foxhunt/ml/tests/model_registry_checkpoint_test.rs

//! Model Registry Checkpoint Integration Tests
//!
//! TDD tests for checkpoint versioning, metadata tracking, and production model registration.
//! Wave 10 Agent 10.8 - Training → Paper Trading Integration

use chrono::Utc;
use ml::model_registry::{ModelRegistry, ModelVersionMetadata};
use ml::{MLResult, ModelType};
use std::path::PathBuf;

// Test database URL
const TEST_DB_URL: &str = "postgresql://foxhunt:foxhunt_dev_password@localhost:5432/foxhunt";
const TEST_S3_PATH: &str = "s3://foxhunt-ml-models-test/";

/// Test 1: Register trained DQN model with checkpoint path
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_register_dqn_checkpoint() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    // Find latest DQN checkpoint
    let checkpoint_path = PathBuf::from(
        "/home/jgrusewski/Work/foxhunt/ml/trained_models/production/dqn/dqn_epoch_30.safetensors",
    );

    let mut metadata = ModelVersionMetadata::new(
        "dqn-production-v1.0.0".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://foxhunt-ml-models/dqn/1.0.0/".to_string(),
    );

    // Add checkpoint metadata
    metadata.add_hyperparameter("epochs", serde_json::json!(30));
    metadata.add_hyperparameter("batch_size", serde_json::json!(128));
    metadata.add_hyperparameter("learning_rate", serde_json::json!(0.0001));

    metadata.add_metric("final_loss", serde_json::json!(0.0342));
    metadata.add_metric("validation_loss", serde_json::json!(0.0356));

    metadata.add_metadata(
        "checkpoint_path",
        checkpoint_path.to_string_lossy().to_string(),
    );
    metadata.add_metadata("training_duration_hours", "2.5".to_string());

    metadata.set_checksum("sha256:dqn_epoch_30_checksum".to_string());

    // Register
    registry.register_version(&metadata).await?;

    // Verify retrieval
    let retrieved = registry
        .get_model_by_version("dqn-production-v1.0.0")
        .await?;
    assert_eq!(retrieved.model_id, "dqn-production-v1.0.0");
    assert_eq!(retrieved.model_type, ModelType::DQN);
    assert_eq!(retrieved.version, "1.0.0");
    assert!(retrieved.metadata.contains_key("checkpoint_path"));

    Ok(())
}

/// Test 2: Register trained PPO model with actor-critic checkpoints
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_register_ppo_checkpoint() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let actor_checkpoint = PathBuf::from("/home/jgrusewski/Work/foxhunt/ml/trained_models/production/ppo/ppo_actor_epoch_420.safetensors");
    let critic_checkpoint = PathBuf::from("/home/jgrusewski/Work/foxhunt/ml/trained_models/production/ppo/ppo_critic_epoch_420.safetensors");

    let mut metadata = ModelVersionMetadata::new(
        "ppo-production-v1.0.0".to_string(),
        ModelType::PPO,
        "1.0.0".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://foxhunt-ml-models/ppo/1.0.0/".to_string(),
    );

    // Add PPO-specific hyperparameters
    metadata.add_hyperparameter("epochs", serde_json::json!(420));
    metadata.add_hyperparameter("batch_size", serde_json::json!(64));
    metadata.add_hyperparameter("learning_rate", serde_json::json!(0.0003));
    metadata.add_hyperparameter("gamma", serde_json::json!(0.99));
    metadata.add_hyperparameter("gae_lambda", serde_json::json!(0.95));

    metadata.add_metric("final_actor_loss", serde_json::json!(0.0152));
    metadata.add_metric("final_critic_loss", serde_json::json!(0.0089));
    metadata.add_metric("avg_reward", serde_json::json!(45.3));

    metadata.add_metadata(
        "actor_checkpoint_path",
        actor_checkpoint.to_string_lossy().to_string(),
    );
    metadata.add_metadata(
        "critic_checkpoint_path",
        critic_checkpoint.to_string_lossy().to_string(),
    );

    metadata.set_checksum("sha256:ppo_epoch_420_checksum".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("ppo-production-v1.0.0")
        .await?;
    assert_eq!(retrieved.model_type, ModelType::PPO);
    assert!(retrieved.metadata.contains_key("actor_checkpoint_path"));
    assert!(retrieved.metadata.contains_key("critic_checkpoint_path"));

