//! 进化引擎中枢
//! 协调 L1/L2/L3 三层进化的触发和执行
//! L1 (记忆进化) 在 GrowthIntegration 中处理
//! L2 (技能进化) 通过 PatternAggregator + SkillGenerator + QualityGate 协调
//! L3 (工作流进化) 通过 WorkflowComposer 协调
//! 反馈闭环通过 FeedbackCollector 管理

use std::sync::Arc;

use crate::experience_store::ExperienceStore;
use crate::feedback_collector::{
    FeedbackCollector, FeedbackEntry, TrustUpdate,
};
use crate::pattern_aggregator::{AggregatedPattern, PatternAggregator};
use crate::quality_gate::{QualityGate, QualityReport};
use crate::skill_generator::{SkillCandidate, SkillGenerator};
use crate::workflow_composer::{ToolChainPattern, WorkflowComposer};
use crate::VikingAdapter;
use zclaw_types::Result;

/// 进化引擎配置
#[derive(Debug, Clone)]
pub struct EvolutionConfig {
    /// 经验复用次数达到此阈值触发 L2
    pub min_reuse_for_skill: u32,
    /// 置信度阈值
    pub quality_confidence_threshold: f32,
    /// 是否启用进化引擎
    pub enabled: bool,
}

impl Default for EvolutionConfig {
    fn default() -> Self {
        Self {
            min_reuse_for_skill: 3,
            quality_confidence_threshold: 0.7,
            enabled: true,
        }
    }
}

/// 进化引擎中枢
pub struct EvolutionEngine {
    viking: Arc<VikingAdapter>,
    feedback: FeedbackCollector,
    config: EvolutionConfig,
}

impl EvolutionEngine {
    pub fn new(viking: Arc<VikingAdapter>) -> Self {
        Self {
            viking: viking.clone(),
            feedback: FeedbackCollector::with_viking(viking),
            config: EvolutionConfig::default(),
        }
    }

    /// Backward-compatible constructor
    /// 从 ExperienceStore 中提取共享的 VikingAdapter 实例
    pub fn from_experience_store(experience_store: Arc<ExperienceStore>) -> Self {
        let viking = experience_store.viking().clone();
        Self {
            viking: viking.clone(),
            feedback: FeedbackCollector::with_viking(viking),
            config: EvolutionConfig::default(),
        }
    }

    pub fn with_config(mut self, config: EvolutionConfig) -> Self {
        self.config = config;
        self
    }

    pub fn set_enabled(&mut self, enabled: bool) {
        self.config.enabled = enabled;
    }

    /// L2 检查：是否有可进化的模式
    pub async fn check_evolvable_patterns(
        &self,
        agent_id: &str,
    ) -> Result<Vec<AggregatedPattern>> {
        if !self.config.enabled {
            return Ok(Vec::new());
        }
        let store = ExperienceStore::new(self.viking.clone());
        let aggregator = PatternAggregator::new(store);
        aggregator
            .find_evolvable_patterns(agent_id, self.config.min_reuse_for_skill)
            .await
    }

    /// L2 执行：为给定模式构建技能生成 prompt
    /// 返回 (prompt_string, pattern) 供上层通过 LLM 调用后 parse
    pub fn build_skill_prompt(&self, pattern: &AggregatedPattern) -> String {
        SkillGenerator::build_prompt(pattern)
    }

    /// L2 执行：解析 LLM 返回的技能 JSON 并进行质量门控
    pub fn validate_skill_candidate(
        &self,
        json_str: &str,
        pattern: &AggregatedPattern,
        existing_triggers: Vec<String>,
    ) -> Result<(SkillCandidate, QualityReport)> {
        let candidate = SkillGenerator::parse_response(json_str, pattern)?;
        let gate = QualityGate::new(self.config.quality_confidence_threshold, existing_triggers);
        let report = gate.validate_skill(&candidate);
        Ok((candidate, report))
    }

    /// 获取当前配置
    pub fn config(&self) -> &EvolutionConfig {
        &self.config
    }

    // -----------------------------------------------------------------------
    // L3: 工作流进化
    // -----------------------------------------------------------------------

    /// L3: 从轨迹数据中提取重复的工具链模式
    pub fn analyze_trajectory_patterns(
        &self,
        trajectories: &[(String, Vec<String>)], // (session_id, tools_used)
    ) -> Vec<(ToolChainPattern, Vec<String>)> {
        if !self.config.enabled {
            return Vec::new();
        }
        WorkflowComposer::extract_patterns(trajectories)
    }

    /// L3: 为给定工具链模式构建工作流生成 prompt
    pub fn build_workflow_prompt(
        &self,
        pattern: &ToolChainPattern,
        frequency: usize,
        industry: Option<&str>,
    ) -> String {
        WorkflowComposer::build_prompt(pattern, frequency, industry)
    }

