Atomizer/atomizer-dashboard/backend/api/routes/optimization.py

"""
Optimization API endpoints
Handles study status, history retrieval, and control operations
"""

from fastapi import APIRouter, HTTPException, UploadFile, File, Form
from fastapi.responses import JSONResponse, FileResponse
from pydantic import BaseModel
from pathlib import Path
from typing import List, Dict, Optional
import json
import sys
import sqlite3
import shutil
import subprocess
import psutil
import signal
from datetime import datetime

# Add project root to path
sys.path.append(str(Path(__file__).parent.parent.parent.parent.parent))

router = APIRouter()

# Base studies directory
STUDIES_DIR = Path(__file__).parent.parent.parent.parent.parent / "studies"


def get_results_dir(study_dir: Path) -> Path:
    """Get the results directory for a study, supporting both 2_results and 3_results."""
    results_dir = study_dir / "2_results"
    if not results_dir.exists():
        results_dir = study_dir / "3_results"
    return results_dir


def resolve_study_path(study_id: str) -> Path:
    """Find study folder by scanning all topic directories.

    Supports nested folder structure: studies/Topic/study_name/
    Study ID is the short name (e.g., 'm1_mirror_adaptive_V14')

    Returns the full path to the study directory.
    Raises HTTPException 404 if not found.
    """
    # First check direct path (backwards compatibility for flat structure)
    direct_path = STUDIES_DIR / study_id
    if direct_path.exists() and direct_path.is_dir():
        # Verify it's actually a study (has 1_setup or config)
        if (direct_path / "1_setup").exists() or (direct_path / "optimization_config.json").exists():
            return direct_path

    # Scan topic folders for nested structure
    for topic_dir in STUDIES_DIR.iterdir():
        if topic_dir.is_dir() and not topic_dir.name.startswith('.'):
            study_dir = topic_dir / study_id
            if study_dir.exists() and study_dir.is_dir():
                # Verify it's actually a study
                if (study_dir / "1_setup").exists() or (study_dir / "optimization_config.json").exists():
                    return study_dir

    raise HTTPException(status_code=404, detail=f"Study not found: {study_id}")


def get_study_topic(study_dir: Path) -> Optional[str]:
    """Get the topic folder name for a study, or None if in root."""
    # Check if parent is a topic folder (not the root studies dir)
    parent = study_dir.parent
    if parent != STUDIES_DIR and parent.parent == STUDIES_DIR:
        return parent.name
    return None


def is_optimization_running(study_id: str) -> bool:
    """Check if an optimization process is currently running for a study.

    Looks for Python processes running run_optimization.py with the study_id in the command line.
    """
    try:
        study_dir = resolve_study_path(study_id)
    except HTTPException:
        return False

    for proc in psutil.process_iter(['pid', 'name', 'cmdline', 'cwd']):
        try:
            cmdline = proc.info.get('cmdline') or []
            cmdline_str = ' '.join(cmdline) if cmdline else ''

            # Check if this is a Python process running run_optimization.py for this study
            if 'python' in cmdline_str.lower() and 'run_optimization' in cmdline_str:
                if study_id in cmdline_str or str(study_dir) in cmdline_str:
                    return True
        except (psutil.NoSuchProcess, psutil.AccessDenied):
            continue

    return False


def get_accurate_study_status(study_id: str, trial_count: int, total_trials: int, has_db: bool) -> str:
    """Determine accurate study status based on multiple factors.

    Status can be:
    - not_started: No database or 0 trials
    - running: Active process found
    - paused: Has trials but no active process and not completed
    - completed: Reached trial target
    - failed: Has error indicators (future enhancement)

    Args:
        study_id: The study identifier
        trial_count: Number of completed trials
        total_trials: Target number of trials from config
        has_db: Whether the study database exists

    Returns:
        Status string: "not_started", "running", "paused", or "completed"
    """
    # No database or no trials = not started
    if not has_db or trial_count == 0:
        return "not_started"

    # Check if we've reached the target
    if trial_count >= total_trials:
        return "completed"

    # Check if process is actively running
    if is_optimization_running(study_id):
        return "running"

    # Has trials but not running and not complete = paused
    return "paused"


def _load_study_info(study_dir: Path, topic: Optional[str] = None) -> Optional[dict]:
    """Load study info from a study directory. Returns None if not a valid study."""
    # Look for optimization config (check multiple locations)
    config_file = study_dir / "optimization_config.json"
    if not config_file.exists():
        config_file = study_dir / "1_setup" / "optimization_config.json"
    if not config_file.exists():
        return None

    # Load config
    with open(config_file) as f:
        config = json.load(f)

    # Check if results directory exists (support both 2_results and 3_results)
    results_dir = study_dir / "2_results"
    if not results_dir.exists():
        results_dir = study_dir / "3_results"

    # Check for Optuna database (Protocol 10) or JSON history (other protocols)
    study_db = results_dir / "study.db"
    history_file = results_dir / "optimization_history_incremental.json"

    trial_count = 0
    best_value = None
    has_db = False

    # Protocol 10: Read from Optuna SQLite database
    if study_db.exists():
        has_db = True
        try:
            # Use timeout to avoid blocking on locked databases
            conn = sqlite3.connect(str(study_db), timeout=2.0)
            cursor = conn.cursor()

            # Get trial count and status
            cursor.execute("SELECT COUNT(*) FROM trials WHERE state = 'COMPLETE'")
            trial_count = cursor.fetchone()[0]

            # Get best trial (for single-objective, or first objective for multi-objective)
            if trial_count > 0:
                cursor.execute("""
                    SELECT value FROM trial_values
                    WHERE trial_id IN (
                        SELECT trial_id FROM trials WHERE state = 'COMPLETE'
                    )
                    ORDER BY value ASC
                    LIMIT 1
                """)
                result = cursor.fetchone()
                if result:
                    best_value = result[0]

            conn.close()

        except Exception as e:
            print(f"Warning: Failed to read Optuna database for {study_dir.name}: {e}")

    # Legacy: Read from JSON history
    elif history_file.exists():
        has_db = True
        with open(history_file) as f:
            history = json.load(f)
            trial_count = len(history)
            if history:
                # Find best trial
                best_trial = min(history, key=lambda x: x['objective'])
                best_value = best_trial['objective']

    # Get total trials from config (supports both formats)
    total_trials = (
        config.get('optimization_settings', {}).get('n_trials') or
        config.get('optimization', {}).get('n_trials') or
        config.get('trials', {}).get('n_trials', 50)
    )

    # Get accurate status using process detection
    status = get_accurate_study_status(study_dir.name, trial_count, total_trials, has_db)

    # Get creation date from directory or config modification time
    created_at = None
    try:
        # First try to get from database (most accurate)
        if study_db.exists():
            created_at = datetime.fromtimestamp(study_db.stat().st_mtime).isoformat()
        elif config_file.exists():
            created_at = datetime.fromtimestamp(config_file.stat().st_mtime).isoformat()
        else:
            created_at = datetime.fromtimestamp(study_dir.stat().st_ctime).isoformat()
    except:
        created_at = None

    # Get last modified time
    last_modified = None
    try:
        if study_db.exists():
            last_modified = datetime.fromtimestamp(study_db.stat().st_mtime).isoformat()
        elif history_file.exists():
            last_modified = datetime.fromtimestamp(history_file.stat().st_mtime).isoformat()
    except:
        last_modified = None

    return {
        "id": study_dir.name,
        "name": study_dir.name.replace("_", " ").title(),
        "topic": topic,  # NEW: topic field for grouping
        "status": status,
        "progress": {
            "current": trial_count,
            "total": total_trials
        },
        "best_value": best_value,
        "target": config.get('target', {}).get('value'),
        "path": str(study_dir),
        "created_at": created_at,
        "last_modified": last_modified
    }


@router.get("/studies")
async def list_studies():
    """List all available optimization studies.

    Supports both flat and nested folder structures:
    - Flat: studies/study_name/
    - Nested: studies/Topic/study_name/

    Returns studies with 'topic' field for frontend grouping.
    """
    try:
        studies = []
        if not STUDIES_DIR.exists():
            return {"studies": []}

        for item in STUDIES_DIR.iterdir():
            if not item.is_dir():
                continue
            if item.name.startswith('.'):
                continue

            # Check if this is a study (flat structure) or a topic folder (nested structure)
            is_study = (item / "1_setup").exists() or (item / "optimization_config.json").exists()

            if is_study:
                # Flat structure: study directly in studies/
                study_info = _load_study_info(item, topic=None)
                if study_info:
                    studies.append(study_info)
            else:
                # Nested structure: this might be a topic folder
                # Check if it contains study subdirectories
                for sub_item in item.iterdir():
                    if not sub_item.is_dir():
                        continue
                    if sub_item.name.startswith('.'):
                        continue

                    # Check if this subdirectory is a study
                    sub_is_study = (sub_item / "1_setup").exists() or (sub_item / "optimization_config.json").exists()
                    if sub_is_study:
                        study_info = _load_study_info(sub_item, topic=item.name)
                        if study_info:
                            studies.append(study_info)

        return {"studies": studies}

    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to list studies: {str(e)}")

@router.get("/studies/{study_id}/status")
async def get_study_status(study_id: str):
    """Get detailed status of a specific study"""
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Load config (check multiple locations)
        config_file = study_dir / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "1_setup" / "optimization_config.json"

        with open(config_file) as f:
            config = json.load(f)

        # Check for results (support both 2_results and 3_results)
        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"
        history_file = results_dir / "optimization_history_incremental.json"

        # Protocol 10: Read from Optuna database
        if study_db.exists():
            conn = sqlite3.connect(str(study_db))
            cursor = conn.cursor()

            # Get trial counts by state
            cursor.execute("SELECT COUNT(*) FROM trials WHERE state = 'COMPLETE'")
            trial_count = cursor.fetchone()[0]

            cursor.execute("SELECT COUNT(*) FROM trials WHERE state = 'PRUNED'")
            pruned_count = cursor.fetchone()[0]

            # Get best trial (first objective for multi-objective)
            best_trial = None
            if trial_count > 0:
                cursor.execute("""
                    SELECT t.trial_id, t.number, tv.value
                    FROM trials t
                    JOIN trial_values tv ON t.trial_id = tv.trial_id
                    WHERE t.state = 'COMPLETE'
                    ORDER BY tv.value ASC
                    LIMIT 1
                """)
                result = cursor.fetchone()
                if result:
                    trial_id, trial_number, best_value = result

                    # Get parameters for this trial
                    cursor.execute("""
                        SELECT param_name, param_value
                        FROM trial_params
                        WHERE trial_id = ?
                    """, (trial_id,))
                    params = {row[0]: row[1] for row in cursor.fetchall()}

                    best_trial = {
                        "trial_number": trial_number,
                        "objective": best_value,
                        "design_variables": params,
                        "results": {"first_frequency": best_value}
                    }

            conn.close()

            total_trials = config.get('optimization_settings', {}).get('n_trials', 50)
            status = get_accurate_study_status(study_id, trial_count, total_trials, True)

            return {
                "study_id": study_id,
                "status": status,
                "progress": {
                    "current": trial_count,
                    "total": total_trials,
                    "percentage": (trial_count / total_trials * 100) if total_trials > 0 else 0
                },
                "best_trial": best_trial,
                "pruned_trials": pruned_count,
                "config": config
            }

        # Legacy: Read from JSON history
        if not history_file.exists():
            return {
                "study_id": study_id,
                "status": "not_started",
                "progress": {"current": 0, "total": config.get('trials', {}).get('n_trials', 50)},
                "config": config
            }

        with open(history_file) as f:
            history = json.load(f)

        trial_count = len(history)
        total_trials = config.get('trials', {}).get('n_trials', 50)

        # Find best trial
        best_trial = None
        if history:
            best_trial = min(history, key=lambda x: x['objective'])

        # Check for pruning data
        pruning_file = results_dir / "pruning_history.json"
        pruned_count = 0
        if pruning_file.exists():
            with open(pruning_file) as f:
                pruning_history = json.load(f)
                pruned_count = len(pruning_history)

        status = "completed" if trial_count >= total_trials else "running"

        return {
            "study_id": study_id,
            "status": status,
            "progress": {
                "current": trial_count,
                "total": total_trials,
                "percentage": (trial_count / total_trials * 100) if total_trials > 0 else 0
            },
            "best_trial": best_trial,
            "pruned_trials": pruned_count,
            "config": config
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get study status: {str(e)}")

@router.get("/studies/{study_id}/history")
async def get_optimization_history(study_id: str, limit: Optional[int] = None):
    """Get optimization history (all trials)"""
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"
        history_file = results_dir / "optimization_history_incremental.json"

        # Protocol 10: Read from Optuna database
        if study_db.exists():
            conn = sqlite3.connect(str(study_db))
            cursor = conn.cursor()