    Ok(())
}

/// Test 3: Register trained MAMBA-2 model with training metrics
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_register_mamba2_checkpoint() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let mut metadata = ModelVersionMetadata::new(
        "mamba2-production-v1.0.0".to_string(),
        ModelType::MAMBA,
        "1.0.0".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://foxhunt-ml-models/mamba2/1.0.0/".to_string(),
    );

    // Add MAMBA-2 hyperparameters
    metadata.add_hyperparameter("epochs", serde_json::json!(24));
    metadata.add_hyperparameter("batch_size", serde_json::json!(32));
    metadata.add_hyperparameter("learning_rate", serde_json::json!(0.0001));
    metadata.add_hyperparameter("d_model", serde_json::json!(256));
    metadata.add_hyperparameter("n_layers", serde_json::json!(6));
    metadata.add_hyperparameter("state_size", serde_json::json!(16));

    metadata.add_metric("best_val_loss", serde_json::json!(1.4318895660848898));
    metadata.add_metric("best_epoch", serde_json::json!(3));
    metadata.add_metric("final_perplexity", serde_json::json!(4.1866025848353));

    metadata.add_metadata(
        "checkpoint_path",
        "/home/jgrusewski/Work/foxhunt/ml/checkpoints/mamba2_dbn/".to_string(),
    );
    metadata.add_metadata("training_duration_hours", "0.031".to_string());

    metadata.set_checksum("sha256:mamba2_epoch_24_checksum".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("mamba2-production-v1.0.0")
        .await?;
    assert_eq!(retrieved.model_type, ModelType::MAMBA);

    // Verify metrics
    let metrics = retrieved.metrics.as_object().unwrap();
    assert!(metrics.contains_key("best_val_loss"));
    assert!(metrics.contains_key("final_perplexity"));

    Ok(())
}

/// Test 4: Register trained TFT model with multiple checkpoints
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_register_tft_checkpoint() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let checkpoint_path = PathBuf::from(
        "/home/jgrusewski/Work/foxhunt/ml/trained_models/production/tft/tft_epoch_100.safetensors",
    );

    let mut metadata = ModelVersionMetadata::new(
        "tft-production-v1.0.0".to_string(),
        ModelType::TFT,
        "1.0.0".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://foxhunt-ml-models/tft/1.0.0/".to_string(),
    );

    // Add TFT hyperparameters
    metadata.add_hyperparameter("epochs", serde_json::json!(100));
    metadata.add_hyperparameter("batch_size", serde_json::json!(256));
    metadata.add_hyperparameter("learning_rate", serde_json::json!(0.0001));
    metadata.add_hyperparameter("hidden_size", serde_json::json!(256));
    metadata.add_hyperparameter("num_attention_heads", serde_json::json!(8));

    metadata.add_metric("final_loss", serde_json::json!(0.0198));
    metadata.add_metric("validation_loss", serde_json::json!(0.0213));
    metadata.add_metric("sharpe_ratio", serde_json::json!(2.4));

    metadata.add_metadata(
        "checkpoint_path",
        checkpoint_path.to_string_lossy().to_string(),
    );

    metadata.set_checksum("sha256:tft_epoch_100_checksum".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("tft-production-v1.0.0")
        .await?;
    assert_eq!(retrieved.model_type, ModelType::TFT);

    // Verify hyperparameters
    let hyperparams = retrieved.hyperparameters.as_object().unwrap();
    assert_eq!(hyperparams.get("epochs").unwrap(), &serde_json::json!(100));

    Ok(())
}

/// Test 5: Register TFT-INT8 quantized model
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_register_tft_int8_checkpoint() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let mut metadata = ModelVersionMetadata::new(
        "tft-int8-production-v1.0.0".to_string(),
        ModelType::TFT,
        "1.0.0-int8".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://foxhunt-ml-models/tft-int8/1.0.0/".to_string(),
    );

    metadata.add_hyperparameter("quantization", serde_json::json!("int8"));
    metadata.add_hyperparameter("epochs", serde_json::json!(100));

    metadata.add_metric("inference_latency_ms", serde_json::json!(3.2));
    metadata.add_metric("model_size_mb", serde_json::json!(128));

    metadata.add_metadata("quantization_method", "static_int8".to_string());
    metadata.add_metadata("optimization_level", "production".to_string());

    metadata.set_checksum("sha256:tft_int8_checksum".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("tft-int8-production-v1.0.0")
        .await?;
    assert_eq!(retrieved.version, "1.0.0-int8");
    assert!(retrieved.metadata.contains_key("quantization_method"));

