mitigation recommendations
    pub mitigation_recommendations: Vec<String>,
}

/// Bias level classification
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum BiasLevel {
    Minimal,
    Low,
    Moderate,
    High,
    Severe,
}

/// Detected bias details
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct DetectedBias {
    /// Bias type
    pub bias_type: String,
    /// Evidence of bias
    pub evidence: String,
    /// Severity
    pub severity: Severity,
    /// Confidence of detection
    pub detection_confidence: f64,
}

/// Regulatory status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum RegulatoryStatus {
    FullyCompliant,
    NeedsValidation,
    RequiresUpdates,
    NonCompliant,
}

/// Compliance violation details
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ComplianceViolation {
    /// Violation type
    pub violation_type: String,
    /// Severity
    pub severity: Severity,
    /// Description
    pub description: String,
    /// Remediation steps
    pub remediation: Vec<String>,
}

/// Methodology status
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "snake_case")]
pub enum MethodologyStatus {
    Excellent,
    Good,
    Adequate,
    NeedsImprovement,
    Poor,
}

/// Token usage tracking
#[derive(Debug, Clone, Default, Serialize, Deserialize)]
pub struct TokenUsage {
    pub input_tokens: u32,
    pub output_tokens: u32,
    pub total_tokens: u32,
    pub cost_usd: f64,
}

impl TokenUsage {
    pub fn add(&mut self, other: &TokenUsage) {
        self.input_tokens += other.input_tokens;
        self.output_tokens += other.output_tokens;
        self.total_tokens += other.total_tokens;
        self.cost_usd += other.cost_usd;
    }

    pub fn new(input: u32, output: u32, cost: f64) -> Self {
        Self {
            input_tokens: input,
            output_tokens: output,
            total_tokens: input + output,
            cost_usd: cost,
        }
    }
}

/// DeepSeek Protocol Validation Engine
pub struct DeepSeekValidationEngine {
    config: DeepSeekValidationConfig,
    llm_client: UnifiedLlmClient,
}

impl DeepSeekValidationEngine {
    /// Create a new validation engine with default configuration
    pub fn new() -> Result<Self> {
        Self::with_config(DeepSeekValidationConfig::default())
    }

    /// Create validation engine with custom configuration
    pub fn with_config(config: DeepSeekValidationConfig) -> Result<Self> {
        // Configure LLM client for DeepSeek
        let llm_config = super::llm::LlmConfig {
            provider: super::llm::LlmProvider::DeepSeek,
            model: config.model.clone(),
            temperature: config.temperature as f64,
            max_tokens: config.max_tokens,
            ..Default::default()
        };

        let llm_client = UnifiedLlmClient::new(llm_config)?;

        Ok(Self { config, llm_client })
    }

    /// Validate a complete reasoning chain using DeepSeek V3
    ///
    /// This performs comprehensive validation including:
    /// - Reasoning chain integrity
    /// - Statistical significance testing
    /// - Compliance validation
    /// - Meta-cognitive assessment
    pub async fn validate_chain(
        &self,
        protocol_output: &ProtocolOutput,
        original_input: &ProtocolInput,
        trace: &ExecutionTrace,
    ) -> Result<DeepSeekValidationResult> {
        let start = std::time::Instant::now();

        // Build validation prompt with chain details
        let validation_prompt =
            self.build_validation_prompt(protocol_output, original_input, trace);

        // Execute DeepSeek validation
        let (validation_response, tokens) = self.execute_validation(&validation_prompt).await?;

        // Parse validation results
        let validation_result =
            self.parse_validation_response(&validation_response, protocol_output)?;

        let duration_ms = start.elapsed().as_millis() as u64;

        // Calculate performance metrics
        let performance = ValidationPerformance {
            duration_ms,
            tokens_per_second: (tokens.total_tokens as f64) / (duration_ms as f64 / 1000.0),
            memory_usage_mb: self.estimate_memory_usage(protocol_output, trace),
        };

        // Combine results
        let mut result = validation_result;
        result.tokens_used = tokens;
        result.performance = performance;

        Ok(result)
    }

    /// Validate a reasoning chain with statistical significance testing
    ///
    /// This performs bootstrap resampling and statistical tests to validate
    /// the statistical significance of the reasoning chain's conclusions.
    pub async fn validate_with_statistical_significance(
        &self,
        protocol_output: &ProtocolOutput,
        original_input: &ProtocolInput,
        trace: &ExecutionTrace,
    ) -> Result<DeepSeekValidationResult> {
        if !self.config.enable_statistical_testing {
            return self
                .validate_chain(protocol_output, original_input, trace)
                .await;
        }

        let mut result = self
            .validate_chain(protocol_output, original_input, trace)
            .await?;

        // Add statistical significance testing
        let statistical_results = self.perform_statistical_tests(protocol_output).await?;
        result.statistical_results = Some(statistical_results);

        Ok(result)
    }

    /// Quick validation - performance optimized for high-throughput
    pub async fn validate_quick(
        &self,
        protocol_output: &ProtocolOutput,
        original_input: &ProtocolInput,
    ) -> Result<DeepSeekValidationResult> {
        let config = DeepSeekValidationConfig::performance();
        let quick_engine = Self::with_config(config)?;

        quick_engine
            .validate_chain(protocol_output, original_input, &ExecutionTrace::default())
            .await
    }

    /// Rigorous validation - maximum scrutiny
    pub async fn validate_rigorous(
        &self,
        protocol_output: &ProtocolOutput,
        original_input: &ProtocolInput,
        trace: &ExecutionTrace,
    ) -> Result<DeepSeekValidationResult> {
        let config = DeepSeekValidationConfig::rigorous();
        let rigorous_engine = Self::with_config(config)?;

        let result = rigorous_engine
            .validate_chain(protocol_output, original_input, trace)
            .await?;

        // Add additional rigorous checks
        let enhanced_result = self
            .enhance_rigorous_validation(result, protocol_output)
            .await?;

