rust-rule-engine 1.0.2-alpha

use rust_rule_engine::engine::engine::RustRuleEngine;
use rust_rule_engine::engine::plugin::{PluginHealth, PluginMetadata, PluginState, RulePlugin};
use rust_rule_engine::errors::Result;
use rust_rule_engine::types::Value;
use std::collections::HashMap;

/// AI/ML Integration Plugin for machine learning and AI operations
pub struct AIMLPlugin {
    metadata: PluginMetadata,
}

impl AIMLPlugin {
    pub fn new() -> Self {
        Self {
            metadata: PluginMetadata {
                name: "ai-ml".to_string(),
                version: "1.0.0".to_string(),
                description: "AI/ML integration for predictions, NLP, and data analysis"
                    .to_string(),
                author: "Rust Rule Engine Team".to_string(),
                state: PluginState::Loaded,
                health: PluginHealth::Healthy,
                actions: vec![
                    "PredictValue".to_string(),
                    "ClassifyText".to_string(),
                    "SentimentAnalysis".to_string(),
                    "GenerateEmbedding".to_string(),
                    "DetectAnomalies".to_string(),
                    "RecommendItems".to_string(),
                    "ChatCompletion".to_string(),
                    "ImageRecognition".to_string(),
                    "TranslateText".to_string(),
                ],
                functions: vec![
                    "calculateSimilarity".to_string(),
                    "normalizeScore".to_string(),
                    "extractKeywords".to_string(),
                    "tokenizeText".to_string(),
                    "preprocessData".to_string(),
                    "evaluateModel".to_string(),
                    "featureEngineering".to_string(),
                ],
                dependencies: vec![
                    "candle-core".to_string(),
                    "tokenizers".to_string(),
                    "reqwest".to_string(),
                    "serde_json".to_string(),
                ],
            },
        }
    }
}

impl RulePlugin for AIMLPlugin {
    fn get_metadata(&self) -> &PluginMetadata {
        &self.metadata
    }

    fn register_actions(&self, engine: &mut RustRuleEngine) -> Result<()> {
        // Predict Value action
        engine.register_action_handler("PredictValue", |params, facts| {
            let model_name = params
                .get("0")
                .map(|v| v.to_string())
                .unwrap_or("default_model".to_string());
            let input_data = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("{}".to_string());
            let result_key = params
                .get("2")
                .map(|v| v.to_string())
                .unwrap_or("prediction".to_string());

            println!("🤖 ML PREDICTION:");
            println!("   Model: {}", model_name);
            println!("   Input: {}", input_data);

            // Simulate prediction based on model type
            let prediction = match model_name.as_str() {
                "fraud_detection" => {
                    println!("   🔍 Fraud Detection Score: 0.15 (Low Risk)");
                    Value::Number(0.15)
                }
                "price_prediction" => {
                    println!("   💰 Predicted Price: $127.45");
                    Value::Number(127.45)
                }
                "customer_churn" => {
                    println!("   📊 Churn Probability: 0.23 (23%)");
                    Value::Number(0.23)
                }
                _ => {
                    println!("   📈 Prediction Score: 0.78");
                    Value::Number(0.78)
                }
            };

            facts.add_value(&result_key, prediction)?;
            facts.add_value("model_confidence", Value::Number(0.92))?;
            Ok(())
        });

        // Classify Text action
        engine.register_action_handler("ClassifyText", |params, facts| {
            let text = params.get("0").map(|v| v.to_string()).unwrap_or_default();
            let categories = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("general".to_string());

            println!("📝 TEXT CLASSIFICATION:");
            println!("   Text: \"{}\"", text);
            println!("   Categories: {}", categories);

            // Simulate classification based on content
            let (category, confidence) =
                if text.to_lowercase().contains("urgent") || text.contains("emergency") {
                    ("urgent", 0.95)
                } else if text.to_lowercase().contains("spam") || text.contains("offer") {
                    ("spam", 0.88)
                } else if text.to_lowercase().contains("question") || text.contains("help") {
                    ("support", 0.76)
                } else if text.to_lowercase().contains("bug") || text.contains("error") {
                    ("technical", 0.84)
                } else {
                    ("general", 0.65)
                };

            println!(
                "   🏷️  Category: {} (confidence: {:.2})",
                category, confidence
            );

            facts.add_value("text_category", Value::String(category.to_string()))?;
            facts.add_value("classification_confidence", Value::Number(confidence))?;
            Ok(())
        });

        // Sentiment Analysis action
        engine.register_action_handler("SentimentAnalysis", |params, facts| {
            let text = params.get("0").map(|v| v.to_string()).unwrap_or_default();

            println!("😊 SENTIMENT ANALYSIS:");
            println!("   Text: \"{}\"", text);