            # Get all completed trials FROM ALL STUDIES in the database
            # This handles adaptive optimizations that create multiple Optuna studies
            # (e.g., v11_fea for FEA trials, v11_iter1_nn for NN trials, etc.)
            cursor.execute("""
                SELECT t.trial_id, t.number, t.datetime_start, t.datetime_complete, s.study_name
                FROM trials t
                JOIN studies s ON t.study_id = s.study_id
                WHERE t.state = 'COMPLETE'
                ORDER BY t.datetime_start DESC
            """ + (f" LIMIT {limit}" if limit else ""))

            trial_rows = cursor.fetchall()

            trials = []
            for trial_id, trial_num, start_time, end_time, study_name in trial_rows:
                # Get objectives for this trial
                cursor.execute("""
                    SELECT value
                    FROM trial_values
                    WHERE trial_id = ?
                    ORDER BY objective
                """, (trial_id,))
                values = [row[0] for row in cursor.fetchall()]

                # Get parameters for this trial
                cursor.execute("""
                    SELECT param_name, param_value
                    FROM trial_params
                    WHERE trial_id = ?
                """, (trial_id,))
                params = {}
                for param_name, param_value in cursor.fetchall():
                    try:
                        params[param_name] = float(param_value) if param_value is not None else None
                    except (ValueError, TypeError):
                        params[param_name] = param_value

                # Get user attributes (extracted results: mass, frequency, stress, displacement, etc.)
                cursor.execute("""
                    SELECT key, value_json
                    FROM trial_user_attributes
                    WHERE trial_id = ?
                """, (trial_id,))
                user_attrs = {}
                for key, value_json in cursor.fetchall():
                    try:
                        user_attrs[key] = json.loads(value_json)
                    except (ValueError, TypeError):
                        user_attrs[key] = value_json

                # Extract ALL numeric metrics from user_attrs for results
                # This ensures multi-objective studies show all Zernike metrics, RMS values, etc.
                results = {}
                excluded_keys = {"design_vars", "constraint_satisfied", "constraint_violations"}
                for key, val in user_attrs.items():
                    if key in excluded_keys:
                        continue
                    # Include numeric values and lists of numbers
                    if isinstance(val, (int, float)):
                        results[key] = val
                    elif isinstance(val, list) and len(val) > 0 and isinstance(val[0], (int, float)):
                        # For lists, store as-is (e.g., Zernike coefficients)
                        results[key] = val
                    elif key == "objectives" and isinstance(val, dict):
                        # Extract nested objectives dict (Zernike multi-objective studies)
                        for obj_key, obj_val in val.items():
                            if isinstance(obj_val, (int, float)):
                                results[obj_key] = obj_val
                # Fallback to first frequency from objectives if available
                if not results and len(values) > 0:
                    results["first_frequency"] = values[0]

                # CRITICAL: Extract design_vars from user_attrs if stored there
                # The optimization code does: trial.set_user_attr("design_vars", design_vars)
                design_vars_from_attrs = user_attrs.get("design_vars", {})

                # Merge with params (prefer user_attrs design_vars if available)
                final_design_vars = {**params, **design_vars_from_attrs} if design_vars_from_attrs else params

                # Extract source for FEA vs NN differentiation
                source = user_attrs.get("source", "FEA")  # Default to FEA for legacy studies

                # Get iter_num from user_attrs if available (this is the actual iteration folder number)
                iter_num = user_attrs.get("iter_num", None)

                # Use iter_num if available, otherwise use trial_id as unique identifier
                # trial_id is unique across all studies in the database
                unique_trial_num = iter_num if iter_num is not None else trial_id

                trials.append({
                    "trial_number": unique_trial_num,
                    "trial_id": trial_id,  # Keep original for debugging
                    "optuna_trial_num": trial_num,  # Keep original Optuna trial number
                    "objective": values[0] if len(values) > 0 else None,  # Primary objective
                    "objectives": values if len(values) > 1 else None,  # All objectives for multi-objective
                    "design_variables": final_design_vars,  # Use merged design vars
                    "results": results,
                    "user_attrs": user_attrs,  # Include all user attributes
                    "source": source,  # FEA or NN
                    "start_time": start_time,
                    "end_time": end_time,
                    "study_name": study_name  # Include for debugging
                })

            conn.close()
            return {"trials": trials}

        # Legacy: Read from JSON history
        if not history_file.exists():
            return {"trials": []}

        with open(history_file) as f:
            history = json.load(f)

        # Apply limit if specified
        if limit:
            history = history[-limit:]

        return {"trials": history}

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get history: {str(e)}")

@router.get("/studies/{study_id}/pruning")
async def get_pruning_history(study_id: str):
    """Get pruning diagnostics from Optuna database or legacy JSON file"""
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"
        pruning_file = results_dir / "pruning_history.json"

        # Protocol 10+: Read from Optuna database
        if study_db.exists():
            conn = sqlite3.connect(str(study_db))
            cursor = conn.cursor()

            # Get all pruned trials from Optuna database
            cursor.execute("""
                SELECT t.trial_id, t.number, t.datetime_start, t.datetime_complete
                FROM trials t
                WHERE t.state = 'PRUNED'
                ORDER BY t.number DESC
            """)
            pruned_rows = cursor.fetchall()

            pruned_trials = []
            for trial_id, trial_num, start_time, end_time in pruned_rows:
                # Get parameters for this trial
                cursor.execute("""
                    SELECT param_name, param_value
                    FROM trial_params
                    WHERE trial_id = ?
                """, (trial_id,))
                params = {row[0]: row[1] for row in cursor.fetchall()}

                # Get user attributes (may contain pruning cause)
                cursor.execute("""
                    SELECT key, value_json
                    FROM trial_user_attributes
                    WHERE trial_id = ?
                """, (trial_id,))
                user_attrs = {}
                for key, value_json in cursor.fetchall():
                    try:
                        user_attrs[key] = json.loads(value_json)
                    except (ValueError, TypeError):
                        user_attrs[key] = value_json

                pruned_trials.append({
                    "trial_number": trial_num,
                    "params": params,
                    "pruning_cause": user_attrs.get("pruning_cause", "Unknown"),
                    "start_time": start_time,
                    "end_time": end_time
                })

            conn.close()
            return {"pruned_trials": pruned_trials, "count": len(pruned_trials)}

        # Legacy: Read from JSON history
        if not pruning_file.exists():
            return {"pruned_trials": [], "count": 0}

        with open(pruning_file) as f:
            pruning_history = json.load(f)

        return {"pruned_trials": pruning_history, "count": len(pruning_history)}

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get pruning history: {str(e)}")

def _infer_objective_unit(objective: Dict) -> str:
    """Infer unit from objective name and description"""
    name = objective.get("name", "").lower()
    desc = objective.get("description", "").lower()

    # Common unit patterns
    if "frequency" in name or "hz" in desc:
        return "Hz"
    elif "stiffness" in name or "n/mm" in desc:
        return "N/mm"
    elif "mass" in name or "kg" in desc:
        return "kg"
    elif "stress" in name or "mpa" in desc or "pa" in desc:
        return "MPa"
    elif "displacement" in name or "mm" in desc:
        return "mm"
    elif "force" in name or "newton" in desc:
        return "N"
    elif "%" in desc or "percent" in desc:
        return "%"

    # Check if unit is explicitly mentioned in description (e.g., "(N/mm)")
    import re
    unit_match = re.search(r'\(([^)]+)\)', desc)
    if unit_match:
        return unit_match.group(1)

    return ""  # No unit found

@router.get("/studies/{study_id}/metadata")
async def get_study_metadata(study_id: str):
    """Read optimization_config.json for objectives, design vars, units (Protocol 13)"""
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Load config (check multiple locations)
        config_file = study_dir / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "1_setup" / "optimization_config.json"

        if not config_file.exists():
            raise HTTPException(status_code=404, detail=f"Config file not found for study {study_id}")

        with open(config_file) as f:
            config = json.load(f)

        # Enhance objectives with inferred units if not present
        objectives = config.get("objectives", [])
        for obj in objectives:
            if "unit" not in obj or not obj["unit"]:
                obj["unit"] = _infer_objective_unit(obj)

        # Get sampler/algorithm info
        optimization = config.get("optimization", {})
        algorithm = optimization.get("algorithm", "TPE")

        # Map algorithm names to Optuna sampler names for frontend display
        sampler_map = {
            "CMA-ES": "CmaEsSampler",
            "cma-es": "CmaEsSampler",
            "cmaes": "CmaEsSampler",
            "TPE": "TPESampler",
            "tpe": "TPESampler",
            "NSGA-II": "NSGAIISampler",
            "nsga-ii": "NSGAIISampler",
            "NSGA-III": "NSGAIIISampler",
            "Random": "RandomSampler",
        }
        sampler = sampler_map.get(algorithm, algorithm)

        return {
            "objectives": objectives,
            "design_variables": config.get("design_variables", []),
            "constraints": config.get("constraints", []),
            "study_name": config.get("study_name", study_id),
            "description": config.get("description", ""),
            "sampler": sampler,
            "algorithm": algorithm,
            "n_trials": optimization.get("n_trials", 100)
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get study metadata: {str(e)}")

@router.get("/studies/{study_id}/optimizer-state")
async def get_optimizer_state(study_id: str):
    """
    Read realtime optimizer state from multiple sources:
    1. dashboard_state.json (written by optimization scripts)
    2. intelligent_optimizer/optimizer_state.json (Protocol 13)
    3. Computed state from database (fallback)
    """
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)

        # Priority 1: Check for dashboard_state.json (written by optimization script)
        dashboard_state_file = results_dir / "dashboard_state.json"
        if dashboard_state_file.exists():
            try:
                with open(dashboard_state_file) as f:
                    state = json.load(f)
                return {"available": True, "source": "dashboard_state", **state}
            except Exception:
                pass

        # Priority 2: Check intelligent_optimizer state
        imso_state_file = results_dir / "intelligent_optimizer" / "optimizer_state.json"
        if imso_state_file.exists():
            try:
                with open(imso_state_file) as f:
                    state = json.load(f)
                return {"available": True, "source": "intelligent_optimizer", **state}
            except Exception:
                pass

        # Priority 3: Compute state from config and database
        config_file = study_dir / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "1_setup" / "optimization_config.json"

        if config_file.exists():
            with open(config_file) as f:
                config = json.load(f)

            # Extract sampler info
            optimizer_config = config.get("optimizer", {})
            sampler_name = optimizer_config.get("sampler", "TPESampler")
            n_trials = optimizer_config.get("n_trials", config.get("n_trials", 100))

            # Get trial count from database
            study_db = results_dir / "study.db"
            completed_trials = 0
            best_value = None

            if study_db.exists():
                try:
                    conn = sqlite3.connect(str(study_db), timeout=2.0)
                    cursor = conn.cursor()
                    cursor.execute("SELECT COUNT(*) FROM trials WHERE state = 'COMPLETE'")
                    completed_trials = cursor.fetchone()[0]

                    # Get best value
                    cursor.execute("""
                        SELECT MIN(value) FROM trial_values
                        WHERE trial_id IN (SELECT trial_id FROM trials WHERE state = 'COMPLETE')
                    """)
                    result = cursor.fetchone()
                    if result and result[0] is not None:
                        best_value = result[0]

                    conn.close()
                except Exception:
                    pass

            # Determine phase
            if completed_trials == 0:
                phase = "initializing"
                phase_description = "Waiting for first trial"
                phase_progress = 0.0
            elif completed_trials < 10:
                phase = "exploration"
                phase_description = "Initial random sampling"
                phase_progress = completed_trials / 10
            elif completed_trials < n_trials * 0.5:
                phase = "exploitation"
                phase_description = "Focusing on promising regions"
                phase_progress = (completed_trials - 10) / (n_trials * 0.5 - 10)
            elif completed_trials < n_trials * 0.9:
                phase = "refinement"
                phase_description = "Fine-tuning best solutions"
                phase_progress = (completed_trials - n_trials * 0.5) / (n_trials * 0.4)
            else:
                phase = "convergence"
                phase_description = "Near completion"
                phase_progress = (completed_trials - n_trials * 0.9) / (n_trials * 0.1)

            # Build objectives list from config
            objectives = []
            if "objectives" in config:
                for obj in config["objectives"]:
                    objectives.append({
                        "name": obj.get("name", "objective"),
                        "direction": obj.get("direction", "minimize"),
                        "current_best": best_value if len(objectives) == 0 else None,
                        "unit": obj.get("units", "")
                    })