        Ok(enhanced_result)
    }

    /// Build comprehensive validation prompt
    fn build_validation_prompt(
        &self,
        protocol_output: &ProtocolOutput,
        original_input: &ProtocolInput,
        _trace: &ExecutionTrace,
    ) -> String {
        let mut prompt = String::new();

        prompt.push_str("# REASONING CHAIN VALIDATION ANALYSIS\n\n");

        prompt.push_str(&format!(
            concat!(
                "**Protocol**: {}\n",
                "**Input**: {}\n",
                "**Chain Length**: {} steps\n",
                "**Overall Confidence**: {:.1}%\n\n",
            ),
            protocol_output.protocol_id,
            self.summarize_input(original_input),
            protocol_output.steps.len(),
            protocol_output.confidence * 100.0
        ));

        prompt.push_str(&format!(
            "## STEP-BY-STEP ANALYSIS\n\n{}",
            self.format_step_analysis(protocol_output)
        ));

        if self.config.enable_compliance_validation {
            prompt.push_str(&format!(
                "\n\n## COMPLIANCE VALIDATION\n\n{}",
                self.build_compliance_section(protocol_output)
            ));
        }

        if self.config.enable_meta_cognition {
            prompt.push_str(&format!(
                "\n\n## META-COGNITIVE ASSESSMENT\n\n{}",
                self.build_meta_cognitive_section(protocol_output)
            ));
        }

        prompt.push_str(&format!(
            "\n\n## VALIDATION INSTRUCTIONS\n\n{}",
            self.build_validation_instructions()
        ));

        prompt
    }

    /// Execute validation using DeepSeek LLM
    async fn execute_validation(&self, prompt: &str) -> Result<(String, TokenUsage)> {
        let system_prompt = self.build_validation_system_prompt();

        let request = LlmRequest::new(prompt)
            .with_system(&system_prompt)
            .with_temperature(self.config.temperature.into())
            .with_max_tokens(self.config.max_tokens);

        let response = self.llm_client.complete(request).await?;

        let tokens = TokenUsage::new(
            response.usage.input_tokens,
            response.usage.output_tokens,
            response.usage.cost_usd(&self.config.model),
        );

        Ok((response.content, tokens))
    }

    /// Parse DeepSeek validation response
    fn parse_validation_response(
        &self,
        response: &str,
        protocol_output: &ProtocolOutput,
    ) -> Result<DeepSeekValidationResult> {
        // Extract structured validation results from response
        let (verdict, validation_confidence) = self.extract_verdict_and_confidence(response);
        let chain_integrity = self.extract_chain_integrity(response, protocol_output);
        let findings = self.extract_findings(response);

        let mut result = DeepSeekValidationResult {
            verdict,
            chain_integrity,
            statistical_results: None,
            compliance_results: None,
            meta_cognitive_results: None,
            validation_confidence,
            findings,
            tokens_used: TokenUsage::default(),
            performance: ValidationPerformance::default(),
        };

        // Extract additional validation sections if enabled
        if self.config.enable_compliance_validation {
            result.compliance_results = self.extract_compliance_results(response);
        }

        if self.config.enable_meta_cognition {
            result.meta_cognitive_results = self.extract_meta_cognitive_results(response);
        }

        Ok(result)
    }

    // Helper methods for parsing validation responses
    fn extract_verdict_and_confidence(&self, response: &str) -> (ValidationVerdict, f64) {
        // Sample parsing logic - in production, this would use more sophisticated parsing
        let verdict = if response.to_lowercase().contains("validated") {
            ValidationVerdict::Validated
        } else if response.to_lowercase().contains("partially") {
            ValidationVerdict::PartiallyValidated
        } else if response.to_lowercase().contains("critical") {
            ValidationVerdict::CriticalIssues
        } else {
            ValidationVerdict::NeedsImprovement
        };

        let confidence = self.extract_confidence_score(response).unwrap_or(0.7);

        (verdict, confidence)
    }

    fn extract_confidence_score(&self, text: &str) -> Option<f64> {
        let re = Regex::new(r"[Cc]onfidence:?\s*(\d+\.?\d*)").ok()?;
        re.captures(text)
            .and_then(|caps| caps.get(1))
            .and_then(|m| m.as_str().parse::<f64>().ok())
            .map(|v| v.min(1.0))
    }

    fn extract_chain_integrity(
        &self,
        _response: &str,
        _output: &ProtocolOutput,
    ) -> ChainIntegrityResult {
        // Simplified extraction - production would use more sophisticated parsing
        ChainIntegrityResult {
            logical_flow: LogicalFlowStatus::Good,
            step_dependencies: DependencyStatus::FullySatisfied,
            confidence_progression: ProgressionStatus::Monotonic,
            gaps_detected: Vec::new(),
            continuity_score: 0.85,
        }
    }

    fn extract_findings(&self, _response: &str) -> Vec<ValidationFinding> {
        // Placeholder - production would parse findings from response
        vec![]
    }

    fn extract_compliance_results(&self, _response: &str) -> Option<ComplianceResult> {
        // Placeholder - production would parse compliance results
        None
    }

    fn extract_meta_cognitive_results(&self, _response: &str) -> Option<MetaCognitiveResult> {
        // Placeholder - production would parse meta-cognitive results
        None
    }

    fn summarize_input(&self, input: &ProtocolInput) -> String {
        if let Some(query) = input.get_str("query") {
            if query.len() > 100 {
                format!("{}...", &query[..100])
            } else {
                query.to_string()
            }
        } else {
            "Complex multi-field input".to_string()
        }
    }

    fn format_step_analysis(&self, output: &ProtocolOutput) -> String {
        let mut analysis = String::new();

        for (i, step) in output.steps.iter().enumerate() {
            analysis.push_str(&format!(
                concat!(
                    "### Step {}: {}\n",
                    "- **Confidence**: {:.1}%\n",
                    "- **Status**: {}\n",
                    "- **Duration**: {}ms\n\n",
                ),
                i + 1,
                step.step_id,
                step.confidence * 100.0,
                if step.success { "Success" } else { "Failed" },
                step.duration_ms
            ));
        }

        analysis
    }

    fn build_compliance_section(&self, _output: &ProtocolOutput) -> String {
        "Evaluate GDPR compliance, bias detection, and regulatory alignment.".to_string()
    }

    fn build_meta_cognitive_section(&self, _output: &ProtocolOutput) -> String {
        "Assess reasoning methodology, cognitive biases, and improvement recommendations."
            .to_string()
    }

    fn build_validation_instructions(&self) -> String {
        let mut instructions = String::new();

        instructions.push_str(&format!(
            concat!(
                "**Validation Criteria**:\n",
                "1. Logical Flow Analysis - Check step sequencing and dependency satisfaction\n",
                "2. Confidence Progression - Analyze confidence trends across steps\n",
                "3. Gap Detection - Identify missing reasoning steps or assumptions\n",
                "4. Statistical Significance - {}\n",
                "5. Compliance Assessment - {}\n",
                "6. Meta-Cognitive Evaluation - {}\n\n",
            ),
            if self.config.enable_statistical_testing {
                "Enabled"
            } else {
                "Disabled"
            },
            if self.config.enable_compliance_validation {
                "Enabled"
            } else {
                "Disabled"
            },
            if self.config.enable_meta_cognition {
                "Enabled"
            } else {
                "Disabled"
            }
        ));