            // Simulate sentiment analysis
            let (sentiment, score) = if text.to_lowercase().contains("love")
                || text.contains("great")
                || text.contains("excellent")
            {
                ("positive", 0.89)
            } else if text.to_lowercase().contains("hate")
                || text.contains("terrible")
                || text.contains("awful")
            {
                ("negative", -0.76)
            } else if text.to_lowercase().contains("okay") || text.contains("fine") {
                ("neutral", 0.12)
            } else {
                ("neutral", 0.05)
            };

            println!("   🎭 Sentiment: {} (score: {:.2})", sentiment, score);

            facts.add_value("sentiment", Value::String(sentiment.to_string()))?;
            facts.add_value("sentiment_score", Value::Number(score))?;
            Ok(())
        });

        // Generate Embedding action
        engine.register_action_handler("GenerateEmbedding", |params, facts| {
            let text = params.get("0").map(|v| v.to_string()).unwrap_or_default();
            let model = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("text-embedding-ada-002".to_string());

            println!("🧠 GENERATE EMBEDDING:");
            println!("   Text: \"{}\"", text);
            println!("   Model: {}", model);

            // Simulate embedding generation
            let embedding_preview = "[0.123, -0.456, 0.789, 0.234, -0.567, ...]";
            println!("   🔢 Embedding (1536 dimensions): {}", embedding_preview);

            facts.add_value("embedding", Value::String(embedding_preview.to_string()))?;
            facts.add_value("embedding_dimensions", Value::Number(1536.0))?;
            Ok(())
        });

        // Detect Anomalies action
        engine.register_action_handler("DetectAnomalies", |params, facts| {
            let data = params
                .get("0")
                .map(|v| v.to_string())
                .unwrap_or("[]".to_string());
            let threshold = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("0.95".to_string());

            println!("🚨 ANOMALY DETECTION:");
            println!("   Data points: {}", data);
            println!("   Threshold: {}", threshold);

            // Simulate anomaly detection
            let anomalies_found = 3;
            let anomaly_score = 0.97;

            println!("   ⚠️  Anomalies found: {} points", anomalies_found);
            println!("   📊 Max anomaly score: {:.2}", anomaly_score);

            facts.add_value("anomalies_count", Value::Number(anomalies_found as f64))?;
            facts.add_value("max_anomaly_score", Value::Number(anomaly_score))?;
            facts.add_value("has_anomalies", Value::Boolean(anomalies_found > 0))?;
            Ok(())
        });

        // Recommend Items action
        engine.register_action_handler("RecommendItems", |params, facts| {
            let user_id = params
                .get("0")
                .map(|v| v.to_string())
                .unwrap_or("user123".to_string());
            let item_type = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("products".to_string());
            let count = params
                .get("2")
                .map(|v| v.to_string())
                .unwrap_or("5".to_string());

            println!("🎯 RECOMMENDATION ENGINE:");
            println!("   User ID: {}", user_id);
            println!("   Item Type: {}", item_type);
            println!("   Count: {}", count);

            // Simulate recommendations based on item type
            let recommendations = match item_type.as_str() {
                "products" => "[{\"id\": 101, \"name\": \"Wireless Headphones\", \"score\": 0.95}]",
                "movies" => "[{\"id\": 201, \"title\": \"Sci-Fi Adventure\", \"score\": 0.88}]",
                "books" => "[{\"id\": 301, \"title\": \"AI Programming Guide\", \"score\": 0.92}]",
                _ => "[{\"id\": 1, \"name\": \"Recommended Item\", \"score\": 0.85}]",
            };

            println!("   ✨ Recommendations generated:");
            println!("      {}", recommendations);

            facts.add_value(
                "recommendations",
                Value::String(recommendations.to_string()),
            )?;
            facts.add_value("recommendation_count", Value::Number(5.0))?;
            Ok(())
        });

        // Chat Completion action
        engine.register_action_handler("ChatCompletion", |params, facts| {
            let prompt = params.get("0").map(|v| v.to_string()).unwrap_or_default();
            let model = params
                .get("1")
                .map(|v| v.to_string())
                .unwrap_or("gpt-3.5-turbo".to_string());

            println!("💬 CHAT COMPLETION:");
            println!("   Prompt: \"{}\"", prompt);
            println!("   Model: {}", model);

            // Simulate AI response based on prompt content
            let response = if prompt.to_lowercase().contains("code") {
                "Here's a code example: `fn hello() { println!(\"Hello!\"); }`"
            } else if prompt.to_lowercase().contains("explain") {
                "Let me explain that concept in simple terms..."
            } else if prompt.to_lowercase().contains("help") {
                "I'd be happy to help you with that!"
            } else {
                "Thank you for your question. Here's my response..."
            };

            println!("   🤖 AI Response: \"{}\"", response);

            facts.add_value("ai_response", Value::String(response.to_string()))?;
            facts.add_value("tokens_used", Value::Number(127.0))?;
            Ok(())
        });

        Ok(())
    }

    fn register_functions(&self, engine: &mut RustRuleEngine) -> Result<()> {
        // Calculate Similarity function
        engine.register_function("calculateSimilarity", |args, _facts| {
            let text1 = args.get(0).map(|v| v.to_string()).unwrap_or_default();
            let text2 = args.get(1).map(|v| v.to_string()).unwrap_or_default();