            # Sampler descriptions
            sampler_descriptions = {
                "TPESampler": "Tree-structured Parzen Estimator - Bayesian optimization using kernel density estimation",
                "NSGAIISampler": "NSGA-II - Multi-objective genetic algorithm with non-dominated sorting",
                "NSGAIIISampler": "NSGA-III - Reference-point based multi-objective optimization",
                "CmaEsSampler": "CMA-ES - Covariance Matrix Adaptation Evolution Strategy",
                "RandomSampler": "Random sampling - Uniform random search across parameter space",
                "QMCSampler": "Quasi-Monte Carlo - Low-discrepancy sequence sampling"
            }

            return {
                "available": True,
                "source": "computed",
                "phase": phase,
                "phase_description": phase_description,
                "phase_progress": min(1.0, max(0.0, phase_progress)),
                "current_strategy": f"{sampler_name} sampling",
                "sampler": {
                    "name": sampler_name,
                    "description": sampler_descriptions.get(sampler_name, f"{sampler_name} optimization algorithm")
                },
                "objectives": objectives,
                "plan": {
                    "total_phases": 4,
                    "current_phase": {"initializing": 0, "exploration": 1, "exploitation": 2, "refinement": 3, "convergence": 4}.get(phase, 0),
                    "phases": ["Exploration", "Exploitation", "Refinement", "Convergence"]
                },
                "completed_trials": completed_trials,
                "total_trials": n_trials
            }

        return {"available": False}

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get optimizer state: {str(e)}")

@router.get("/studies/{study_id}/pareto-front")
async def get_pareto_front(study_id: str):
    """Get Pareto-optimal solutions for multi-objective studies (Protocol 13)"""
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"

        if not study_db.exists():
            return {"is_multi_objective": False, "pareto_front": []}

        # Import optuna here to avoid loading it for all endpoints
        import optuna

        storage = optuna.storages.RDBStorage(f"sqlite:///{study_db}")
        study = optuna.load_study(study_name=study_id, storage=storage)

        # Check if multi-objective
        if len(study.directions) == 1:
            return {"is_multi_objective": False, "pareto_front": []}

        # Get Pareto front
        pareto_trials = study.best_trials

        return {
            "is_multi_objective": True,
            "pareto_front": [
                {
                    "trial_number": t.number,
                    "values": t.values,
                    "params": t.params,
                    "user_attrs": dict(t.user_attrs),
                    "constraint_satisfied": t.user_attrs.get("constraint_satisfied", True)
                }
                for t in pareto_trials
            ]
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get Pareto front: {str(e)}")


@router.get("/studies/{study_id}/nn-pareto-front")
async def get_nn_pareto_front(study_id: str):
    """Get NN surrogate Pareto front from nn_pareto_front.json"""
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)
        nn_pareto_file = results_dir / "nn_pareto_front.json"

        if not nn_pareto_file.exists():
            return {"has_nn_results": False, "pareto_front": []}

        with open(nn_pareto_file) as f:
            nn_pareto = json.load(f)

        # Transform to match Trial interface format
        transformed = []
        for trial in nn_pareto:
            transformed.append({
                "trial_number": trial.get("trial_number"),
                "values": [trial.get("mass"), trial.get("frequency")],
                "params": trial.get("params", {}),
                "user_attrs": {
                    "source": "NN",
                    "feasible": trial.get("feasible", False),
                    "predicted_stress": trial.get("predicted_stress"),
                    "predicted_displacement": trial.get("predicted_displacement"),
                    "mass": trial.get("mass"),
                    "frequency": trial.get("frequency")
                },
                "constraint_satisfied": trial.get("feasible", False),
                "source": "NN"
            })

        return {
            "has_nn_results": True,
            "pareto_front": transformed,
            "count": len(transformed)
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get NN Pareto front: {str(e)}")


@router.get("/studies/{study_id}/nn-state")
async def get_nn_optimization_state(study_id: str):
    """Get NN optimization state/summary from nn_optimization_state.json"""
    try:
        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)
        nn_state_file = results_dir / "nn_optimization_state.json"

        if not nn_state_file.exists():
            return {"has_nn_state": False}

        with open(nn_state_file) as f:
            state = json.load(f)

        return {
            "has_nn_state": True,
            "total_fea_count": state.get("total_fea_count", 0),
            "total_nn_count": state.get("total_nn_count", 0),
            "pareto_front_size": state.get("pareto_front_size", 0),
            "best_mass": state.get("best_mass"),
            "best_frequency": state.get("best_frequency"),
            "timestamp": state.get("timestamp")
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get NN state: {str(e)}")


@router.post("/studies")
async def create_study(
    config: str = Form(...),
    prt_file: Optional[UploadFile] = File(None),
    sim_file: Optional[UploadFile] = File(None),
    fem_file: Optional[UploadFile] = File(None)
):
    """
    Create a new optimization study
    Accepts:
    - config: JSON string with study configuration
    - prt_file: NX part file (optional if using existing study)
    - sim_file: NX simulation file (optional)
    - fem_file: NX FEM file (optional)
    """
    try:
        # Parse config
        config_data = json.loads(config)
        study_name = config_data.get("name") # Changed from study_name to name to match frontend

        if not study_name:
            raise HTTPException(status_code=400, detail="name is required in config")

        # Create study directory structure
        study_dir = STUDIES_DIR / study_name
        if study_dir.exists():
            raise HTTPException(status_code=400, detail=f"Study {study_name} already exists")

        setup_dir = study_dir / "1_setup"
        model_dir = setup_dir / "model"
        results_dir = study_dir / "2_results"

        setup_dir.mkdir(parents=True, exist_ok=True)
        model_dir.mkdir(parents=True, exist_ok=True)
        results_dir.mkdir(parents=True, exist_ok=True)

        # Save config file
        config_file = setup_dir / "optimization_config.json"
        with open(config_file, 'w') as f:
            json.dump(config_data, f, indent=2)

        # Save uploaded files
        files_saved = {}
        if prt_file:
            prt_path = model_dir / prt_file.filename
            with open(prt_path, 'wb') as f:
                content = await prt_file.read()
                f.write(content)
            files_saved['prt_file'] = str(prt_path)

        if sim_file:
            sim_path = model_dir / sim_file.filename
            with open(sim_path, 'wb') as f:
                content = await sim_file.read()
                f.write(content)
            files_saved['sim_file'] = str(sim_path)

        if fem_file:
            fem_path = model_dir / fem_file.filename
            with open(fem_path, 'wb') as f:
                content = await fem_file.read()
                f.write(content)
            files_saved['fem_file'] = str(fem_path)

        return JSONResponse(
            status_code=201,
            content={
                "status": "created",
                "study_id": study_name,
                "study_path": str(study_dir),
                "config_path": str(config_file),
                "files_saved": files_saved,
                "message": f"Study {study_name} created successfully. Ready to run optimization."
            }
        )

    except json.JSONDecodeError as e:
        raise HTTPException(status_code=400, detail=f"Invalid JSON in config: {str(e)}")
    except Exception as e:
        # Clean up on error
        if 'study_dir' in locals() and study_dir.exists():
            shutil.rmtree(study_dir)
        raise HTTPException(status_code=500, detail=f"Failed to create study: {str(e)}")

@router.post("/studies/{study_id}/convert-mesh")
async def convert_study_mesh(study_id: str):
    """
    Convert study mesh to GLTF for 3D visualization
    Creates a web-viewable 3D model with FEA results as vertex colors
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Import mesh converter
        sys.path.append(str(Path(__file__).parent.parent.parent.parent.parent))
        from optimization_engine.nx.mesh_converter import convert_study_mesh

        # Convert mesh
        output_path = convert_study_mesh(study_dir)

        if output_path and output_path.exists():
            return {
                "status": "success",
                "gltf_path": str(output_path),
                "gltf_url": f"/api/optimization/studies/{study_id}/mesh/model.gltf",
                "metadata_url": f"/api/optimization/studies/{study_id}/mesh/model.json",
                "message": "Mesh converted successfully"
            }
        else:
            raise HTTPException(status_code=500, detail="Mesh conversion failed")

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to convert mesh: {str(e)}")

@router.get("/studies/{study_id}/mesh/{filename}")
async def get_mesh_file(study_id: str, filename: str):
    """
    Serve GLTF mesh files and metadata
    Supports .gltf, .bin, and .json files
    """
    try:
        # Validate filename to prevent directory traversal
        if '..' in filename or '/' in filename or '\\' in filename:
            raise HTTPException(status_code=400, detail="Invalid filename")

        study_dir = resolve_study_path(study_id)
        visualization_dir = study_dir / "3_visualization"

        file_path = visualization_dir / filename

        if not file_path.exists():
            raise HTTPException(status_code=404, detail=f"File {filename} not found")

        # Determine content type
        suffix = file_path.suffix.lower()
        content_types = {
            '.gltf': 'model/gltf+json',
            '.bin': 'application/octet-stream',
            '.json': 'application/json',
            '.glb': 'model/gltf-binary'
        }

        content_type = content_types.get(suffix, 'application/octet-stream')

        return FileResponse(
            path=str(file_path),
            media_type=content_type,
            filename=filename
        )

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"File not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to serve mesh file: {str(e)}")

@router.get("/studies/{study_id}/optuna-url")
async def get_optuna_dashboard_url(study_id: str):
    """
    Get the Optuna dashboard URL for a specific study.
    Returns the URL to access the study in Optuna dashboard.

    The Optuna dashboard should be started with a relative path from the Atomizer root:
    sqlite:///studies/{study_id}/2_results/study.db
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"

        if not study_db.exists():
            raise HTTPException(status_code=404, detail=f"No Optuna database found for study {study_id}")

        # Get the study name from the database (may differ from folder name)
        import optuna
        storage = optuna.storages.RDBStorage(f"sqlite:///{study_db}")
        studies = storage.get_all_studies()

        if not studies:
            raise HTTPException(status_code=404, detail=f"No Optuna study found in database for {study_id}")

        # Use the actual study name from the database
        optuna_study_name = studies[0].study_name

        # Return URL info for the frontend
        # The dashboard should be running on port 8081 with the correct database
        return {
            "study_id": study_id,
            "optuna_study_name": optuna_study_name,
            "database_path": f"studies/{study_id}/2_results/study.db",
            "dashboard_url": f"http://localhost:8081/dashboard/studies/{studies[0]._study_id}",
            "dashboard_base": "http://localhost:8081",
            "note": "Optuna dashboard must be started with: sqlite:///studies/{study_id}/2_results/study.db"
        }

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get Optuna URL: {str(e)}")

@router.post("/studies/{study_id}/generate-report")
async def generate_report(
    study_id: str,
    format: str = "markdown",
    include_llm_summary: bool = False
):
    """
    Generate an optimization report in the specified format

    Args:
        study_id: Study identifier
        format: Report format ('markdown', 'html', or 'pdf')
        include_llm_summary: Whether to include LLM-generated executive summary

    Returns:
        Information about the generated report including download URL
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Validate format
        valid_formats = ['markdown', 'md', 'html', 'pdf']
        if format.lower() not in valid_formats:
            raise HTTPException(status_code=400, detail=f"Invalid format. Must be one of: {', '.join(valid_formats)}")

        # Import report generator
        sys.path.append(str(Path(__file__).parent.parent.parent.parent.parent))
        from optimization_engine.future.report_generator import generate_study_report

        # Generate report
        output_path = generate_study_report(
            study_dir=study_dir,
            output_format=format.lower(),
            include_llm_summary=include_llm_summary
        )

        if output_path and output_path.exists():
            # Get relative path for URL
            rel_path = output_path.relative_to(study_dir)

            return {
                "status": "success",
                "format": format,
                "file_path": str(output_path),
                "download_url": f"/api/optimization/studies/{study_id}/reports/{output_path.name}",
                "file_size": output_path.stat().st_size,
                "message": f"Report generated successfully in {format} format"
            }
        else:
            raise HTTPException(status_code=500, detail="Report generation failed")

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Study {study_id} not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to generate report: {str(e)}")

@router.get("/studies/{study_id}/reports/{filename}")
async def download_report(study_id: str, filename: str):
    """
    Download a generated report file

    Args:
        study_id: Study identifier
        filename: Report filename

    Returns:
        Report file for download
    """
    try:
        # Validate filename to prevent directory traversal
        if '..' in filename or '/' in filename or '\\' in filename:
            raise HTTPException(status_code=400, detail="Invalid filename")

        study_dir = resolve_study_path(study_id)
        results_dir = get_results_dir(study_dir)

        file_path = results_dir / filename

        if not file_path.exists():
            raise HTTPException(status_code=404, detail=f"Report file {filename} not found")

        # Determine content type
        suffix = file_path.suffix.lower()
        content_types = {
            '.md': 'text/markdown',
            '.html': 'text/html',
            '.pdf': 'application/pdf',
            '.json': 'application/json'
        }

        content_type = content_types.get(suffix, 'application/octet-stream')

        return FileResponse(
            path=str(file_path),
            media_type=content_type,
            filename=filename,
            headers={"Content-Disposition": f"attachment; filename={filename}"}
        )

    except FileNotFoundError:
        raise HTTPException(status_code=404, detail=f"Report file not found")
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to download report: {str(e)}")