        instructions
    }

    fn build_validation_system_prompt(&self) -> String {
        concat!(
            "You are DeepSeek V3 (671B parameters), a reasoning chain validation expert.\n",
            "Your task is to validate AI reasoning chains for logical integrity, statistical significance, compliance, and reasoning quality.\n",
            "\n",
            "**Validation Guidelines**:\n",
            "1. BE BRUTALLY HONEST in your assessment\n",
            "2. Use DeepSeek's 671B parameter scale to detect subtle reasoning flaws\n",
            "3. Apply cross-cultural intelligence to detect biases\n",
            "4. Provide statistical rigor in confidence assessment\n",
            "5. Assess compliance with GDPR and other regulations\n",
            "6. Evaluate reasoning methodology and cognitive processes\n",
            "\n",
            "**Output Format**:\n",
            "- Overall verdict (Validated/PartiallyValidated/Invalid/CriticalIssues)\n",
            "- Chain integrity analysis\n",
            "- Statistical significance (if applicable)\n",
            "- Compliance assessment\n",
            "- Meta-cognitive evaluation\n",
            "- Detailed findings with severity levels\n",
            "- Confidence score for validation (0.0-1.0)\n",
        )
        .to_string()
    }

    async fn perform_statistical_tests(
        &self,
        _output: &ProtocolOutput,
    ) -> Result<StatisticalResult> {
        // Placeholder for statistical testing implementation
        // In production, this would perform bootstrap resampling, hypothesis testing, etc.
        Ok(StatisticalResult {
            significant: true,
            p_value: Some(0.03),
            confidence_interval: Some((0.75, 0.95)),
            sample_size: Some(1000),
        })
    }

    fn estimate_memory_usage(&self, _output: &ProtocolOutput, _trace: &ExecutionTrace) -> f64 {
        // Estimate memory usage based on chain complexity
        50.0 // Conservative estimate in MB
    }

    async fn enhance_rigorous_validation(
        &self,
        mut result: DeepSeekValidationResult,
        _output: &ProtocolOutput,
    ) -> Result<DeepSeekValidationResult> {
        // Add rigorous validation enhancements
        result.validation_confidence *= 0.9; // Apply rigorous discount
        Ok(result)
    }
}

impl Default for DeepSeekValidationEngine {
    fn default() -> Self {
        Self::new().expect("Failed to create default validation engine")
    }
}

// ═══════════════════════════════════════════════════════════════════════════════════════════════
// COMPREHENSIVE TEST SUITE
// ═══════════════════════════════════════════════════════════════════════════════════════════════

#[cfg(test)]
mod tests {
    use super::*;
    use proptest::prelude::*;

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // CONFIGURATION TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod config_tests {
        use super::*;

        #[test]
        fn test_default_config_values() {
            let config = DeepSeekValidationConfig::default();

            assert_eq!(config.model, "deepseek-chat");
            assert!((config.temperature - 0.1).abs() < f32::EPSILON);
            assert_eq!(config.max_tokens, 4000);
            assert!(!config.enable_statistical_testing);
            assert!((config.alpha - 0.05).abs() < f64::EPSILON);
            assert!(config.enable_compliance_validation);
            assert!(!config.enable_meta_cognition);
            assert!((config.min_confidence - 0.70).abs() < f64::EPSILON);
            assert_eq!(config.max_chain_length, 20);
        }

        #[test]
        fn test_rigorous_config() {
            let config = DeepSeekValidationConfig::rigorous();

            assert!(config.enable_statistical_testing);
            assert!(config.enable_compliance_validation);
            assert!(config.enable_meta_cognition);
            assert!((config.temperature - 0.05).abs() < f32::EPSILON);
            assert!((config.min_confidence - 0.85).abs() < f64::EPSILON);
        }

        #[test]
        fn test_performance_config() {
            let config = DeepSeekValidationConfig::performance();

            assert!(!config.enable_statistical_testing);
            assert!(config.enable_compliance_validation);
            assert!(!config.enable_meta_cognition);
            assert!((config.temperature - 0.2).abs() < f32::EPSILON);
            assert_eq!(config.max_tokens, 2000);
        }

        #[test]
        fn test_config_serialization_roundtrip() {
            let config = DeepSeekValidationConfig::rigorous();
            let json = serde_json::to_string(&config).unwrap();
            let deserialized: DeepSeekValidationConfig = serde_json::from_str(&json).unwrap();

            assert_eq!(config.model, deserialized.model);
            assert!((config.temperature - deserialized.temperature).abs() < f32::EPSILON);
            assert_eq!(
                config.enable_statistical_testing,
                deserialized.enable_statistical_testing
            );
        }

        #[test]
        fn test_config_deserialization_with_defaults() {
            // Minimal JSON - should use defaults for missing fields
            let json = r#"{"model": "custom-model"}"#;
            let config: DeepSeekValidationConfig = serde_json::from_str(json).unwrap();

            assert_eq!(config.model, "custom-model");
            // All other fields should have defaults
            assert!((config.temperature - 0.1).abs() < f32::EPSILON);
            assert_eq!(config.max_tokens, 4000);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // TOKEN USAGE TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod token_usage_tests {
        use super::*;

        #[test]
        fn test_token_usage_new() {
            let usage = TokenUsage::new(100, 50, 0.01);

            assert_eq!(usage.input_tokens, 100);
            assert_eq!(usage.output_tokens, 50);
            assert_eq!(usage.total_tokens, 150);
            assert!((usage.cost_usd - 0.01).abs() < f64::EPSILON);
        }

        #[test]
        fn test_token_usage_default() {
            let usage = TokenUsage::default();

            assert_eq!(usage.input_tokens, 0);
            assert_eq!(usage.output_tokens, 0);
            assert_eq!(usage.total_tokens, 0);
            assert!((usage.cost_usd - 0.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_token_usage_add() {
            let mut usage1 = TokenUsage::new(100, 50, 0.01);
            let usage2 = TokenUsage::new(200, 100, 0.02);

            usage1.add(&usage2);

            assert_eq!(usage1.input_tokens, 300);
            assert_eq!(usage1.output_tokens, 150);
            assert_eq!(usage1.total_tokens, 450);
            assert!((usage1.cost_usd - 0.03).abs() < f64::EPSILON);
        }

        #[test]
        fn test_token_usage_add_to_zero() {
            let mut usage = TokenUsage::default();
            let other = TokenUsage::new(500, 250, 0.05);

            usage.add(&other);

            assert_eq!(usage.input_tokens, 500);
            assert_eq!(usage.output_tokens, 250);
            assert_eq!(usage.total_tokens, 750);
        }

        #[test]
        fn test_token_usage_boundary_max_u32() {
            let usage = TokenUsage::new(u32::MAX - 10, 5, 0.0);
            assert_eq!(usage.total_tokens, u32::MAX - 5);
        }

        #[test]
        fn test_token_usage_serialization() {
            let usage = TokenUsage::new(1000, 500, 0.123);
            let json = serde_json::to_string(&usage).unwrap();
            let deserialized: TokenUsage = serde_json::from_str(&json).unwrap();

            assert_eq!(usage.input_tokens, deserialized.input_tokens);
            assert_eq!(usage.output_tokens, deserialized.output_tokens);
            assert_eq!(usage.total_tokens, deserialized.total_tokens);
            assert!((usage.cost_usd - deserialized.cost_usd).abs() < 0.0001);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // CHAIN INTEGRITY RESULT ARITHMETIC TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod chain_integrity_arithmetic_tests {
        use super::*;

        fn create_chain_integrity(continuity_score: f64) -> ChainIntegrityResult {
            ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Good,
                step_dependencies: DependencyStatus::FullySatisfied,
                confidence_progression: ProgressionStatus::Monotonic,
                gaps_detected: vec![],
                continuity_score,
            }
        }