    Ok(())
}

/// Test 6: Version increment handling
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_version_increment() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    // Register v1.0.0
    let mut metadata_v1 = ModelVersionMetadata::new(
        "dqn-version-test-v1.0.0".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/1.0.0/".to_string(),
    );
    metadata_v1.add_metric("loss", serde_json::json!(0.05));
    registry.register_version(&metadata_v1).await?;

    // Register v1.1.0 (improvement)
    let mut metadata_v1_1 = ModelVersionMetadata::new(
        "dqn-version-test-v1.1.0".to_string(),
        ModelType::DQN,
        "1.1.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/1.1.0/".to_string(),
    );
    metadata_v1_1.add_metric("loss", serde_json::json!(0.03));
    registry.register_version(&metadata_v1_1).await?;

    // Register v2.0.0 (major update)
    let mut metadata_v2 = ModelVersionMetadata::new(
        "dqn-version-test-v2.0.0".to_string(),
        ModelType::DQN,
        "2.0.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/2.0.0/".to_string(),
    );
    metadata_v2.add_metric("loss", serde_json::json!(0.01));
    registry.register_version(&metadata_v2).await?;

    // Verify all versions exist
    let v1 = registry
        .get_model_by_version("dqn-version-test-v1.0.0")
        .await?;
    assert_eq!(v1.version, "1.0.0");

    let v1_1 = registry
        .get_model_by_version("dqn-version-test-v1.1.0")
        .await?;
    assert_eq!(v1_1.version, "1.1.0");

    let v2 = registry
        .get_model_by_version("dqn-version-test-v2.0.0")
        .await?;
    assert_eq!(v2.version, "2.0.0");

    Ok(())
}

/// Test 7: Checkpoint path validation
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_checkpoint_path_metadata() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let checkpoint_path = PathBuf::from(
        "/home/jgrusewski/Work/foxhunt/ml/trained_models/production/dqn/dqn_epoch_30.safetensors",
    );

    let mut metadata = ModelVersionMetadata::new(
        "dqn-checkpoint-path-test".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/1.0.0/".to_string(),
    );

    metadata.add_metadata(
        "checkpoint_path",
        checkpoint_path.to_string_lossy().to_string(),
    );
    metadata.add_metadata("checkpoint_format", "safetensors".to_string());
    metadata.add_metadata("checkpoint_size_mb", "256".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("dqn-checkpoint-path-test")
        .await?;

    assert!(retrieved.metadata.contains_key("checkpoint_path"));
    assert_eq!(
        retrieved.metadata.get("checkpoint_format").unwrap(),
        "safetensors"
    );

    Ok(())
}

/// Test 8: Multi-model registry query
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_multi_model_registry_query() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    // Register multiple models
    let model_types = vec![
        (ModelType::DQN, "dqn-multi-test"),
        (ModelType::PPO, "ppo-multi-test"),
        (ModelType::MAMBA, "mamba-multi-test"),
        (ModelType::TFT, "tft-multi-test"),
    ];

    for (model_type, model_id) in model_types {
        let metadata = ModelVersionMetadata::new(
            model_id.to_string(),
            model_type,
            "1.0.0".to_string(),
            "test_data".to_string(),
            format!("s3://test/{}/1.0.0/", model_id),
        );
        registry.register_version(&metadata).await?;
    }

    // Query by type
    let dqn_models = registry.get_models_by_type(ModelType::DQN).await?;
    assert!(dqn_models.iter().any(|m| m.model_id == "dqn-multi-test"));

    let ppo_models = registry.get_models_by_type(ModelType::PPO).await?;
    assert!(ppo_models.iter().any(|m| m.model_id == "ppo-multi-test"));

    Ok(())
}

/// Test 9: Production model promotion workflow
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_production_promotion_workflow() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let metadata = ModelVersionMetadata::new(
        "dqn-promotion-test".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/1.0.0/".to_string(),
    );

    // Start as experimental
    assert!(metadata.is_experimental);
    assert!(!metadata.is_production);

    registry.register_version(&metadata).await?;

    // Promote to production
    registry.mark_production("dqn-promotion-test").await?;

    // Verify production status
    let retrieved = registry.get_model_by_version("dqn-promotion-test").await?;
    assert!(retrieved.is_production);
    assert!(!retrieved.is_experimental);

    // Verify in production query
    let production_models = registry.get_production_models().await?;
    assert!(production_models
        .iter()
        .any(|m| m.model_id == "dqn-promotion-test"));

    Ok(())
}

/// Test 10: Training metrics metadata
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_training_metrics_metadata() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let mut metadata = ModelVersionMetadata::new(
        "dqn-metrics-test".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "test_data".to_string(),
        "s3://test/dqn/1.0.0/".to_string(),
    );