@router.get("/studies/{study_id}/console")
async def get_console_output(study_id: str, lines: int = 200):
    """
    Get the latest console output/logs from the optimization run

    Args:
        study_id: Study identifier
        lines: Number of lines to return (default: 200)

    Returns:
        JSON with console output lines
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Look for log files in various locations
        log_paths = [
            study_dir / "optimization.log",
            study_dir / "2_results" / "optimization.log",
            study_dir / "3_results" / "optimization.log",
            study_dir / "run.log",
        ]

        log_content = None
        log_path_used = None

        for log_path in log_paths:
            if log_path.exists():
                log_path_used = log_path
                break

        if log_path_used is None:
            return {
                "lines": [],
                "total_lines": 0,
                "log_file": None,
                "message": "No log file found. Optimization may not have started yet."
            }

        # Read the last N lines efficiently
        with open(log_path_used, 'r', encoding='utf-8', errors='replace') as f:
            all_lines = f.readlines()

        # Get last N lines
        last_lines = all_lines[-lines:] if len(all_lines) > lines else all_lines

        # Clean up lines (remove trailing newlines)
        last_lines = [line.rstrip('\n\r') for line in last_lines]

        return {
            "lines": last_lines,
            "total_lines": len(all_lines),
            "displayed_lines": len(last_lines),
            "log_file": str(log_path_used),
            "timestamp": datetime.now().isoformat()
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to read console output: {str(e)}")


@router.get("/studies/{study_id}/report")
async def get_study_report(study_id: str):
    """
    Get the STUDY_REPORT.md file content for a study

    Args:
        study_id: Study identifier

    Returns:
        JSON with the markdown content
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Look for STUDY_REPORT.md in the study root
        report_path = study_dir / "STUDY_REPORT.md"

        if not report_path.exists():
            raise HTTPException(status_code=404, detail="No STUDY_REPORT.md found for this study")

        with open(report_path, 'r', encoding='utf-8') as f:
            content = f.read()

        return {
            "content": content,
            "path": str(report_path),
            "study_id": study_id
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to read study report: {str(e)}")


# ============================================================================
# Study README and Config Endpoints
# ============================================================================

@router.get("/studies/{study_id}/readme")
async def get_study_readme(study_id: str):
    """
    Get the README.md file content for a study (from 1_setup folder)

    Args:
        study_id: Study identifier

    Returns:
        JSON with the markdown content
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Look for README.md in various locations
        readme_paths = [
            study_dir / "README.md",
            study_dir / "1_setup" / "README.md",
            study_dir / "readme.md",
        ]

        readme_content = None
        readme_path = None

        for path in readme_paths:
            if path.exists():
                readme_path = path
                with open(path, 'r', encoding='utf-8') as f:
                    readme_content = f.read()
                break

        if readme_content is None:
            # Generate a basic README from config if none exists
            config_file = study_dir / "1_setup" / "optimization_config.json"
            if not config_file.exists():
                config_file = study_dir / "optimization_config.json"

            if config_file.exists():
                with open(config_file) as f:
                    config = json.load(f)

                readme_content = f"""# {config.get('study_name', study_id)}

{config.get('description', 'No description available.')}

## Design Variables
{chr(10).join([f"- **{dv['name']}**: {dv.get('min', '?')} - {dv.get('max', '?')} {dv.get('units', '')}" for dv in config.get('design_variables', [])])}

## Objectives
{chr(10).join([f"- **{obj['name']}**: {obj.get('description', '')} ({obj.get('direction', 'minimize')})" for obj in config.get('objectives', [])])}
"""
            else:
                readme_content = f"# {study_id}\n\nNo README or configuration found for this study."

        return {
            "content": readme_content,
            "path": str(readme_path) if readme_path else None,
            "study_id": study_id
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to read README: {str(e)}")


@router.get("/studies/{study_id}/image/{image_path:path}")
async def get_study_image(study_id: str, image_path: str):
    """
    Serve images from a study directory.

    Supports images in:
    - study_dir/image.png
    - study_dir/1_setup/image.png
    - study_dir/3_results/image.png
    - study_dir/assets/image.png

    Args:
        study_id: Study identifier
        image_path: Relative path to the image within the study

    Returns:
        FileResponse with the image
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Sanitize path to prevent directory traversal
        image_path = image_path.replace('..', '').lstrip('/')

        # Try multiple locations for the image
        possible_paths = [
            study_dir / image_path,
            study_dir / "1_setup" / image_path,
            study_dir / "3_results" / image_path,
            study_dir / "2_results" / image_path,
            study_dir / "assets" / image_path,
        ]

        image_file = None
        for path in possible_paths:
            if path.exists() and path.is_file():
                image_file = path
                break

        if image_file is None:
            raise HTTPException(status_code=404, detail=f"Image not found: {image_path}")

        # Determine media type
        suffix = image_file.suffix.lower()
        media_types = {
            '.png': 'image/png',
            '.jpg': 'image/jpeg',
            '.jpeg': 'image/jpeg',
            '.gif': 'image/gif',
            '.svg': 'image/svg+xml',
            '.webp': 'image/webp',
        }
        media_type = media_types.get(suffix, 'application/octet-stream')

        return FileResponse(image_file, media_type=media_type)

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to serve image: {str(e)}")


@router.get("/studies/{study_id}/config")
async def get_study_config(study_id: str):
    """
    Get the full optimization_config.json for a study

    Args:
        study_id: Study identifier

    Returns:
        JSON with the complete configuration
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Look for config in various locations
        config_file = study_dir / "1_setup" / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "optimization_config.json"

        if not config_file.exists():
            raise HTTPException(status_code=404, detail=f"Config file not found for study {study_id}")

        with open(config_file) as f:
            config = json.load(f)

        return {
            "config": config,
            "path": str(config_file),
            "study_id": study_id
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to read config: {str(e)}")


# ============================================================================
# Process Control Endpoints
# ============================================================================

# Track running processes by study_id
_running_processes: Dict[str, int] = {}

# Track paused processes by study_id
_paused_processes: Dict[str, bool] = {}


def _find_optimization_process(study_id: str) -> Optional[psutil.Process]:
    """Find a running optimization process for a given study.

    Uses multiple detection strategies:
    1. Check tracked processes dictionary (most reliable)
    2. Check process working directory (Windows-friendly)
    3. Check cmdline for script names
    """
    study_dir = resolve_study_path(study_id)
    study_dir_str = str(study_dir).lower()
    study_dir_normalized = study_dir_str.replace('/', '\\')

    # Strategy 1: Check tracked processes first (most reliable)
    if study_id in _running_processes:
        tracked_pid = _running_processes[study_id]
        try:
            proc = psutil.Process(tracked_pid)
            if proc.is_running() and proc.status() != psutil.STATUS_ZOMBIE:
                return proc
            else:
                # Process died, clean up tracking
                del _running_processes[study_id]
        except (psutil.NoSuchProcess, psutil.AccessDenied):
            del _running_processes[study_id]

    # Strategy 2 & 3: Scan all Python processes
    for proc in psutil.process_iter(['pid', 'name', 'cmdline', 'cwd']):
        try:
            # Only check Python processes
            proc_name = (proc.info.get('name') or '').lower()
            if 'python' not in proc_name:
                continue

            cmdline = proc.info.get('cmdline') or []
            cmdline_str = ' '.join(cmdline) if cmdline else ''
            cmdline_lower = cmdline_str.lower()

            # Check if this is an optimization script (run_optimization.py or run_sat_optimization.py)
            is_opt_script = ('run_optimization' in cmdline_lower or
                           'run_sat_optimization' in cmdline_lower or
                           'run_imso' in cmdline_lower)

            if not is_opt_script:
                continue

            # Strategy 2: Check CWD (most reliable on Windows)
            try:
                proc_cwd = proc.cwd().lower().replace('/', '\\')
                if proc_cwd == study_dir_normalized or study_dir_normalized in proc_cwd:
                    # Track this process for future lookups
                    _running_processes[study_id] = proc.pid
                    return proc
            except (psutil.NoSuchProcess, psutil.AccessDenied, OSError):
                pass

            # Strategy 3: Check cmdline for study path or study_id
            cmdline_normalized = cmdline_lower.replace('/', '\\')
            if study_dir_normalized in cmdline_normalized:
                _running_processes[study_id] = proc.pid
                return proc

            # Also check if study_id appears in cmdline (e.g., as part of path)
            if study_id.lower() in cmdline_lower:
                _running_processes[study_id] = proc.pid
                return proc

        except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess):
            continue

    return None


@router.get("/studies/{study_id}/process")
async def get_process_status(study_id: str):
    """
    Get the process status for a study's optimization run

    Args:
        study_id: Study identifier

    Returns:
        JSON with process status (is_running, pid, iteration counts)
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Check if process is running
        proc = _find_optimization_process(study_id)
        is_running = proc is not None
        pid = proc.pid if proc else None

        # Get iteration counts from database
        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"

        fea_count = 0
        nn_count = 0
        iteration = None
        total_trials = None
        time_per_trial = None
        eta_seconds = None
        eta_formatted = None
        rate_per_hour = None
        start_time = None

        # Try to get total_trials from config
        config_file = study_dir / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "1_setup" / "optimization_config.json"
        if config_file.exists():
            try:
                with open(config_file) as f:
                    config = json.load(f)
                    total_trials = config.get("optimizer", {}).get("n_trials")
                    if not total_trials:
                        total_trials = config.get("n_trials")
            except Exception:
                pass

        if study_db.exists():
            try:
                conn = sqlite3.connect(str(study_db))
                cursor = conn.cursor()

                # Count FEA trials (from main study or studies with "_fea" suffix)
                cursor.execute("""
                    SELECT COUNT(*) FROM trials t
                    JOIN studies s ON t.study_id = s.study_id
                    WHERE t.state = 'COMPLETE'
                    AND (s.study_name LIKE '%_fea' OR s.study_name NOT LIKE '%_nn%')
                """)
                fea_count = cursor.fetchone()[0]

                # Count NN trials
                cursor.execute("""
                    SELECT COUNT(*) FROM trials t
                    JOIN studies s ON t.study_id = s.study_id
                    WHERE t.state = 'COMPLETE'
                    AND s.study_name LIKE '%_nn%'
                """)
                nn_count = cursor.fetchone()[0]

                # Try to get current iteration from study names
                cursor.execute("""
                    SELECT study_name FROM studies
                    WHERE study_name LIKE '%_iter%'
                    ORDER BY study_name DESC LIMIT 1
                """)
                result = cursor.fetchone()
                if result:
                    import re
                    match = re.search(r'iter(\d+)', result[0])
                    if match:
                        iteration = int(match.group(1))

                # Calculate ETA from trial timestamps
                # Get timestamps from trial_user_attributes or trial datetime_complete
                cursor.execute("""
                    SELECT datetime_complete FROM trials
                    WHERE state = 'COMPLETE' AND datetime_complete IS NOT NULL
                    ORDER BY datetime_complete DESC
                    LIMIT 20
                """)
                timestamps = cursor.fetchall()

                if len(timestamps) >= 2:
                    from datetime import datetime as dt
                    try:
                        # Parse timestamps and calculate average time between trials
                        times = []
                        for ts in timestamps:
                            if ts[0]:
                                # Handle different timestamp formats
                                ts_str = ts[0]
                                try:
                                    times.append(dt.fromisoformat(ts_str.replace('Z', '+00:00')))
                                except ValueError:
                                    try:
                                        times.append(dt.strptime(ts_str, "%Y-%m-%d %H:%M:%S.%f"))
                                    except ValueError:
                                        times.append(dt.strptime(ts_str, "%Y-%m-%d %H:%M:%S"))

                        if len(times) >= 2:
                            # Sort ascending and calculate time differences
                            times.sort()
                            diffs = [(times[i+1] - times[i]).total_seconds() for i in range(len(times)-1)]
                            # Filter out outliers (e.g., breaks in optimization)
                            diffs = [d for d in diffs if d > 0 and d < 7200]  # Max 2 hours between trials
                            if diffs:
                                time_per_trial = sum(diffs) / len(diffs)
                                rate_per_hour = 3600 / time_per_trial if time_per_trial > 0 else 0

                                # Calculate ETA
                                completed = fea_count + nn_count
                                if total_trials and completed < total_trials:
                                    remaining = total_trials - completed
                                    eta_seconds = time_per_trial * remaining

                                    # Format ETA nicely
                                    if eta_seconds < 60:
                                        eta_formatted = f"{int(eta_seconds)}s"
                                    elif eta_seconds < 3600:
                                        eta_formatted = f"{int(eta_seconds / 60)}m"
                                    else:
                                        hours = int(eta_seconds / 3600)
                                        minutes = int((eta_seconds % 3600) / 60)
                                        eta_formatted = f"{hours}h {minutes}m"