        #[test]
        fn test_chain_integrity_add_f32() {
            let result = create_chain_integrity(0.85);
            let sum = result + 0.15f32;
            assert!((sum - 1.0).abs() < 0.0001);
        }

        #[test]
        fn test_chain_integrity_sub_f32_ref() {
            let result = create_chain_integrity(0.85);
            let diff = result - &0.35f32;
            assert!((diff - 0.5).abs() < 0.0001);
        }

        #[test]
        fn test_chain_integrity_mul_f64() {
            let result = create_chain_integrity(0.5);
            let product = result * 2.0f64;
            assert!((product - 1.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_chain_integrity_mul_f32() {
            let result = create_chain_integrity(0.5);
            let product = result * 4.0f32;
            assert!((product - 2.0).abs() < 0.0001);
        }

        #[test]
        fn test_chain_integrity_boundary_zero() {
            let result = create_chain_integrity(0.0);
            let product = result * 100.0f64;
            assert!((product - 0.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_chain_integrity_boundary_one() {
            let result = create_chain_integrity(1.0);
            let product = result * 0.5f64;
            assert!((product - 0.5).abs() < f64::EPSILON);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // VALIDATION PERFORMANCE TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod validation_performance_tests {
        use super::*;

        #[test]
        fn test_validation_performance_new() {
            let perf = ValidationPerformance::new(1000, 500.0, 128.5);

            assert_eq!(perf.duration_ms, 1000);
            assert!((perf.tokens_per_second - 500.0).abs() < f64::EPSILON);
            assert!((perf.memory_usage_mb - 128.5).abs() < f64::EPSILON);
        }

        #[test]
        fn test_validation_performance_default() {
            let perf = ValidationPerformance::default();

            assert_eq!(perf.duration_ms, 0);
            assert!((perf.tokens_per_second - 0.0).abs() < f64::EPSILON);
            assert!((perf.memory_usage_mb - 0.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_validation_performance_serialization() {
            let perf = ValidationPerformance::new(5000, 1200.5, 256.0);
            let json = serde_json::to_string(&perf).unwrap();
            let deserialized: ValidationPerformance = serde_json::from_str(&json).unwrap();

            assert_eq!(perf.duration_ms, deserialized.duration_ms);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // ENUM SERIALIZATION TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod enum_serialization_tests {
        use super::*;

        #[test]
        fn test_validation_verdict_serialization() {
            let cases = [
                (ValidationVerdict::Validated, "\"validated\""),
                (
                    ValidationVerdict::PartiallyValidated,
                    "\"partially_validated\"",
                ),
                (ValidationVerdict::NeedsImprovement, "\"needs_improvement\""),
                (ValidationVerdict::Invalid, "\"invalid\""),
                (ValidationVerdict::CriticalIssues, "\"critical_issues\""),
            ];

            for (verdict, expected_json) in cases {
                let json = serde_json::to_string(&verdict).unwrap();
                assert_eq!(json, expected_json);

                let deserialized: ValidationVerdict = serde_json::from_str(&json).unwrap();
                assert_eq!(verdict, deserialized);
            }
        }

        #[test]
        fn test_severity_serialization() {
            let cases = [
                (Severity::Critical, "\"critical\""),
                (Severity::High, "\"high\""),
                (Severity::Medium, "\"medium\""),
                (Severity::Low, "\"low\""),
                (Severity::Info, "\"info\""),
            ];

            for (severity, expected_json) in cases {
                let json = serde_json::to_string(&severity).unwrap();
                assert_eq!(json, expected_json);

                let deserialized: Severity = serde_json::from_str(&json).unwrap();
                assert_eq!(severity, deserialized);
            }
        }

        #[test]
        fn test_bias_level_serialization() {
            let cases = [
                (BiasLevel::Minimal, "\"minimal\""),
                (BiasLevel::Low, "\"low\""),
                (BiasLevel::Moderate, "\"moderate\""),
                (BiasLevel::High, "\"high\""),
                (BiasLevel::Severe, "\"severe\""),
            ];

            for (level, expected_json) in cases {
                let json = serde_json::to_string(&level).unwrap();
                assert_eq!(json, expected_json);
            }
        }

        #[test]
        fn test_logical_flow_status_serialization() {
            let status = LogicalFlowStatus::Excellent;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"excellent\"");
        }

        #[test]
        fn test_dependency_status_serialization() {
            let status = DependencyStatus::PartiallySatisfied;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"partially_satisfied\"");
        }

        #[test]
        fn test_progression_status_serialization() {
            let status = ProgressionStatus::SlowlyDecaying;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"slowly_decaying\"");
        }

        #[test]
        fn test_compliance_status_serialization() {
            let status = ComplianceStatus::CriticalNonCompliance;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"critical_non_compliance\"");
        }

        #[test]
        fn test_regulatory_status_serialization() {
            let status = RegulatoryStatus::NeedsValidation;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"needs_validation\"");
        }

        #[test]
        fn test_methodology_status_serialization() {
            let status = MethodologyStatus::NeedsImprovement;
            let json = serde_json::to_string(&status).unwrap();
            assert_eq!(json, "\"needs_improvement\"");
        }

        #[test]
        fn test_validation_category_serialization() {
            let category = ValidationCategory::EnterpriseCompliance;
            let json = serde_json::to_string(&category).unwrap();
            assert_eq!(json, "\"enterprise_compliance\"");
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // STATISTICAL RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod statistical_result_tests {
        use super::*;

        #[test]
        fn test_statistical_result_significant() {
            let result = StatisticalResult {
                significant: true,
                p_value: Some(0.01),
                confidence_interval: Some((0.80, 0.95)),
                sample_size: Some(1000),
            };

            assert!(result.significant);
            assert!(result.p_value.unwrap() < 0.05);
        }

        #[test]
        fn test_statistical_result_not_significant() {
            let result = StatisticalResult {
                significant: false,
                p_value: Some(0.15),
                confidence_interval: None,
                sample_size: Some(50),
            };

            assert!(!result.significant);
            assert!(result.p_value.unwrap() > 0.05);
        }

        #[test]
        fn test_statistical_result_serialization() {
            let result = StatisticalResult {
                significant: true,
                p_value: Some(0.03),
                confidence_interval: Some((0.75, 0.95)),
                sample_size: Some(500),
            };

            let json = serde_json::to_string(&result).unwrap();
            let deserialized: StatisticalResult = serde_json::from_str(&json).unwrap();

            assert_eq!(result.significant, deserialized.significant);
            assert_eq!(result.p_value, deserialized.p_value);
            assert_eq!(result.confidence_interval, deserialized.confidence_interval);
        }

