foxhunt/MAIN_RS_WIRING_INSTRUCTIONS.md

Main.rs Wiring Instructions for Tuning Progress Streaming

Required Changes to /services/ml_training_service/src/main.rs

1. Import TuningManager

Add to imports section (around line 19):

use ml_training_service::{database, encryption, gpu_config, orchestrator, service, storage, tuning_manager};

2. Initialize TuningManager

Add after orchestrator initialization (around line 318):

info!("Training orchestrator started");

// Initialize tuning manager
let tuning_script_path = std::env::var("TUNING_SCRIPT_PATH")
    .unwrap_or_else(|_| "/opt/foxhunt/scripts/optuna_tuner.py".to_string());
let tuning_working_dir = std::env::var("TUNING_WORKING_DIR")
    .unwrap_or_else(|_| "/var/lib/foxhunt/tuning".to_string());

let tuning_manager = Arc::new(tuning_manager::TuningManager::new(
    tuning_script_path,
    tuning_working_dir,
));

info!("Tuning manager initialized");

3. Update Service Creation

Change service creation (around line 327):

// Create gRPC service WITH TuningManager
let training_service = MLTrainingServiceImpl::new(
    Arc::clone(&orchestrator),
    Arc::clone(&tuning_manager),  // <-- ADD THIS
    ml_config.clone()
);

4. Environment Variables

Add to .env (if not exists):

# Hyperparameter Tuning Configuration
TUNING_SCRIPT_PATH=/opt/foxhunt/scripts/optuna_tuner.py
TUNING_WORKING_DIR=/var/lib/foxhunt/tuning

---

Complete Code Snippet

// After line 318 (after orchestrator started)
info!("Training orchestrator started");

// Initialize tuning manager
let tuning_script_path = std::env::var("TUNING_SCRIPT_PATH")
    .unwrap_or_else(|_| {
        warn!("TUNING_SCRIPT_PATH not set, using default");
        "/opt/foxhunt/scripts/optuna_tuner.py".to_string()
    });

let tuning_working_dir = std::env::var("TUNING_WORKING_DIR")
    .unwrap_or_else(|_| {
        warn!("TUNING_WORKING_DIR not set, using default");
        "/var/lib/foxhunt/tuning".to_string()
    });

info!(
    "Tuning manager configuration: script={}, working_dir={}",
    tuning_script_path, tuning_working_dir
);

let tuning_manager = Arc::new(tuning_manager::TuningManager::new(
    tuning_script_path,
    tuning_working_dir,
));

info!("Tuning manager initialized with broadcast channel (capacity: 100)");

// Initialize TLS configuration for mTLS (line 321)
let tls_config = MLTrainingServiceTlsConfig::from_config(&config_manager).await
    .context("Failed to initialize TLS configuration")?;

info!("TLS configuration initialized with mutual TLS");

// Create gRPC service (line 327)
let training_service = MLTrainingServiceImpl::new(
    Arc::clone(&orchestrator),
    Arc::clone(&tuning_manager),  // <-- NEW PARAMETER
    ml_config.clone()
);

---

Verification Steps

After applying changes:

1. Compile check:

   cargo build -p ml_training_service
   

2. Run service:

   cargo run -p ml_training_service
   

3. Verify logs:

   INFO ml_training_service: Training orchestrator started
   INFO ml_training_service: Tuning manager configuration: script=/opt/foxhunt/scripts/optuna_tuner.py, working_dir=/var/lib/foxhunt/tuning
   INFO ml_training_service: Tuning manager initialized with broadcast channel (capacity: 100)
   INFO ml_training_service: TLS configuration initialized with mutual TLS
   

4. Test streaming:

   tli tune start --model DQN --trials 10 --config tuning_config.yaml --watch
   

---

Troubleshooting

Error: "cannot find tuning_manager in the crate root"

Solution: Add to src/lib.rs:

pub mod tuning_manager;

Error: "no method named new found for struct MLTrainingServiceImpl"

Cause: Service constructor signature changed (added tuning_manager parameter)

Solution: Update all service instantiations to pass Arc<TuningManager>

Error: "mismatched types: expected 3 arguments, found 2"

Cause: Old service constructor called without tuning_manager

Solution: Find all MLTrainingServiceImpl::new() calls and add the new parameter:

// BEFORE
MLTrainingServiceImpl::new(orchestrator, config)

// AFTER
MLTrainingServiceImpl::new(orchestrator, tuning_manager, config)

---

Testing Checklist

• Service compiles successfully
• Service starts without errors
• Logs show "Tuning manager initialized"
• gRPC health check passes
• tli tune start creates job
• tli tune start --watch streams progress
• Progress updates appear in real-time
• Stream closes on job completion
• Multiple clients can subscribe simultaneously

---

Files to Modify

1. ✅ services/ml_training_service/src/main.rs (wiring)
2. ✅ services/ml_training_service/src/lib.rs (export tuning_manager module)
3. ✅ .env (environment variables)

---

Status: Ready for implementation Estimated Time: 10-15 minutes Risk: Low (backward compatible, additive changes only)