                                # Get start time (first trial)
                                if times:
                                    start_time = times[0].isoformat()
                    except Exception as e:
                        print(f"Warning: Failed to calculate ETA: {e}")

                conn.close()
            except Exception as e:
                print(f"Warning: Failed to read database for process status: {e}")

        # Check if paused
        is_paused = _paused_processes.get(study_id, False) if is_running else False

        return {
            "is_running": is_running,
            "is_paused": is_paused,
            "pid": pid,
            "iteration": iteration,
            "fea_count": fea_count,
            "nn_count": nn_count,
            "total_trials": total_trials,
            "completed_trials": fea_count + nn_count,
            "time_per_trial_seconds": round(time_per_trial, 1) if time_per_trial else None,
            "eta_seconds": round(eta_seconds) if eta_seconds else None,
            "eta_formatted": eta_formatted,
            "rate_per_hour": round(rate_per_hour, 2) if rate_per_hour else None,
            "start_time": start_time,
            "study_id": study_id
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get process status: {str(e)}")


class StartOptimizationRequest(BaseModel):
    freshStart: bool = False
    maxIterations: int = 100
    trials: int = 300  # For SAT scripts
    feaBatchSize: int = 5
    tuneTrials: int = 30
    ensembleSize: int = 3
    patience: int = 5


def _detect_script_type(script_path: Path) -> str:
    """Detect the type of optimization script (SAT, IMSO, etc.)"""
    try:
        content = script_path.read_text(encoding='utf-8', errors='ignore')
        if 'run_sat_optimization' in content or 'SAT' in content or 'Self-Aware Turbo' in content:
            return 'sat'
        elif 'IMSO' in content or 'intelligent_optimizer' in content:
            return 'imso'
        else:
            return 'generic'
    except:
        return 'generic'


@router.post("/studies/{study_id}/start")
async def start_optimization(study_id: str, request: StartOptimizationRequest = None):
    """
    Start the optimization process for a study

    Args:
        study_id: Study identifier
        request: Optional start options

    Returns:
        JSON with process info
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Check if already running
        existing_proc = _find_optimization_process(study_id)
        if existing_proc:
            return {
                "success": False,
                "message": f"Optimization already running (PID: {existing_proc.pid})",
                "pid": existing_proc.pid
            }

        # Find run_optimization.py or run_sat_optimization.py
        run_script = study_dir / "run_sat_optimization.py"
        if not run_script.exists():
            run_script = study_dir / "run_optimization.py"
        if not run_script.exists():
            raise HTTPException(status_code=404, detail=f"No optimization script found for study {study_id}")

        # Detect script type and build appropriate command
        script_type = _detect_script_type(run_script)
        python_exe = sys.executable
        cmd = [python_exe, str(run_script)]

        if request:
            if script_type == 'sat':
                # SAT scripts use --trials
                cmd.extend(["--trials", str(request.trials)])
                if request.freshStart:
                    # SAT scripts might support --resume, so fresh = no --resume
                    pass  # Default is fresh start for SAT
            else:
                # IMSO/generic scripts use different flags
                cmd.append("--start")
                if request.freshStart:
                    cmd.append("--fresh")
                cmd.extend(["--fea-batch", str(request.feaBatchSize)])
                cmd.extend(["--tune-trials", str(request.tuneTrials)])
                cmd.extend(["--ensemble-size", str(request.ensembleSize)])
                cmd.extend(["--patience", str(request.patience)])

        # Start process in background
        proc = subprocess.Popen(
            cmd,
            cwd=str(study_dir),
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            start_new_session=True
        )

        _running_processes[study_id] = proc.pid

        return {
            "success": True,
            "message": f"Optimization started successfully ({script_type} script)",
            "pid": proc.pid,
            "command": ' '.join(cmd),
            "script_type": script_type
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to start optimization: {str(e)}")


class StopRequest(BaseModel):
    force: bool = True  # Default to force kill


@router.post("/studies/{study_id}/stop")
async def stop_optimization(study_id: str, request: StopRequest = None):
    """
    Stop the optimization process for a study (hard kill by default)

    Args:
        study_id: Study identifier
        request.force: If True (default), immediately kill. If False, try graceful first.

    Returns:
        JSON with result
    """
    if request is None:
        request = StopRequest()

    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Find running process
        proc = _find_optimization_process(study_id)

        if not proc:
            return {
                "success": False,
                "message": "No running optimization process found"
            }

        pid = proc.pid
        killed_pids = []

        try:
            # FIRST: Get all children BEFORE killing parent
            children = []
            try:
                children = proc.children(recursive=True)
            except (psutil.NoSuchProcess, psutil.AccessDenied):
                pass

            if request.force:
                # Hard kill: immediately kill parent and all children
                # Kill children first (bottom-up)
                for child in reversed(children):
                    try:
                        child.kill()  # SIGKILL on Unix, TerminateProcess on Windows
                        killed_pids.append(child.pid)
                    except (psutil.NoSuchProcess, psutil.AccessDenied):
                        pass

                # Then kill parent
                try:
                    proc.kill()
                    killed_pids.append(pid)
                except psutil.NoSuchProcess:
                    pass
            else:
                # Graceful: try SIGTERM first, then force
                try:
                    proc.terminate()
                    proc.wait(timeout=5)
                except psutil.TimeoutExpired:
                    # Didn't stop gracefully, force kill
                    for child in reversed(children):
                        try:
                            child.kill()
                            killed_pids.append(child.pid)
                        except (psutil.NoSuchProcess, psutil.AccessDenied):
                            pass
                    proc.kill()
                    killed_pids.append(pid)
                except psutil.NoSuchProcess:
                    pass

            # Clean up tracking
            if study_id in _running_processes:
                del _running_processes[study_id]

            return {
                "success": True,
                "message": f"Optimization killed (PID: {pid}, +{len(children)} children)",
                "pid": pid,
                "killed_pids": killed_pids
            }

        except psutil.NoSuchProcess:
            if study_id in _running_processes:
                del _running_processes[study_id]
            return {
                "success": True,
                "message": "Process already terminated"
            }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to stop optimization: {str(e)}")


@router.post("/studies/{study_id}/pause")
async def pause_optimization(study_id: str):
    """
    Pause the optimization process for a study.

    Uses psutil.Process.suspend() which sends SIGSTOP on Unix or suspends on Windows.

    Args:
        study_id: Study identifier

    Returns:
        JSON with result
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Find running process
        proc = _find_optimization_process(study_id)

        if not proc:
            return {
                "success": False,
                "message": "No running optimization process found"
            }

        # Check if already paused
        if _paused_processes.get(study_id, False):
            return {
                "success": False,
                "message": "Optimization is already paused"
            }

        pid = proc.pid

        try:
            # Suspend the main process and all children
            children = []
            try:
                children = proc.children(recursive=True)
            except (psutil.NoSuchProcess, psutil.AccessDenied):
                pass

            # Suspend children first
            for child in children:
                try:
                    child.suspend()
                except (psutil.NoSuchProcess, psutil.AccessDenied):
                    pass

            # Suspend parent
            proc.suspend()

            # Mark as paused
            _paused_processes[study_id] = True

            return {
                "success": True,
                "message": f"Optimization paused (PID: {pid}, +{len(children)} children)",
                "pid": pid
            }

        except psutil.NoSuchProcess:
            return {
                "success": False,
                "message": "Process no longer exists"
            }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to pause optimization: {str(e)}")


@router.post("/studies/{study_id}/resume")
async def resume_optimization(study_id: str):
    """
    Resume a paused optimization process for a study.

    Uses psutil.Process.resume() which sends SIGCONT on Unix or resumes on Windows.

    Args:
        study_id: Study identifier

    Returns:
        JSON with result
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Find running process
        proc = _find_optimization_process(study_id)

        if not proc:
            return {
                "success": False,
                "message": "No optimization process found"
            }

        # Check if actually paused
        if not _paused_processes.get(study_id, False):
            return {
                "success": False,
                "message": "Optimization is not paused"
            }

        pid = proc.pid

        try:
            # Resume the main process and all children
            children = []
            try:
                children = proc.children(recursive=True)
            except (psutil.NoSuchProcess, psutil.AccessDenied):
                pass

            # Resume parent first
            proc.resume()

            # Resume children
            for child in children:
                try:
                    child.resume()
                except (psutil.NoSuchProcess, psutil.AccessDenied):
                    pass

            # Mark as not paused
            _paused_processes[study_id] = False

            return {
                "success": True,
                "message": f"Optimization resumed (PID: {pid}, +{len(children)} children)",
                "pid": pid
            }

        except psutil.NoSuchProcess:
            _paused_processes[study_id] = False
            return {
                "success": False,
                "message": "Process no longer exists"
            }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to resume optimization: {str(e)}")


class ValidateRequest(BaseModel):
    topN: int = 5


@router.post("/studies/{study_id}/validate")
async def validate_optimization(study_id: str, request: ValidateRequest = None):
    """
    Run final FEA validation on top NN predictions

    Args:
        study_id: Study identifier
        request: Validation options (topN)

    Returns:
        JSON with process info
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Check if optimization is still running
        existing_proc = _find_optimization_process(study_id)
        if existing_proc:
            return {
                "success": False,
                "message": "Cannot validate while optimization is running. Stop optimization first."
            }

        # Look for final_validation.py script
        validation_script = study_dir / "final_validation.py"

        if not validation_script.exists():
            # Fall back to run_optimization.py with --validate flag if script doesn't exist
            run_script = study_dir / "run_optimization.py"
            if not run_script.exists():
                raise HTTPException(status_code=404, detail="No validation script found")

            python_exe = sys.executable
            top_n = request.topN if request else 5
            cmd = [python_exe, str(run_script), "--validate", "--top", str(top_n)]
        else:
            python_exe = sys.executable
            top_n = request.topN if request else 5
            cmd = [python_exe, str(validation_script), "--top", str(top_n)]

        # Start validation process
        proc = subprocess.Popen(
            cmd,
            cwd=str(study_dir),
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            start_new_session=True
        )

        return {
            "success": True,
            "message": f"Validation started for top {top_n} NN predictions",
            "pid": proc.pid,
            "command": ' '.join(cmd)
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to start validation: {str(e)}")


# ============================================================================
# Optuna Dashboard Launch
# ============================================================================

_optuna_processes: Dict[str, subprocess.Popen] = {}

@router.post("/studies/{study_id}/optuna-dashboard")
async def launch_optuna_dashboard(study_id: str):
    """
    Launch Optuna dashboard for a specific study

    Args:
        study_id: Study identifier

    Returns:
        JSON with dashboard URL and process info
    """
    import time
    import socket

    def is_port_in_use(port: int) -> bool:
        """Check if a port is already in use"""
        with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
            return s.connect_ex(('localhost', port)) == 0

    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        results_dir = get_results_dir(study_dir)
        study_db = results_dir / "study.db"

        if not study_db.exists():
            raise HTTPException(status_code=404, detail=f"No Optuna database found for study {study_id}")

        port = 8081

        # Check if dashboard is already running on this port
        if is_port_in_use(port):
            # Check if it's our process
            if study_id in _optuna_processes:
                proc = _optuna_processes[study_id]
                if proc.poll() is None:  # Still running
                    return {
                        "success": True,
                        "url": f"http://localhost:{port}",
                        "pid": proc.pid,
                        "message": "Optuna dashboard already running"
                    }
            # Port in use but not by us - still return success since dashboard is available
            return {
                "success": True,
                "url": f"http://localhost:{port}",
                "pid": None,
                "message": "Optuna dashboard already running on port 8081"
            }

        # Launch optuna-dashboard using CLI command (more robust than Python import)
        # Use absolute path with POSIX format for SQLite URL
        abs_db_path = study_db.absolute().as_posix()
        storage_url = f"sqlite:///{abs_db_path}"

        # Find optuna-dashboard executable - use full path for conda environment
        import platform
        import shutil as sh

        optuna_dashboard_cmd = "optuna-dashboard"

        if platform.system() == 'Windows':
            # Try to find optuna-dashboard in the conda environment
            conda_paths = [
                Path(r"C:\Users\antoi\anaconda3\envs\atomizer\Scripts\optuna-dashboard.exe"),
                Path(r"C:\Users\antoi\anaconda3\envs\atomizer\Library\bin\optuna-dashboard.exe"),
            ]
            for conda_path in conda_paths:
                if conda_path.exists():
                    optuna_dashboard_cmd = str(conda_path)
                    break
            else:
                # Fallback: check if it's in PATH
                found_path = sh.which("optuna-dashboard")
                if found_path:
                    optuna_dashboard_cmd = found_path
                else:
                    # Return helpful error message
                    raise HTTPException(
                        status_code=500,
                        detail="optuna-dashboard not found. Install with: pip install optuna-dashboard"
                    )

        cmd = [optuna_dashboard_cmd, storage_url, "--port", str(port), "--host", "0.0.0.0"]