        #[test]
        fn test_statistical_result_boundary_p_value() {
            // Test boundary at alpha = 0.05
            let exactly_significant = StatisticalResult {
                significant: true,
                p_value: Some(0.049),
                confidence_interval: None,
                sample_size: None,
            };
            assert!(exactly_significant.p_value.unwrap() < 0.05);

            let not_significant = StatisticalResult {
                significant: false,
                p_value: Some(0.051),
                confidence_interval: None,
                sample_size: None,
            };
            assert!(not_significant.p_value.unwrap() > 0.05);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // VALIDATION FINDING TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod validation_finding_tests {
        use super::*;

        #[test]
        fn test_validation_finding_creation() {
            let finding = ValidationFinding {
                category: ValidationCategory::LogicalFlow,
                severity: Severity::High,
                description: "Gap detected between steps 2 and 3".to_string(),
                affected_steps: vec!["step_2".to_string(), "step_3".to_string()],
                evidence: vec!["Missing intermediate reasoning".to_string()],
                recommendations: vec!["Add bridging step".to_string()],
            };

            assert_eq!(finding.category, ValidationCategory::LogicalFlow);
            assert_eq!(finding.severity, Severity::High);
            assert_eq!(finding.affected_steps.len(), 2);
        }

        #[test]
        fn test_validation_finding_with_unicode() {
            let finding = ValidationFinding {
                category: ValidationCategory::BiasDetection,
                severity: Severity::Medium,
                description: "Unicode test: \u{1F4A1} \u{1F4CA} \u{2705}".to_string(),
                affected_steps: vec!["step_\u{03B1}".to_string()],
                evidence: vec!["Evidence with emoji: \u{1F914}".to_string()],
                recommendations: vec!["\u{2728} Improve coverage".to_string()],
            };

            let json = serde_json::to_string(&finding).unwrap();
            let deserialized: ValidationFinding = serde_json::from_str(&json).unwrap();

            assert!(deserialized.description.contains('\u{1F4A1}'));
        }

        #[test]
        fn test_validation_finding_empty_vectors() {
            let finding = ValidationFinding {
                category: ValidationCategory::Methodology,
                severity: Severity::Info,
                description: "Minor observation".to_string(),
                affected_steps: vec![],
                evidence: vec![],
                recommendations: vec![],
            };

            assert!(finding.affected_steps.is_empty());
            assert!(finding.evidence.is_empty());
            assert!(finding.recommendations.is_empty());
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // DETECTED BIAS TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod detected_bias_tests {
        use super::*;

        #[test]
        fn test_detected_bias_creation() {
            let bias = DetectedBias {
                bias_type: "confirmation_bias".to_string(),
                evidence: "Selectively interpreting data".to_string(),
                severity: Severity::Medium,
                detection_confidence: 0.85,
            };

            assert_eq!(bias.bias_type, "confirmation_bias");
            assert!((bias.detection_confidence - 0.85).abs() < f64::EPSILON);
        }

        #[test]
        fn test_detected_bias_confidence_boundaries() {
            let low_confidence = DetectedBias {
                bias_type: "anchoring".to_string(),
                evidence: "Weak signal".to_string(),
                severity: Severity::Low,
                detection_confidence: 0.0,
            };
            assert!((low_confidence.detection_confidence - 0.0).abs() < f64::EPSILON);

            let high_confidence = DetectedBias {
                bias_type: "anchoring".to_string(),
                evidence: "Strong signal".to_string(),
                severity: Severity::Critical,
                detection_confidence: 1.0,
            };
            assert!((high_confidence.detection_confidence - 1.0).abs() < f64::EPSILON);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // COMPLIANCE VIOLATION TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod compliance_violation_tests {
        use super::*;

        #[test]
        fn test_compliance_violation_creation() {
            let violation = ComplianceViolation {
                violation_type: "GDPR_ARTICLE_17".to_string(),
                severity: Severity::Critical,
                description: "Right to erasure not implemented".to_string(),
                remediation: vec![
                    "Implement data deletion endpoint".to_string(),
                    "Add audit logging".to_string(),
                ],
            };

            assert_eq!(violation.violation_type, "GDPR_ARTICLE_17");
            assert_eq!(violation.remediation.len(), 2);
        }

        #[test]
        fn test_compliance_violation_serialization() {
            let violation = ComplianceViolation {
                violation_type: "SOX_COMPLIANCE".to_string(),
                severity: Severity::High,
                description: "Audit trail incomplete".to_string(),
                remediation: vec!["Add comprehensive logging".to_string()],
            };

            let json = serde_json::to_string(&violation).unwrap();
            assert!(json.contains("SOX_COMPLIANCE"));

            let deserialized: ComplianceViolation = serde_json::from_str(&json).unwrap();
            assert_eq!(violation.violation_type, deserialized.violation_type);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // BIAS DETECTION RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod bias_detection_result_tests {
        use super::*;

        #[test]
        fn test_bias_detection_result_no_biases() {
            let result = BiasDetectionResult {
                overall_bias: BiasLevel::Minimal,
                detected_biases: vec![],
                mitigation_recommendations: vec![],
            };

            assert_eq!(result.overall_bias, BiasLevel::Minimal);
            assert!(result.detected_biases.is_empty());
        }

        #[test]
        fn test_bias_detection_result_multiple_biases() {
            let result = BiasDetectionResult {
                overall_bias: BiasLevel::Moderate,
                detected_biases: vec![
                    DetectedBias {
                        bias_type: "confirmation".to_string(),
                        evidence: "Evidence 1".to_string(),
                        severity: Severity::Medium,
                        detection_confidence: 0.7,
                    },
                    DetectedBias {
                        bias_type: "availability".to_string(),
                        evidence: "Evidence 2".to_string(),
                        severity: Severity::Low,
                        detection_confidence: 0.6,
                    },
                ],
                mitigation_recommendations: vec![
                    "Consider alternative perspectives".to_string(),
                    "Use structured decision frameworks".to_string(),
                ],
            };

            assert_eq!(result.detected_biases.len(), 2);
            assert_eq!(result.mitigation_recommendations.len(), 2);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // META-COGNITIVE RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod meta_cognitive_result_tests {
        use super::*;

        #[test]
        fn test_meta_cognitive_result_creation() {
            let result = MetaCognitiveResult {
                reasoning_quality: 85.5,
                methodology_quality: MethodologyStatus::Good,
                cognitive_biases: vec!["overconfidence".to_string()],
                recommendations: vec!["Seek external validation".to_string()],
            };

            assert!((result.reasoning_quality - 85.5).abs() < f64::EPSILON);
            assert_eq!(result.methodology_quality, MethodologyStatus::Good);
        }