            // Simulate cosine similarity calculation
            let similarity = if text1.is_empty() || text2.is_empty() {
                0.0
            } else if text1 == text2 {
                1.0
            } else {
                // Simple word overlap similarity
                let words1: std::collections::HashSet<_> = text1.split_whitespace().collect();
                let words2: std::collections::HashSet<_> = text2.split_whitespace().collect();
                let intersection = words1.intersection(&words2).count();
                let union = words1.union(&words2).count();
                intersection as f64 / union as f64
            };

            println!("🔍 Similarity calculated: {:.3}", similarity);

            Ok(Value::Number(similarity))
        });

        // Extract Keywords function
        engine.register_function("extractKeywords", |args, _facts| {
            let text = args.get(0).map(|v| v.to_string()).unwrap_or_default();
            let max_keywords = args.get(1).map(|v| v.to_string()).unwrap_or("5".to_string());
            
            // Simulate keyword extraction
            let keywords = if text.to_lowercase().contains("ai") || text.contains("machine learning") {
                "artificial intelligence, machine learning, neural networks, deep learning, automation"
            } else if text.to_lowercase().contains("database") || text.contains("sql") {
                "database, sql, query, data, storage"
            } else if text.to_lowercase().contains("web") || text.contains("http") {
                "web, http, api, rest, server"
            } else {
                "general, content, text, analysis, processing"
            };
            
            println!("🏷️  Keywords extracted ({}): {}", max_keywords, keywords);
            
            Ok(Value::String(keywords.to_string()))
        });

        // Tokenize Text function
        engine.register_function("tokenizeText", |args, _facts| {
            let text = args.get(0).map(|v| v.to_string()).unwrap_or_default();

            // Simulate tokenization
            let word_count = text.split_whitespace().count();
            let tokens = format!(
                "[\"{}\"]",
                text.split_whitespace().collect::<Vec<_>>().join("\", \"")
            );

            println!(
                "🔤 Tokenized {} words: {}",
                word_count,
                if tokens.len() > 100 {
                    format!("{}...", &tokens[..100])
                } else {
                    tokens.clone()
                }
            );

            Ok(Value::String(tokens))
        });

        // Normalize Score function
        engine.register_function("normalizeScore", |args, _facts| {
            let score = args
                .get(0)
                .and_then(|v| v.to_string().parse::<f64>().ok())
                .unwrap_or(0.0);
            let min_val = args
                .get(1)
                .and_then(|v| v.to_string().parse::<f64>().ok())
                .unwrap_or(0.0);
            let max_val = args
                .get(2)
                .and_then(|v| v.to_string().parse::<f64>().ok())
                .unwrap_or(1.0);

            // Normalize score to 0-1 range
            let normalized = if max_val == min_val {
                0.5
            } else {
                (score - min_val) / (max_val - min_val)
            }
            .max(0.0)
            .min(1.0);

            println!("📊 Score normalized: {} -> {:.3}", score, normalized);

            Ok(Value::Number(normalized))
        });

        // Evaluate Model function
        engine.register_function("evaluateModel", |args, _facts| {
            let model_name = args
                .get(0)
                .map(|v| v.to_string())
                .unwrap_or("model".to_string());
            let metric = args
                .get(1)
                .map(|v| v.to_string())
                .unwrap_or("accuracy".to_string());

            // Simulate model evaluation
            let score = match metric.as_str() {
                "accuracy" => 0.94,
                "precision" => 0.91,
                "recall" => 0.88,
                "f1_score" => 0.89,
                "auc" => 0.96,
                _ => 0.85,
            };

            println!("📈 Model {} {} score: {:.3}", model_name, metric, score);

            Ok(Value::Number(score))
        });

        Ok(())
    }

    fn unload(&mut self) -> Result<()> {
        self.metadata.state = PluginState::Unloaded;
        println!("🤖 AI/ML Plugin unloaded");
        Ok(())
    }

    fn health_check(&mut self) -> PluginHealth {
        println!("🏥 AI/ML health check: Models loaded and ready");
        PluginHealth::Healthy
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use rust_rule_engine::engine::knowledge_base::KnowledgeBase;
    use rust_rule_engine::Facts;

    #[test]
    fn test_aiml_plugin() {
        let kb = KnowledgeBase::new("AIMLTest");
        let mut engine = RustRuleEngine::new(kb);
        let facts = Facts::new();

        let plugin = AIMLPlugin::new();

        // Test plugin registration
        assert!(plugin.register_actions(&mut engine).is_ok());
        assert!(plugin.register_functions(&mut engine).is_ok());

        // Test function availability
        assert!(engine.has_function("calculateSimilarity"));
        assert!(engine.has_function("extractKeywords"));
        assert!(engine.has_function("normalizeScore"));

        // Test action availability
        assert!(engine.has_action_handler("PredictValue"));
        assert!(engine.has_action_handler("ClassifyText"));
        assert!(engine.has_action_handler("SentimentAnalysis"));
    }
}