        # On Windows, use CREATE_NEW_PROCESS_GROUP and DETACHED_PROCESS flags
        if platform.system() == 'Windows':
            # Windows-specific: create detached process
            DETACHED_PROCESS = 0x00000008
            CREATE_NEW_PROCESS_GROUP = 0x00000200
            proc = subprocess.Popen(
                cmd,
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
                creationflags=DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP
            )
        else:
            proc = subprocess.Popen(
                cmd,
                stdout=subprocess.PIPE,
                stderr=subprocess.PIPE,
                start_new_session=True
            )

        _optuna_processes[study_id] = proc

        # Wait for dashboard to start (check port repeatedly)
        max_wait = 5  # seconds
        start_time = time.time()
        while time.time() - start_time < max_wait:
            if is_port_in_use(port):
                return {
                    "success": True,
                    "url": f"http://localhost:{port}",
                    "pid": proc.pid,
                    "message": "Optuna dashboard launched successfully"
                }
            # Check if process died
            if proc.poll() is not None:
                stderr = ""
                try:
                    stderr = proc.stderr.read().decode() if proc.stderr else ""
                except:
                    pass
                return {
                    "success": False,
                    "message": f"Failed to start Optuna dashboard: {stderr}"
                }
            time.sleep(0.5)

        # Timeout - process might still be starting
        if proc.poll() is None:
            return {
                "success": True,
                "url": f"http://localhost:{port}",
                "pid": proc.pid,
                "message": "Optuna dashboard starting (may take a moment)"
            }
        else:
            stderr = ""
            try:
                stderr = proc.stderr.read().decode() if proc.stderr else ""
            except:
                pass
            return {
                "success": False,
                "message": f"Failed to start Optuna dashboard: {stderr}"
            }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to launch Optuna dashboard: {str(e)}")


@router.get("/optuna-dashboard/check")
async def check_optuna_dashboard_available():
    """
    Check if optuna-dashboard is installed and available.

    Returns:
        JSON with availability status and installation instructions if not available.
    """
    import platform
    import shutil as sh

    if platform.system() == 'Windows':
        # Check conda environment paths
        conda_paths = [
            Path(r"C:\Users\antoi\anaconda3\envs\atomizer\Scripts\optuna-dashboard.exe"),
            Path(r"C:\Users\antoi\anaconda3\envs\atomizer\Library\bin\optuna-dashboard.exe"),
        ]
        for conda_path in conda_paths:
            if conda_path.exists():
                return {
                    "available": True,
                    "path": str(conda_path),
                    "message": "optuna-dashboard is installed"
                }

    # Fallback: check system PATH
    found_path = sh.which("optuna-dashboard")
    if found_path:
        return {
            "available": True,
            "path": found_path,
            "message": "optuna-dashboard is installed"
        }

    return {
        "available": False,
        "path": None,
        "message": "optuna-dashboard not found",
        "install_instructions": "pip install optuna-dashboard"
    }


# ============================================================================
# Model Files Endpoint
# ============================================================================

@router.get("/studies/{study_id}/model-files")
async def get_model_files(study_id: str):
    """
    Get list of NX model files (.prt, .sim, .fem, .bdf, .dat, .op2) for a study

    Args:
        study_id: Study identifier

    Returns:
        JSON with list of model files and their paths
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Look for model directory (check multiple locations)
        model_dirs = [
            study_dir / "1_setup" / "model",
            study_dir / "model",
            study_dir / "1_setup",
            study_dir
        ]

        model_files = []
        model_dir_path = None

        # NX and FEA file extensions to look for
        nx_extensions = {'.prt', '.sim', '.fem', '.bdf', '.dat', '.op2', '.f06', '.inp'}

        for model_dir in model_dirs:
            if model_dir.exists() and model_dir.is_dir():
                for file_path in model_dir.iterdir():
                    if file_path.is_file() and file_path.suffix.lower() in nx_extensions:
                        model_files.append({
                            "name": file_path.name,
                            "path": str(file_path),
                            "extension": file_path.suffix.lower(),
                            "size_bytes": file_path.stat().st_size,
                            "size_display": _format_file_size(file_path.stat().st_size),
                            "modified": datetime.fromtimestamp(file_path.stat().st_mtime).isoformat()
                        })
                        if model_dir_path is None:
                            model_dir_path = str(model_dir)

        # Sort by extension for better display (prt first, then sim, fem, etc.)
        extension_order = {'.prt': 0, '.sim': 1, '.fem': 2, '.bdf': 3, '.dat': 4, '.op2': 5, '.f06': 6, '.inp': 7}
        model_files.sort(key=lambda x: (extension_order.get(x['extension'], 99), x['name']))

        return {
            "study_id": study_id,
            "model_dir": model_dir_path or str(study_dir / "1_setup" / "model"),
            "files": model_files,
            "count": len(model_files)
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get model files: {str(e)}")


def _format_file_size(size_bytes: int) -> str:
    """Format file size in human-readable form"""
    if size_bytes < 1024:
        return f"{size_bytes} B"
    elif size_bytes < 1024 * 1024:
        return f"{size_bytes / 1024:.1f} KB"
    elif size_bytes < 1024 * 1024 * 1024:
        return f"{size_bytes / (1024 * 1024):.1f} MB"
    else:
        return f"{size_bytes / (1024 * 1024 * 1024):.2f} GB"


@router.post("/studies/{study_id}/open-folder")
async def open_model_folder(study_id: str, folder_type: str = "model"):
    """
    Open the model folder in system file explorer

    Args:
        study_id: Study identifier
        folder_type: Type of folder to open (model, results, setup)

    Returns:
        JSON with success status
    """
    import os
    import platform

    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Determine which folder to open
        if folder_type == "model":
            target_dir = study_dir / "1_setup" / "model"
            if not target_dir.exists():
                target_dir = study_dir / "1_setup"
        elif folder_type == "results":
            target_dir = get_results_dir(study_dir)
        elif folder_type == "setup":
            target_dir = study_dir / "1_setup"
        else:
            target_dir = study_dir

        if not target_dir.exists():
            target_dir = study_dir

        # Open in file explorer based on platform
        system = platform.system()
        try:
            if system == "Windows":
                os.startfile(str(target_dir))
            elif system == "Darwin":  # macOS
                subprocess.Popen(["open", str(target_dir)])
            else:  # Linux
                subprocess.Popen(["xdg-open", str(target_dir)])

            return {
                "success": True,
                "message": f"Opened {target_dir}",
                "path": str(target_dir)
            }
        except Exception as e:
            return {
                "success": False,
                "message": f"Failed to open folder: {str(e)}",
                "path": str(target_dir)
            }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to open folder: {str(e)}")


@router.get("/studies/{study_id}/best-solution")
async def get_best_solution(study_id: str):
    """Get the best trial(s) for a study with improvement metrics"""
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study '{study_id}' not found")

        results_dir = get_results_dir(study_dir)
        db_path = results_dir / "study.db"

        if not db_path.exists():
            return {
                "study_id": study_id,
                "best_trial": None,
                "first_trial": None,
                "improvements": {},
                "total_trials": 0
            }

        conn = sqlite3.connect(str(db_path))
        conn.row_factory = sqlite3.Row
        cursor = conn.cursor()

        # Get best trial (single objective - minimize by default)
        cursor.execute("""
            SELECT t.trial_id, t.number, tv.value as objective,
                   datetime(tv.value_id, 'unixepoch') as timestamp
            FROM trials t
            JOIN trial_values tv ON t.trial_id = tv.trial_id
            WHERE t.state = 'COMPLETE'
            ORDER BY tv.value ASC
            LIMIT 1
        """)
        best_row = cursor.fetchone()

        # Get first completed trial for comparison
        cursor.execute("""
            SELECT t.trial_id, t.number, tv.value as objective
            FROM trials t
            JOIN trial_values tv ON t.trial_id = tv.trial_id
            WHERE t.state = 'COMPLETE'
            ORDER BY t.number ASC
            LIMIT 1
        """)
        first_row = cursor.fetchone()

        # Get total trial count
        cursor.execute("SELECT COUNT(*) FROM trials WHERE state = 'COMPLETE'")
        total_trials = cursor.fetchone()[0]

        best_trial = None
        first_trial = None
        improvements = {}

        if best_row:
            best_trial_id = best_row['trial_id']

            # Get design variables
            cursor.execute("""
                SELECT param_name, param_value
                FROM trial_params
                WHERE trial_id = ?
            """, (best_trial_id,))
            params = {row['param_name']: row['param_value'] for row in cursor.fetchall()}

            # Get user attributes (including results)
            cursor.execute("""
                SELECT key, value_json
                FROM trial_user_attributes
                WHERE trial_id = ?
            """, (best_trial_id,))
            user_attrs = {}
            for row in cursor.fetchall():
                try:
                    user_attrs[row['key']] = json.loads(row['value_json'])
                except:
                    user_attrs[row['key']] = row['value_json']

            best_trial = {
                "trial_number": best_row['number'],
                "objective": best_row['objective'],
                "design_variables": params,
                "user_attrs": user_attrs,
                "timestamp": best_row['timestamp']
            }

        if first_row:
            first_trial_id = first_row['trial_id']

            cursor.execute("""
                SELECT param_name, param_value
                FROM trial_params
                WHERE trial_id = ?
            """, (first_trial_id,))
            first_params = {row['param_name']: row['param_value'] for row in cursor.fetchall()}

            first_trial = {
                "trial_number": first_row['number'],
                "objective": first_row['objective'],
                "design_variables": first_params
            }

            # Calculate improvement
            if best_row and first_row['objective'] != 0:
                improvement_pct = ((first_row['objective'] - best_row['objective']) / abs(first_row['objective'])) * 100
                improvements["objective"] = {
                    "initial": first_row['objective'],
                    "final": best_row['objective'],
                    "improvement_pct": round(improvement_pct, 2),
                    "absolute_change": round(first_row['objective'] - best_row['objective'], 6)
                }

        conn.close()

        return {
            "study_id": study_id,
            "best_trial": best_trial,
            "first_trial": first_trial,
            "improvements": improvements,
            "total_trials": total_trials
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get best solution: {str(e)}")


@router.get("/studies/{study_id}/runs")
async def get_study_runs(study_id: str):
    """
    Get all optimization runs/studies in the database for comparison.
    Many studies have multiple Optuna studies (e.g., v11_fea, v11_iter1_nn, v11_iter2_nn).
    This endpoint returns metrics for each sub-study.
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study '{study_id}' not found")

        results_dir = get_results_dir(study_dir)
        db_path = results_dir / "study.db"

        if not db_path.exists():
            return {"runs": [], "total_runs": 0}

        conn = sqlite3.connect(str(db_path))
        conn.row_factory = sqlite3.Row
        cursor = conn.cursor()

        # Get all Optuna studies in this database
        cursor.execute("""
            SELECT study_id, study_name
            FROM studies
            ORDER BY study_id
        """)
        studies = cursor.fetchall()

        runs = []
        for study_row in studies:
            optuna_study_id = study_row['study_id']
            study_name = study_row['study_name']

            # Get trial count
            cursor.execute("""
                SELECT COUNT(*) FROM trials
                WHERE study_id = ? AND state = 'COMPLETE'
            """, (optuna_study_id,))
            trial_count = cursor.fetchone()[0]

            if trial_count == 0:
                continue

            # Get best value (first objective)
            cursor.execute("""
                SELECT MIN(tv.value) as best_value
                FROM trial_values tv
                JOIN trials t ON tv.trial_id = t.trial_id
                WHERE t.study_id = ? AND t.state = 'COMPLETE' AND tv.objective = 0
            """, (optuna_study_id,))
            best_result = cursor.fetchone()
            best_value = best_result['best_value'] if best_result else None

            # Get average value
            cursor.execute("""
                SELECT AVG(tv.value) as avg_value
                FROM trial_values tv
                JOIN trials t ON tv.trial_id = t.trial_id
                WHERE t.study_id = ? AND t.state = 'COMPLETE' AND tv.objective = 0
            """, (optuna_study_id,))
            avg_result = cursor.fetchone()
            avg_value = avg_result['avg_value'] if avg_result else None

            # Get time range
            cursor.execute("""
                SELECT MIN(datetime_start) as first_trial, MAX(datetime_complete) as last_trial
                FROM trials
                WHERE study_id = ? AND state = 'COMPLETE'
            """, (optuna_study_id,))
            time_result = cursor.fetchone()

            # Determine source type (FEA or NN)
            source = "NN" if "_nn" in study_name.lower() else "FEA"

            runs.append({
                "run_id": optuna_study_id,
                "name": study_name,
                "source": source,
                "trial_count": trial_count,
                "best_value": best_value,
                "avg_value": avg_value,
                "first_trial": time_result['first_trial'] if time_result else None,
                "last_trial": time_result['last_trial'] if time_result else None
            })

        conn.close()

        return {
            "runs": runs,
            "total_runs": len(runs),
            "study_id": study_id
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get runs: {str(e)}")


class UpdateConfigRequest(BaseModel):
    config: dict


@router.put("/studies/{study_id}/config")
async def update_study_config(study_id: str, request: UpdateConfigRequest):
    """
    Update the optimization_config.json for a study

    Args:
        study_id: Study identifier
        request: New configuration data

    Returns:
        JSON with success status
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Check if optimization is running - don't allow config changes while running
        if is_optimization_running(study_id):
            raise HTTPException(
                status_code=409,
                detail="Cannot modify config while optimization is running. Stop the optimization first."
            )

        # Find config file location
        config_file = study_dir / "1_setup" / "optimization_config.json"
        if not config_file.exists():
            config_file = study_dir / "optimization_config.json"

        if not config_file.exists():
            raise HTTPException(status_code=404, detail=f"Config file not found for study {study_id}")

        # Backup existing config
        backup_file = config_file.with_suffix('.json.backup')
        shutil.copy(config_file, backup_file)

        # Write new config
        with open(config_file, 'w') as f:
            json.dump(request.config, f, indent=2)

        return {
            "success": True,
            "message": "Configuration updated successfully",
            "path": str(config_file),
            "backup_path": str(backup_file)
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to update config: {str(e)}")


# ============================================================================
# Zernike Analysis Endpoints
# ============================================================================

@router.get("/studies/{study_id}/zernike-available")
async def get_zernike_available_trials(study_id: str):
    """
    Get list of trial numbers that have Zernike analysis available (OP2 files).