        #[test]
        fn test_meta_cognitive_result_boundary_scores() {
            let min_score = MetaCognitiveResult {
                reasoning_quality: 0.0,
                methodology_quality: MethodologyStatus::Poor,
                cognitive_biases: vec![],
                recommendations: vec![],
            };
            assert!((min_score.reasoning_quality - 0.0).abs() < f64::EPSILON);

            let max_score = MetaCognitiveResult {
                reasoning_quality: 100.0,
                methodology_quality: MethodologyStatus::Excellent,
                cognitive_biases: vec![],
                recommendations: vec![],
            };
            assert!((max_score.reasoning_quality - 100.0).abs() < f64::EPSILON);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // CHAIN INTEGRITY RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod chain_integrity_result_tests {
        use super::*;

        #[test]
        fn test_chain_integrity_result_creation() {
            let result = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Excellent,
                step_dependencies: DependencyStatus::FullySatisfied,
                confidence_progression: ProgressionStatus::Monotonic,
                gaps_detected: vec![],
                continuity_score: 0.95,
            };

            assert_eq!(result.logical_flow, LogicalFlowStatus::Excellent);
            assert!((result.continuity_score - 0.95).abs() < f64::EPSILON);
        }

        #[test]
        fn test_chain_integrity_result_with_gaps() {
            let result = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::NeedsImprovement,
                step_dependencies: DependencyStatus::PartiallySatisfied,
                confidence_progression: ProgressionStatus::Erratic,
                gaps_detected: vec![
                    "Missing justification between step 1 and 2".to_string(),
                    "Unexplained confidence drop at step 4".to_string(),
                ],
                continuity_score: 0.45,
            };

            assert_eq!(result.gaps_detected.len(), 2);
            assert!(result.continuity_score < 0.5);
        }

        #[test]
        fn test_chain_integrity_continuity_score_boundaries() {
            let zero_score = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Poor,
                step_dependencies: DependencyStatus::Unsatisfied,
                confidence_progression: ProgressionStatus::Unstable,
                gaps_detected: vec![],
                continuity_score: 0.0,
            };
            assert!((zero_score.continuity_score - 0.0).abs() < f64::EPSILON);

            let perfect_score = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Excellent,
                step_dependencies: DependencyStatus::FullySatisfied,
                confidence_progression: ProgressionStatus::Monotonic,
                gaps_detected: vec![],
                continuity_score: 1.0,
            };
            assert!((perfect_score.continuity_score - 1.0).abs() < f64::EPSILON);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // DEEPSEEK VALIDATION RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod deepseek_validation_result_tests {
        use super::*;

        fn create_minimal_result() -> DeepSeekValidationResult {
            DeepSeekValidationResult {
                verdict: ValidationVerdict::Validated,
                chain_integrity: ChainIntegrityResult {
                    logical_flow: LogicalFlowStatus::Good,
                    step_dependencies: DependencyStatus::FullySatisfied,
                    confidence_progression: ProgressionStatus::Monotonic,
                    gaps_detected: vec![],
                    continuity_score: 0.85,
                },
                statistical_results: None,
                compliance_results: None,
                meta_cognitive_results: None,
                validation_confidence: 0.90,
                findings: vec![],
                tokens_used: TokenUsage::default(),
                performance: ValidationPerformance::default(),
            }
        }

        #[test]
        fn test_validation_result_minimal() {
            let result = create_minimal_result();

            assert_eq!(result.verdict, ValidationVerdict::Validated);
            assert!(result.statistical_results.is_none());
            assert!(result.compliance_results.is_none());
            assert!(result.meta_cognitive_results.is_none());
        }

        #[test]
        fn test_validation_result_serialization_skip_none() {
            let result = create_minimal_result();
            let json = serde_json::to_string(&result).unwrap();

            // Optional None fields should be skipped due to skip_serializing_if
            assert!(!json.contains("statistical_results"));
            assert!(!json.contains("compliance_results"));
            assert!(!json.contains("meta_cognitive_results"));
        }

        #[test]
        fn test_validation_result_with_all_options() {
            let result = DeepSeekValidationResult {
                verdict: ValidationVerdict::PartiallyValidated,
                chain_integrity: ChainIntegrityResult {
                    logical_flow: LogicalFlowStatus::Satisfactory,
                    step_dependencies: DependencyStatus::MostlySatisfied,
                    confidence_progression: ProgressionStatus::SlowlyDecaying,
                    gaps_detected: vec!["Minor gap".to_string()],
                    continuity_score: 0.75,
                },
                statistical_results: Some(StatisticalResult {
                    significant: true,
                    p_value: Some(0.02),
                    confidence_interval: Some((0.70, 0.90)),
                    sample_size: Some(100),
                }),
                compliance_results: Some(ComplianceResult {
                    gdpr_compliance: ComplianceStatus::MinorIssues,
                    bias_detection: BiasDetectionResult {
                        overall_bias: BiasLevel::Low,
                        detected_biases: vec![],
                        mitigation_recommendations: vec![],
                    },
                    regulatory_alignment: RegulatoryStatus::NeedsValidation,
                    violations: vec![],
                }),
                meta_cognitive_results: Some(MetaCognitiveResult {
                    reasoning_quality: 78.0,
                    methodology_quality: MethodologyStatus::Good,
                    cognitive_biases: vec![],
                    recommendations: vec![],
                }),
                validation_confidence: 0.80,
                findings: vec![ValidationFinding {
                    category: ValidationCategory::LogicalFlow,
                    severity: Severity::Low,
                    description: "Minor observation".to_string(),
                    affected_steps: vec![],
                    evidence: vec![],
                    recommendations: vec![],
                }],
                tokens_used: TokenUsage::new(1000, 500, 0.05),
                performance: ValidationPerformance::new(2000, 750.0, 64.0),
            };

            let json = serde_json::to_string(&result).unwrap();
            let deserialized: DeepSeekValidationResult = serde_json::from_str(&json).unwrap();

            assert_eq!(result.verdict, deserialized.verdict);
            assert!(deserialized.statistical_results.is_some());
            assert!(deserialized.compliance_results.is_some());
            assert!(deserialized.meta_cognitive_results.is_some());
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // COMPLIANCE RESULT TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod compliance_result_tests {
        use super::*;

        #[test]
        fn test_compliance_result_fully_compliant() {
            let result = ComplianceResult {
                gdpr_compliance: ComplianceStatus::Compliant,
                bias_detection: BiasDetectionResult {
                    overall_bias: BiasLevel::Minimal,
                    detected_biases: vec![],
                    mitigation_recommendations: vec![],
                },
                regulatory_alignment: RegulatoryStatus::FullyCompliant,
                violations: vec![],
            };

            assert_eq!(result.gdpr_compliance, ComplianceStatus::Compliant);
            assert!(result.violations.is_empty());
        }