    // -----------------------------------------------------------------------
    // 反馈闭环
    // -----------------------------------------------------------------------

    /// 提交反馈并获取信任度更新，自动持久化
    pub async fn submit_feedback(&mut self, entry: FeedbackEntry) -> TrustUpdate {
        let update = self.feedback.submit_feedback(entry);
        // 非阻塞持久化：失败仅打日志，不影响返回值
        if let Err(e) = self.feedback.save().await {
            tracing::warn!("[EvolutionEngine] Failed to persist trust records: {}", e);
        }
        update
    }

    /// 获取需要优化的进化产物
    pub fn get_artifacts_needing_optimization(&self) -> Vec<String> {
        self.feedback
            .get_artifacts_needing_optimization()
            .iter()
            .map(|r| r.artifact_id.clone())
            .collect()
    }

    /// 获取建议归档的进化产物
    pub fn get_artifacts_to_archive(&self) -> Vec<String> {
        self.feedback
            .get_artifacts_to_archive()
            .iter()
            .map(|r| r.artifact_id.clone())
            .collect()
    }

    /// 获取推荐产物
    pub fn get_recommended_artifacts(&self) -> Vec<String> {
        self.feedback
            .get_recommended_artifacts()
            .iter()
            .map(|r| r.artifact_id.clone())
            .collect()
    }

    /// 获取反馈收集器的引用（用于高级查询）
    pub fn feedback(&self) -> &FeedbackCollector {
        &self.feedback
    }

    /// 启动时加载已持久化的信任度记录
    pub async fn load_feedback(&mut self) -> Result<usize> {
        self.feedback
            .load()
            .await
            .map_err(|e| zclaw_types::ZclawError::Internal(e))
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::experience_store::Experience;

    #[tokio::test]
    async fn test_disabled_returns_empty() {
        let viking = Arc::new(crate::VikingAdapter::in_memory());
        let mut engine = EvolutionEngine::new(viking);
        engine.set_enabled(false);

        let patterns = engine.check_evolvable_patterns("agent-1").await.unwrap();
        assert!(patterns.is_empty());
    }

    #[tokio::test]
    async fn test_no_evolvable_patterns() {
        let viking = Arc::new(crate::VikingAdapter::in_memory());
        let engine = EvolutionEngine::new(viking);

        let patterns = engine.check_evolvable_patterns("unknown-agent").await.unwrap();
        assert!(patterns.is_empty());
    }

    #[tokio::test]
    async fn test_finds_evolvable_pattern() {
        let viking = Arc::new(crate::VikingAdapter::in_memory());
        let store_inner = ExperienceStore::new(viking.clone());

        let mut exp = Experience::new(
            "agent-1",
            "report generation",
            "researcher",
            vec!["query db".into(), "format".into()],
            "success",
        );
        exp.reuse_count = 5;
        store_inner.store_experience(&exp).await.unwrap();

        let engine = EvolutionEngine::new(viking);

        let patterns = engine.check_evolvable_patterns("agent-1").await.unwrap();
        assert_eq!(patterns.len(), 1);
        assert_eq!(patterns[0].pain_pattern, "report generation");
    }

    #[test]
    fn test_build_skill_prompt() {
        let viking = Arc::new(crate::VikingAdapter::in_memory());
        let engine = EvolutionEngine::new(viking);

        let exp = Experience::new(
            "a", "report", "researcher", vec!["step1".into()], "ok",
        );
        let pattern = AggregatedPattern {
            pain_pattern: "report".to_string(),
            experiences: vec![exp],
            common_steps: vec!["step1".into()],
            total_reuse: 5,
            tools_used: vec!["researcher".into()],
            industry_context: None,
        };
        let prompt = engine.build_skill_prompt(&pattern);
        assert!(prompt.contains("report"));
    }

    #[test]
    fn test_validate_skill_candidate() {
        let viking = Arc::new(crate::VikingAdapter::in_memory());
        let engine = EvolutionEngine::new(viking);

        let exp = Experience::new(
            "a", "report", "researcher", vec!["step1".into()], "ok",
        );
        let pattern = AggregatedPattern {
            pain_pattern: "report".to_string(),
            experiences: vec![exp],
            common_steps: vec!["step1".into()],
            total_reuse: 5,
            tools_used: vec!["researcher".into()],
            industry_context: None,
        };

        let json = r##"{"name":"报表技能","description":"生成报表","triggers":["报表","日报"],"tools":["researcher"],"body_markdown":"# 报表\n步骤","confidence":0.9}"##;
        let (candidate, report) = engine
            .validate_skill_candidate(json, &pattern, vec!["搜索".to_string()])
            .unwrap();
        assert_eq!(candidate.name, "报表技能");
        assert!(report.passed);
    }
}