    Returns:
        JSON with list of trial numbers that have iteration folders with OP2 files
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study '{study_id}' not found")

        iter_base = study_dir / "2_iterations"
        if not iter_base.exists():
            return {"study_id": study_id, "available_trials": [], "count": 0}

        available_trials = []
        for d in iter_base.iterdir():
            if d.is_dir() and d.name.startswith('iter'):
                # Check for OP2 file
                op2_files = list(d.glob("*.op2"))
                if op2_files:
                    iter_num_str = d.name.replace('iter', '')
                    try:
                        iter_num = int(iter_num_str)
                        # Map iter number to trial number (iter1 -> trial 0, etc.)
                        # But also keep iter_num as possibility
                        if iter_num != 9999:
                            available_trials.append(iter_num - 1)  # 0-indexed trial
                        else:
                            available_trials.append(9999)  # Special test iteration
                    except ValueError:
                        pass

        available_trials.sort()
        return {
            "study_id": study_id,
            "available_trials": available_trials,
            "count": len(available_trials)
        }

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get available trials: {str(e)}")


@router.get("/studies/{study_id}/trials/{trial_number}/zernike")
async def get_trial_zernike(study_id: str, trial_number: int):
    """
    Generate or retrieve Zernike analysis HTML for a specific trial.

    This endpoint generates interactive Zernike wavefront analysis for mirror
    optimization trials. It produces 3D surface residual plots, RMS metrics,
    and coefficient bar charts for each angle comparison (40_vs_20, 60_vs_20, 90_vs_20).

    Args:
        study_id: Study identifier
        trial_number: Trial/iteration number

    Returns:
        JSON with HTML content for each comparison, or error if OP2 not found
    """
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study '{study_id}' not found")

        # Find iteration directory
        # Trial numbers in Optuna DB may differ from iteration folder numbers
        # Common patterns:
        # 1. iter{trial_number} - direct mapping
        # 2. iter{trial_number + 1} - 0-indexed trials vs 1-indexed folders
        # 3. Check for actual folder existence

        iter_dir = None
        possible_iter_nums = [trial_number, trial_number + 1]

        for iter_num in possible_iter_nums:
            candidate = study_dir / "2_iterations" / f"iter{iter_num}"
            if candidate.exists():
                iter_dir = candidate
                break

        if iter_dir is None:
            raise HTTPException(
                status_code=404,
                detail=f"No FEA results for trial {trial_number}. This trial may have used surrogate model (NN) prediction instead of full FEA simulation. Zernike analysis requires OP2 results from actual FEA runs."
            )

        # Check for OP2 file BEFORE doing expensive imports
        op2_files = list(iter_dir.glob("*.op2"))
        if not op2_files:
            raise HTTPException(
                status_code=404,
                detail=f"No OP2 results file found in {iter_dir.name}. FEA may not have completed."
            )

        # Only import heavy dependencies after we know we have an OP2 file
        sys.path.append(str(Path(__file__).parent.parent.parent.parent.parent))
        from optimization_engine.extractors.extract_zernike_figure import ZernikeOPDExtractor
        from optimization_engine.extractors import ZernikeExtractor
        import numpy as np
        from math import factorial
        import plotly.graph_objects as go
        from plotly.subplots import make_subplots
        from matplotlib.tri import Triangulation

        # Also find BDF/DAT geometry file for OPD extractor
        bdf_files = list(iter_dir.glob("*.dat")) + list(iter_dir.glob("*.bdf"))
        bdf_path = bdf_files[0] if bdf_files else None

        # Configuration
        N_MODES = 50
        AMP = 0.5  # Reduced deformation scaling (0.5x)
        PANCAKE = 3.0  # Z-axis range multiplier
        PLOT_DOWNSAMPLE = 5000  # Reduced for faster loading
        FILTER_LOW_ORDERS = 4
        COLORSCALE = 'Turbo'  # Colorscale: 'RdBu_r', 'Viridis', 'Plasma', 'Turbo'

        SUBCASE_MAP = {
            '1': '90', '2': '20', '3': '40', '4': '60',
        }
        REF_SUBCASE = '2'

        def noll_indices(j: int):
            if j < 1:
                raise ValueError("Noll index j must be >= 1")
            count = 0
            n = 0
            while True:
                if n == 0:
                    ms = [0]
                elif n % 2 == 0:
                    ms = [0] + [m for k in range(1, n//2 + 1) for m in (-2*k, 2*k)]
                else:
                    ms = [m for k in range(0, (n+1)//2) for m in (-(2*k+1), (2*k+1))]
                for m in ms:
                    count += 1
                    if count == j:
                        return n, m
                n += 1

        def zernike_noll(j: int, r: np.ndarray, th: np.ndarray) -> np.ndarray:
            n, m = noll_indices(j)
            R = np.zeros_like(r)
            for s in range((n-abs(m))//2 + 1):
                c = ((-1)**s * factorial(n-s) /
                     (factorial(s) *
                      factorial((n+abs(m))//2 - s) *
                      factorial((n-abs(m))//2 - s)))
                R += c * r**(n-2*s)
            if m == 0:
                return R
            return R * (np.cos(m*th) if m > 0 else np.sin(-m*th))

        def zernike_common_name(n: int, m: int) -> str:
            names = {
                (0, 0): "Piston", (1, -1): "Tilt X", (1, 1): "Tilt Y",
                (2, 0): "Defocus", (2, -2): "Astig 45°", (2, 2): "Astig 0°",
                (3, -1): "Coma X", (3, 1): "Coma Y", (3, -3): "Trefoil X", (3, 3): "Trefoil Y",
                (4, 0): "Primary Spherical", (4, -2): "Sec Astig X", (4, 2): "Sec Astig Y",
                (4, -4): "Quadrafoil X", (4, 4): "Quadrafoil Y",
                (5, -1): "Sec Coma X", (5, 1): "Sec Coma Y",
                (5, -3): "Sec Trefoil X", (5, 3): "Sec Trefoil Y",
                (5, -5): "Pentafoil X", (5, 5): "Pentafoil Y",
                (6, 0): "Sec Spherical",
            }
            return names.get((n, m), f"Z(n={n}, m={m})")

        def zernike_label(j: int) -> str:
            n, m = noll_indices(j)
            return f"J{j:02d} - {zernike_common_name(n, m)}"

        def compute_manufacturing_metrics(coefficients: np.ndarray) -> dict:
            """Compute manufacturing-related aberration metrics."""
            return {
                'defocus_nm': float(abs(coefficients[3])),  # J4
                'astigmatism_rms': float(np.sqrt(coefficients[4]**2 + coefficients[5]**2)),  # J5+J6
                'coma_rms': float(np.sqrt(coefficients[6]**2 + coefficients[7]**2)),  # J7+J8
                'trefoil_rms': float(np.sqrt(coefficients[8]**2 + coefficients[9]**2)),  # J9+J10
                'spherical_nm': float(abs(coefficients[10])) if len(coefficients) > 10 else 0.0,  # J11
            }

        def compute_rms_filter_j1to3(X, Y, W_nm, coefficients, R):
            """Compute RMS with J1-J3 filtered (keeping defocus for optician workload)."""
            Xc = X - np.mean(X)
            Yc = Y - np.mean(Y)
            r = np.hypot(Xc/R, Yc/R)
            th = np.arctan2(Yc, Xc)
            Z_j1to3 = np.column_stack([zernike_noll(j, r, th) for j in range(1, 4)])
            W_filter_j1to3 = W_nm - Z_j1to3 @ coefficients[:3]
            return float(np.sqrt(np.mean(W_filter_j1to3**2)))

        def generate_zernike_html(
            title: str,
            X: np.ndarray,
            Y: np.ndarray,
            W_nm: np.ndarray,
            coefficients: np.ndarray,
            rms_global: float,
            rms_filtered: float,
            ref_title: str = "20 deg",
            abs_pair = None,
            is_manufacturing: bool = False,
            mfg_metrics: dict = None,
            correction_metrics: dict = None
        ) -> str:
            """Generate HTML string for Zernike visualization with full tables."""
            # Compute residual surface (filtered)
            Xc = X - np.mean(X)
            Yc = Y - np.mean(Y)
            R = float(np.max(np.hypot(Xc, Yc)))
            r = np.hypot(Xc/R, Yc/R)
            th = np.arctan2(Yc, Xc)

            Z = np.column_stack([zernike_noll(j, r, th) for j in range(1, N_MODES+1)])
            W_res_filt = W_nm - Z[:, :FILTER_LOW_ORDERS].dot(coefficients[:FILTER_LOW_ORDERS])

            # Compute J1-J3 filtered RMS (optician workload metric)
            rms_filter_j1to3 = compute_rms_filter_j1to3(X, Y, W_nm, coefficients, R)

            # Downsample for display
            n = len(X)
            if n > PLOT_DOWNSAMPLE:
                rng = np.random.default_rng(42)
                sel = rng.choice(n, size=PLOT_DOWNSAMPLE, replace=False)
                Xp, Yp, Wp = X[sel], Y[sel], W_res_filt[sel]
            else:
                Xp, Yp, Wp = X, Y, W_res_filt

            res_amp = AMP * Wp
            max_amp = float(np.max(np.abs(res_amp))) if res_amp.size else 1.0

            # Create smooth shaded SURFACE mesh with lighting
            surface_trace = None
            try:
                tri = Triangulation(Xp, Yp)
                if tri.triangles is not None and len(tri.triangles) > 0:
                    i_idx, j_idx, k_idx = tri.triangles.T
                    surface_trace = go.Mesh3d(
                        x=Xp.tolist(), y=Yp.tolist(), z=res_amp.tolist(),
                        i=i_idx.tolist(), j=j_idx.tolist(), k=k_idx.tolist(),
                        intensity=res_amp.tolist(),
                        colorscale=COLORSCALE,
                        opacity=1.0,
                        flatshading=False,  # Smooth shading
                        lighting=dict(
                            ambient=0.4,
                            diffuse=0.8,
                            specular=0.3,
                            roughness=0.5,
                            fresnel=0.2
                        ),
                        lightposition=dict(x=100, y=200, z=300),
                        showscale=True,
                        colorbar=dict(
                            title=dict(text="Residual (nm)", side='right'),
                            thickness=15,
                            len=0.6,
                            tickformat=".1f"
                        ),
                        hovertemplate="X: %{x:.1f}<br>Y: %{y:.1f}<br>Residual: %{z:.2f} nm<extra></extra>"
                    )
            except Exception as e:
                print(f"Triangulation failed: {e}")

            labels = [zernike_label(j) for j in range(1, N_MODES+1)]
            coeff_abs = np.abs(coefficients)
            mfg = compute_manufacturing_metrics(coefficients)