        #[test]
        fn test_compliance_result_with_violations() {
            let result = ComplianceResult {
                gdpr_compliance: ComplianceStatus::NonCompliant,
                bias_detection: BiasDetectionResult {
                    overall_bias: BiasLevel::High,
                    detected_biases: vec![DetectedBias {
                        bias_type: "demographic".to_string(),
                        evidence: "Unequal treatment detected".to_string(),
                        severity: Severity::High,
                        detection_confidence: 0.9,
                    }],
                    mitigation_recommendations: vec!["Implement fairness constraints".to_string()],
                },
                regulatory_alignment: RegulatoryStatus::NonCompliant,
                violations: vec![
                    ComplianceViolation {
                        violation_type: "GDPR_CONSENT".to_string(),
                        severity: Severity::Critical,
                        description: "Missing consent mechanism".to_string(),
                        remediation: vec!["Add consent UI".to_string()],
                    },
                    ComplianceViolation {
                        violation_type: "GDPR_DATA_RETENTION".to_string(),
                        severity: Severity::High,
                        description: "Data retained beyond policy".to_string(),
                        remediation: vec!["Implement data lifecycle management".to_string()],
                    },
                ],
            };

            assert_eq!(result.violations.len(), 2);
            assert_eq!(result.bias_detection.detected_biases.len(), 1);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // UNICODE AND SPECIAL CHARACTER TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod unicode_tests {
        use super::*;

        #[test]
        fn test_unicode_in_chain_integrity_gaps() {
            let result = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Satisfactory,
                step_dependencies: DependencyStatus::MostlySatisfied,
                confidence_progression: ProgressionStatus::SlowlyDecaying,
                gaps_detected: vec![
                    "\u{4E2D}\u{6587}\u{6D4B}\u{8BD5}".to_string(), // Chinese: "Chinese test"
                    "\u{65E5}\u{672C}\u{8A9E}".to_string(),         // Japanese: "Japanese"
                    "\u{D55C}\u{AD6D}\u{C5B4}".to_string(),         // Korean: "Korean"
                    "\u{0410}\u{0411}\u{0412}".to_string(),         // Russian: "ABV"
                    "\u{05D0}\u{05D1}\u{05D2}".to_string(),         // Hebrew: "Alef Bet Gimel"
                    "\u{0627}\u{0628}\u{062A}".to_string(),         // Arabic: letters
                ],
                continuity_score: 0.7,
            };

            let json = serde_json::to_string(&result).unwrap();
            let deserialized: ChainIntegrityResult = serde_json::from_str(&json).unwrap();

            assert_eq!(result.gaps_detected.len(), deserialized.gaps_detected.len());
            for (original, parsed) in result
                .gaps_detected
                .iter()
                .zip(deserialized.gaps_detected.iter())
            {
                assert_eq!(original, parsed);
            }
        }

        #[test]
        fn test_unicode_emoji_in_descriptions() {
            let finding = ValidationFinding {
                category: ValidationCategory::Methodology,
                severity: Severity::Info,
                description: "Status: \u{2705} Pass \u{274C} Fail \u{26A0} Warning".to_string(),
                affected_steps: vec!["\u{1F3AF} Target Step".to_string()],
                evidence: vec!["\u{1F4CA} Chart data shows trend".to_string()],
                recommendations: vec!["\u{1F4A1} Consider improvement".to_string()],
            };

            let json = serde_json::to_string(&finding).unwrap();
            let deserialized: ValidationFinding = serde_json::from_str(&json).unwrap();

            assert!(deserialized.description.contains('\u{2705}'));
            assert!(deserialized.description.contains('\u{274C}'));
        }

        #[test]
        fn test_special_characters_in_strings() {
            let violation = ComplianceViolation {
                violation_type: "TEST_<>\"'&".to_string(),
                severity: Severity::Low,
                description: "Special chars: <script>alert('xss')</script>".to_string(),
                remediation: vec!["Use proper escaping: &amp; &lt; &gt;".to_string()],
            };

            let json = serde_json::to_string(&violation).unwrap();
            let deserialized: ComplianceViolation = serde_json::from_str(&json).unwrap();

            assert!(deserialized.violation_type.contains('<'));
            assert!(deserialized.violation_type.contains('>'));
        }

        #[test]
        fn test_empty_strings() {
            let bias = DetectedBias {
                bias_type: "".to_string(),
                evidence: "".to_string(),
                severity: Severity::Info,
                detection_confidence: 0.0,
            };

            let json = serde_json::to_string(&bias).unwrap();
            let deserialized: DetectedBias = serde_json::from_str(&json).unwrap();

            assert!(deserialized.bias_type.is_empty());
            assert!(deserialized.evidence.is_empty());
        }

        #[test]
        fn test_very_long_strings() {
            let long_string = "A".repeat(100_000);

            let finding = ValidationFinding {
                category: ValidationCategory::TokenEfficiency,
                severity: Severity::Low,
                description: long_string.clone(),
                affected_steps: vec![],
                evidence: vec![],
                recommendations: vec![],
            };

            let json = serde_json::to_string(&finding).unwrap();
            let deserialized: ValidationFinding = serde_json::from_str(&json).unwrap();

            assert_eq!(deserialized.description.len(), 100_000);
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // PROPERTY-BASED TESTS (using proptest)
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod property_tests {
        use super::*;

        proptest! {
            #[test]
            fn test_token_usage_add_is_associative(
                input1 in 0u32..1000,
                output1 in 0u32..1000,
                cost1 in 0.0f64..10.0,
                input2 in 0u32..1000,
                output2 in 0u32..1000,
                cost2 in 0.0f64..10.0,
            ) {
                let mut usage1 = TokenUsage::new(input1, output1, cost1);
                let usage2 = TokenUsage::new(input2, output2, cost2);

                usage1.add(&usage2);

                prop_assert_eq!(usage1.input_tokens, input1 + input2);
                prop_assert_eq!(usage1.output_tokens, output1 + output2);
                prop_assert_eq!(usage1.total_tokens, input1 + output1 + input2 + output2);
                prop_assert!((usage1.cost_usd - (cost1 + cost2)).abs() < 0.0001);
            }

            #[test]
            fn test_token_usage_new_total_is_sum(input in 0u32..u32::MAX/2, output in 0u32..u32::MAX/2) {
                let usage = TokenUsage::new(input, output, 0.0);
                prop_assert_eq!(usage.total_tokens, input + output);
            }

            #[test]
            fn test_chain_integrity_mul_f64_preserves_zero(rhs in -1000.0f64..1000.0) {
                let result = ChainIntegrityResult {
                    logical_flow: LogicalFlowStatus::Good,
                    step_dependencies: DependencyStatus::FullySatisfied,
                    confidence_progression: ProgressionStatus::Monotonic,
                    gaps_detected: vec![],
                    continuity_score: 0.0,
                };

                let product = result * rhs;
                prop_assert!((product - 0.0).abs() < f64::EPSILON);
            }