    // Add comprehensive metrics
    metadata.add_metric("final_loss", serde_json::json!(0.0342));
    metadata.add_metric("validation_loss", serde_json::json!(0.0356));
    metadata.add_metric("best_epoch", serde_json::json!(28));
    metadata.add_metric("total_epochs", serde_json::json!(30));
    metadata.add_metric("training_duration_hours", serde_json::json!(2.5));
    metadata.add_metric("gpu_memory_used_gb", serde_json::json!(3.2));
    metadata.add_metric("avg_epoch_time_seconds", serde_json::json!(300));

    registry.register_version(&metadata).await?;

    let retrieved = registry.get_model_by_version("dqn-metrics-test").await?;

    let metrics = retrieved.metrics.as_object().unwrap();
    assert_eq!(
        metrics.get("final_loss").unwrap(),
        &serde_json::json!(0.0342)
    );
    assert_eq!(metrics.get("best_epoch").unwrap(), &serde_json::json!(28));
    assert!(metrics.contains_key("gpu_memory_used_gb"));

    Ok(())
}

/// Test 11: List all checkpoints for a model type
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_list_checkpoints_by_type() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    // Register multiple DQN checkpoints
    for epoch in [10, 20, 30] {
        let mut metadata = ModelVersionMetadata::new(
            format!("dqn-checkpoint-epoch-{}", epoch),
            ModelType::DQN,
            format!("1.0.{}", epoch),
            "test_data".to_string(),
            format!("s3://test/dqn/1.0.{}/", epoch),
        );
        metadata.add_metadata("epoch", epoch.to_string());
        registry.register_version(&metadata).await?;
    }

    let dqn_models = registry.get_models_by_type(ModelType::DQN).await?;
    let checkpoint_models: Vec<_> = dqn_models
        .iter()
        .filter(|m| m.model_id.starts_with("dqn-checkpoint-epoch-"))
        .collect();

    assert!(checkpoint_models.len() >= 3);

    Ok(())
}

/// Test 12: Checkpoint metadata completeness
#[tokio::test]
#[ignore] // Requires PostgreSQL
async fn test_checkpoint_metadata_completeness() -> MLResult<()> {
    let registry = ModelRegistry::new(TEST_DB_URL, TEST_S3_PATH).await?;

    let mut metadata = ModelVersionMetadata::new(
        "complete-metadata-test".to_string(),
        ModelType::DQN,
        "1.0.0".to_string(),
        "ES.FUT_2024_Q4".to_string(),
        "s3://test/dqn/1.0.0/".to_string(),
    );

    // Add comprehensive metadata
    metadata.add_hyperparameter("epochs", serde_json::json!(30));
    metadata.add_hyperparameter("batch_size", serde_json::json!(128));
    metadata.add_hyperparameter("learning_rate", serde_json::json!(0.0001));
    metadata.add_hyperparameter("gamma", serde_json::json!(0.99));
    metadata.add_hyperparameter("epsilon_start", serde_json::json!(1.0));
    metadata.add_hyperparameter("epsilon_end", serde_json::json!(0.01));

    metadata.add_metric("final_loss", serde_json::json!(0.0342));
    metadata.add_metric("validation_loss", serde_json::json!(0.0356));
    metadata.add_metric("sharpe_ratio", serde_json::json!(2.1));
    metadata.add_metric("max_drawdown", serde_json::json!(0.12));

    metadata.add_metadata(
        "checkpoint_path",
        "/path/to/checkpoint.safetensors".to_string(),
    );
    metadata.add_metadata("training_date", Utc::now().to_rfc3339());
    metadata.add_metadata("cuda_version", "12.1".to_string());
    metadata.add_metadata("pytorch_version", "2.0.0".to_string());

    metadata.set_checksum("sha256:complete_metadata_checksum".to_string());

    registry.register_version(&metadata).await?;

    let retrieved = registry
        .get_model_by_version("complete-metadata-test")
        .await?;

    // Verify hyperparameters
    let hyperparams = retrieved.hyperparameters.as_object().unwrap();
    assert_eq!(hyperparams.len(), 6);

    // Verify metrics
    let metrics = retrieved.metrics.as_object().unwrap();
    assert_eq!(metrics.len(), 4);

    // Verify metadata
    assert_eq!(retrieved.metadata.len(), 4);
    assert!(retrieved.metadata.contains_key("checkpoint_path"));
    assert!(retrieved.metadata.contains_key("cuda_version"));

    Ok(())
}