            #[test]
            fn test_chain_integrity_mul_identity(score in 0.0f64..1.0) {
                let result = ChainIntegrityResult {
                    logical_flow: LogicalFlowStatus::Good,
                    step_dependencies: DependencyStatus::FullySatisfied,
                    confidence_progression: ProgressionStatus::Monotonic,
                    gaps_detected: vec![],
                    continuity_score: score,
                };

                let product = result * 1.0f64;
                prop_assert!((product - score).abs() < f64::EPSILON);
            }

            #[test]
            fn test_validation_performance_serialization_roundtrip(
                duration in 0u64..u64::MAX,
                tps in 0.0f64..10000.0,
                memory in 0.0f64..10000.0,
            ) {
                let perf = ValidationPerformance::new(duration, tps, memory);
                let json = serde_json::to_string(&perf).unwrap();
                let deserialized: ValidationPerformance = serde_json::from_str(&json).unwrap();

                prop_assert_eq!(perf.duration_ms, deserialized.duration_ms);
                // Floating point comparison with tolerance
                prop_assert!((perf.tokens_per_second - deserialized.tokens_per_second).abs() < 0.0001);
            }

            #[test]
            fn test_config_temperature_in_valid_range(temp in 0.0f32..2.0) {
                // Temperature should be preserved through serialization
                let config = DeepSeekValidationConfig {
                    temperature: temp,
                    ..Default::default()
                };

                let json = serde_json::to_string(&config).unwrap();
                let deserialized: DeepSeekValidationConfig = serde_json::from_str(&json).unwrap();

                prop_assert!((config.temperature - deserialized.temperature).abs() < 0.0001);
            }

            #[test]
            fn test_arbitrary_string_in_gaps(s in "\\PC*") {
                // Test that any valid UTF-8 string can be stored in gaps
                let result = ChainIntegrityResult {
                    logical_flow: LogicalFlowStatus::Good,
                    step_dependencies: DependencyStatus::FullySatisfied,
                    confidence_progression: ProgressionStatus::Monotonic,
                    gaps_detected: vec![s.clone()],
                    continuity_score: 0.5,
                };

                let json = serde_json::to_string(&result).unwrap();
                let deserialized: ChainIntegrityResult = serde_json::from_str(&json).unwrap();

                prop_assert_eq!(result.gaps_detected, deserialized.gaps_detected);
            }
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // VALIDATION ENGINE TESTS (Requires API key - skipped in CI)
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod engine_tests {
        use super::*;

        #[tokio::test]
        async fn test_validation_engine_creation() {
            // This test may fail if DEEPSEEK_API_KEY is not set
            // In production, this would be mocked
            let result = DeepSeekValidationEngine::new();

            // Engine creation should succeed regardless of API key (key checked on use)
            assert!(result.is_ok());

            if let Ok(engine) = result {
                assert_eq!(engine.config.model, "deepseek-chat");
            }
        }

        #[test]
        fn test_configuration_options() {
            let config = DeepSeekValidationConfig::rigorous();
            assert!(config.enable_statistical_testing);
            assert!(config.enable_compliance_validation);
            assert!(config.enable_meta_cognition);

            let perf_config = DeepSeekValidationConfig::performance();
            assert!(!perf_config.enable_statistical_testing);
            assert!(perf_config.enable_compliance_validation);
        }

        #[test]
        fn test_validation_engine_with_custom_config() {
            let config = DeepSeekValidationConfig {
                model: "deepseek-coder".to_string(),
                temperature: 0.0,
                max_tokens: 8000,
                enable_statistical_testing: true,
                alpha: 0.01,
                enable_compliance_validation: true,
                enable_meta_cognition: true,
                min_confidence: 0.90,
                max_chain_length: 50,
            };

            let result = DeepSeekValidationEngine::with_config(config);
            assert!(result.is_ok());

            if let Ok(engine) = result {
                assert_eq!(engine.config.model, "deepseek-coder");
                assert_eq!(engine.config.max_tokens, 8000);
                assert!((engine.config.alpha - 0.01).abs() < f64::EPSILON);
            }
        }
    }

    // ─────────────────────────────────────────────────────────────────────────────────────────────
    // EDGE CASE TESTS
    // ─────────────────────────────────────────────────────────────────────────────────────────────

    mod edge_case_tests {
        use super::*;

        #[test]
        fn test_token_usage_zero_values() {
            let usage = TokenUsage::new(0, 0, 0.0);
            assert_eq!(usage.total_tokens, 0);
            assert!((usage.cost_usd - 0.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_validation_performance_zero_duration() {
            let perf = ValidationPerformance::new(0, 0.0, 0.0);
            // Division by zero should be handled gracefully in actual usage
            assert_eq!(perf.duration_ms, 0);
        }

        #[test]
        fn test_continuity_score_exactly_zero() {
            let result = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Poor,
                step_dependencies: DependencyStatus::Unsatisfied,
                confidence_progression: ProgressionStatus::Unstable,
                gaps_detected: vec![],
                continuity_score: 0.0,
            };

            // Multiplying by any value should still be zero
            let product = result.clone() * 100.0f64;
            assert!((product - 0.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_continuity_score_exactly_one() {
            let result = ChainIntegrityResult {
                logical_flow: LogicalFlowStatus::Excellent,
                step_dependencies: DependencyStatus::FullySatisfied,
                confidence_progression: ProgressionStatus::Monotonic,
                gaps_detected: vec![],
                continuity_score: 1.0,
            };

            // Multiplying by identity should preserve value
            let product = result * 1.0f64;
            assert!((product - 1.0).abs() < f64::EPSILON);
        }

        #[test]
        fn test_negative_cost_handling() {
            // While unusual, negative cost might be used for credits/refunds
            let usage = TokenUsage::new(100, 50, -0.01);
            assert!(usage.cost_usd < 0.0);
        }

        #[test]
        fn test_very_small_confidence_values() {
            let result = StatisticalResult {
                significant: true,
                p_value: Some(1e-100),
                confidence_interval: Some((0.999999, 0.9999999)),
                sample_size: Some(1_000_000),
            };

            let json = serde_json::to_string(&result).unwrap();
            let deserialized: StatisticalResult = serde_json::from_str(&json).unwrap();

            // Very small p-values should be preserved
            assert!(deserialized.p_value.unwrap() < 1e-50);
        }

        #[test]
        fn test_default_function_values() {
            // Test all default functions directly
            assert_eq!(default_deepseek_model(), "deepseek-chat");
            assert!((default_validation_temperature() - 0.1).abs() < f32::EPSILON);
            assert_eq!(default_max_tokens(), 4000);
            assert!((default_alpha() - 0.05).abs() < f64::EPSILON);
            assert!((default_min_confidence() - 0.70).abs() < f64::EPSILON);
            assert_eq!(default_max_chain_length(), 20);
        }
    }
}