            # Determine layout based on whether this is manufacturing (90 deg) view
            if is_manufacturing and mfg_metrics and correction_metrics:
                # Manufacturing view: 5 rows
                fig = make_subplots(
                    rows=5, cols=1,
                    specs=[[{"type": "scene"}],
                           [{"type": "table"}],
                           [{"type": "table"}],
                           [{"type": "table"}],
                           [{"type": "xy"}]],
                    row_heights=[0.35, 0.10, 0.15, 0.15, 0.25],
                    vertical_spacing=0.025,
                    subplot_titles=[
                        f"<b>Surface Residual (relative to {ref_title})</b>",
                        "<b>RMS Metrics</b>",
                        "<b>Mode Magnitudes (Absolute 90 deg)</b>",
                        "<b>Pre-Correction (90 deg - 20 deg)</b>",
                        f"<b>Zernike Coefficients ({N_MODES} modes)</b>"
                    ]
                )
            else:
                # Standard relative view: 4 rows with full coefficient table
                fig = make_subplots(
                    rows=4, cols=1,
                    specs=[[{"type": "scene"}],
                           [{"type": "table"}],
                           [{"type": "table"}],
                           [{"type": "xy"}]],
                    row_heights=[0.40, 0.12, 0.28, 0.20],
                    vertical_spacing=0.03,
                    subplot_titles=[
                        f"<b>Surface Residual (relative to {ref_title})</b>",
                        "<b>RMS Metrics</b>",
                        f"<b>Zernike Coefficients ({N_MODES} modes)</b>",
                        "<b>Top 20 |Zernike Coefficients| (nm)</b>"
                    ]
                )

            # Add surface mesh (or fallback to scatter)
            if surface_trace is not None:
                fig.add_trace(surface_trace, row=1, col=1)
            else:
                # Fallback to scatter if triangulation failed
                fig.add_trace(go.Scatter3d(
                    x=Xp.tolist(), y=Yp.tolist(), z=res_amp.tolist(),
                    mode='markers',
                    marker=dict(size=2, color=res_amp.tolist(), colorscale=COLORSCALE, showscale=True),
                    showlegend=False
                ), row=1, col=1)

            fig.update_scenes(
                camera=dict(
                    eye=dict(x=1.2, y=1.2, z=0.8),
                    up=dict(x=0, y=0, z=1)
                ),
                xaxis=dict(
                    title="X (mm)",
                    showgrid=True,
                    gridcolor='rgba(128,128,128,0.3)',
                    showbackground=True,
                    backgroundcolor='rgba(240,240,240,0.9)'
                ),
                yaxis=dict(
                    title="Y (mm)",
                    showgrid=True,
                    gridcolor='rgba(128,128,128,0.3)',
                    showbackground=True,
                    backgroundcolor='rgba(240,240,240,0.9)'
                ),
                zaxis=dict(
                    title="Residual (nm)",
                    range=[-max_amp * PANCAKE, max_amp * PANCAKE],
                    showgrid=True,
                    gridcolor='rgba(128,128,128,0.3)',
                    showbackground=True,
                    backgroundcolor='rgba(230,230,250,0.9)'
                ),
                aspectmode='manual',
                aspectratio=dict(x=1, y=1, z=0.4)
            )

            # Row 2: RMS table with all metrics
            if abs_pair is not None:
                abs_global, abs_filtered = abs_pair
                fig.add_trace(go.Table(
                    header=dict(
                        values=["<b>Metric</b>", "<b>Relative (nm)</b>", "<b>Absolute (nm)</b>"],
                        align="left",
                        fill_color='rgb(55, 83, 109)',
                        font=dict(color='white', size=12)
                    ),
                    cells=dict(
                        values=[
                            ["Global RMS", "Filtered RMS (J1-J4)", "Filtered RMS (J1-J3, w/ defocus)"],
                            [f"{rms_global:.2f}", f"{rms_filtered:.2f}", f"{rms_filter_j1to3:.2f}"],
                            [f"{abs_global:.2f}", f"{abs_filtered:.2f}", "-"],
                        ],
                        align="left",
                        fill_color='rgb(243, 243, 243)'
                    )
                ), row=2, col=1)
            else:
                fig.add_trace(go.Table(
                    header=dict(
                        values=["<b>Metric</b>", "<b>Value (nm)</b>"],
                        align="left",
                        fill_color='rgb(55, 83, 109)',
                        font=dict(color='white', size=12)
                    ),
                    cells=dict(
                        values=[
                            ["Global RMS", "Filtered RMS (J1-J4)", "Filtered RMS (J1-J3, w/ defocus)"],
                            [f"{rms_global:.2f}", f"{rms_filtered:.2f}", f"{rms_filter_j1to3:.2f}"]
                        ],
                        align="left",
                        fill_color='rgb(243, 243, 243)'
                    )
                ), row=2, col=1)

            if is_manufacturing and mfg_metrics and correction_metrics:
                # Row 3: Mode magnitudes at 90 deg (absolute)
                fig.add_trace(go.Table(
                    header=dict(
                        values=["<b>Mode</b>", "<b>Value (nm)</b>"],
                        align="left",
                        fill_color='rgb(55, 83, 109)',
                        font=dict(color='white', size=11)
                    ),
                    cells=dict(
                        values=[
                            ["Defocus (J4)", "Astigmatism (J5+J6)", "Coma (J7+J8)", "Trefoil (J9+J10)", "Spherical (J11)"],
                            [f"{mfg_metrics['defocus_nm']:.2f}", f"{mfg_metrics['astigmatism_rms']:.2f}",
                             f"{mfg_metrics['coma_rms']:.2f}", f"{mfg_metrics['trefoil_rms']:.2f}",
                             f"{mfg_metrics['spherical_nm']:.2f}"]
                        ],
                        align="left",
                        fill_color='rgb(243, 243, 243)'
                    )
                ), row=3, col=1)

                # Row 4: Pre-correction (90 deg - 20 deg)
                fig.add_trace(go.Table(
                    header=dict(
                        values=["<b>Correction Mode</b>", "<b>Value (nm)</b>"],
                        align="left",
                        fill_color='rgb(55, 83, 109)',
                        font=dict(color='white', size=11)
                    ),
                    cells=dict(
                        values=[
                            ["Total RMS (J1-J3 filter)", "Defocus (J4)", "Astigmatism (J5+J6)", "Coma (J7+J8)"],
                            [f"{correction_metrics.get('rms_filter_j1to3', 0):.2f}",
                             f"{correction_metrics['defocus_nm']:.2f}",
                             f"{correction_metrics['astigmatism_rms']:.2f}",
                             f"{correction_metrics['coma_rms']:.2f}"]
                        ],
                        align="left",
                        fill_color='rgb(243, 243, 243)'
                    )
                ), row=4, col=1)

                # Row 5: Bar chart
                sorted_idx = np.argsort(coeff_abs)[::-1][:20]
                fig.add_trace(
                    go.Bar(
                        x=[float(coeff_abs[i]) for i in sorted_idx],
                        y=[labels[i] for i in sorted_idx],
                        orientation='h',
                        marker_color='rgb(55, 83, 109)',
                        hovertemplate="%{y}<br>|Coeff| = %{x:.3f} nm<extra></extra>",
                        showlegend=False
                    ),
                    row=5, col=1
                )
            else:
                # Row 3: Full coefficient table
                fig.add_trace(go.Table(
                    header=dict(
                        values=["<b>Noll j</b>", "<b>Mode Name</b>", "<b>Coeff (nm)</b>", "<b>|Coeff| (nm)</b>"],
                        align="left",
                        fill_color='rgb(55, 83, 109)',
                        font=dict(color='white', size=11)
                    ),
                    cells=dict(
                        values=[
                            list(range(1, N_MODES+1)),
                            labels,
                            [f"{c:+.3f}" for c in coefficients],
                            [f"{abs(c):.3f}" for c in coefficients]
                        ],
                        align="left",
                        fill_color='rgb(243, 243, 243)',
                        font=dict(size=10),
                        height=22
                    )
                ), row=3, col=1)

                # Row 4: Bar chart - top 20 modes by magnitude
                sorted_idx = np.argsort(coeff_abs)[::-1][:20]
                fig.add_trace(
                    go.Bar(
                        x=[float(coeff_abs[i]) for i in sorted_idx],
                        y=[labels[i] for i in sorted_idx],
                        orientation='h',
                        marker_color='rgb(55, 83, 109)',
                        hovertemplate="%{y}<br>|Coeff| = %{x:.3f} nm<extra></extra>",
                        showlegend=False
                    ),
                    row=4, col=1
                )

            fig.update_layout(
                width=1400,
                height=1800 if is_manufacturing else 1600,
                margin=dict(t=80, b=20, l=20, r=20),
                title=dict(
                    text=f"<b>{title}</b>",
                    font=dict(size=20),
                    x=0.5
                ),
                paper_bgcolor='white',
                plot_bgcolor='white'
            )

            return fig.to_html(include_plotlyjs='cdn', full_html=True)

        # =====================================================================
        # NEW: Use OPD method (accounts for lateral X,Y displacement)
        # =====================================================================
        op2_path = op2_files[0]

        # Try OPD extractor first (more accurate), fall back to Standard if no BDF
        use_opd = bdf_path is not None
        if use_opd:
            try:
                opd_extractor = ZernikeOPDExtractor(
                    str(op2_path),
                    bdf_path=str(bdf_path),
                    n_modes=N_MODES,
                    filter_orders=FILTER_LOW_ORDERS
                )
            except Exception as e:
                print(f"OPD extractor failed, falling back to Standard: {e}")
                use_opd = False

        # Also create Standard extractor for comparison
        std_extractor = ZernikeExtractor(str(op2_path), displacement_unit='mm', n_modes=N_MODES)

        def generate_dual_method_html(
            title: str,
            target_sc: str,
            ref_sc: str,
            is_manufacturing: bool = False
        ) -> tuple:
            """Generate HTML with OPD method and displacement component views.

            Returns: (html_content, rms_global_opd, rms_filtered_opd, lateral_stats)
            """
            target_angle = SUBCASE_MAP.get(target_sc, target_sc)
            ref_angle = SUBCASE_MAP.get(ref_sc, ref_sc)

            # Extract using OPD method (primary)
            if use_opd:
                opd_rel = opd_extractor.extract_relative(target_sc, ref_sc)
                opd_abs = opd_extractor.extract_subcase(target_sc)
            else:
                opd_rel = None
                opd_abs = None

            # Extract using Standard method (for comparison)
            std_rel = std_extractor.extract_relative(target_sc, ref_sc, include_coefficients=True)
            std_abs = std_extractor.extract_subcase(target_sc, include_coefficients=True)

            # Get OPD data with full arrays for visualization
            if use_opd:
                opd_data = opd_extractor._build_figure_opd_data(target_sc)
                opd_ref_data = opd_extractor._build_figure_opd_data(ref_sc)

                # Build relative displacement arrays (node-by-node)
                ref_map = {int(nid): i for i, nid in enumerate(opd_ref_data['node_ids'])}
                X_list, Y_list, WFE_list = [], [], []
                dx_list, dy_list, dz_list = [], [], []

                for i, nid in enumerate(opd_data['node_ids']):
                    nid = int(nid)
                    if nid not in ref_map:
                        continue
                    ref_idx = ref_map[nid]

                    # Use deformed coordinates from OPD
                    X_list.append(opd_data['x_deformed'][i])
                    Y_list.append(opd_data['y_deformed'][i])
                    WFE_list.append(opd_data['wfe_nm'][i] - opd_ref_data['wfe_nm'][ref_idx])

                    # Relative displacements (target - reference)
                    dx_list.append(opd_data['dx'][i] - opd_ref_data['dx'][ref_idx])
                    dy_list.append(opd_data['dy'][i] - opd_ref_data['dy'][ref_idx])
                    dz_list.append(opd_data['dz'][i] - opd_ref_data['dz'][ref_idx])

                X = np.array(X_list)
                Y = np.array(Y_list)
                W = np.array(WFE_list)
                dx = np.array(dx_list) * 1000.0  # mm to µm
                dy = np.array(dy_list) * 1000.0
                dz = np.array(dz_list) * 1000.0

                # Lateral displacement magnitude
                lateral_um = np.sqrt(dx**2 + dy**2)
                max_lateral = float(np.max(np.abs(lateral_um)))
                rms_lateral = float(np.sqrt(np.mean(lateral_um**2)))

                rms_global_opd = opd_rel['relative_global_rms_nm']
                rms_filtered_opd = opd_rel['relative_filtered_rms_nm']
                coefficients = np.array(opd_rel.get('delta_coefficients', std_rel['coefficients']))
            else:
                # Fallback to Standard method arrays
                target_disp = std_extractor.displacements[target_sc]
                ref_disp = std_extractor.displacements[ref_sc]
                ref_map = {int(nid): i for i, nid in enumerate(ref_disp['node_ids'])}

                X_list, Y_list, W_list = [], [], []
                dx_list, dy_list, dz_list = [], [], []

                for i, nid in enumerate(target_disp['node_ids']):
                    nid = int(nid)
                    if nid not in ref_map:
                        continue
                    geo = std_extractor.node_geometry.get(nid)
                    if geo is None:
                        continue
                    ref_idx = ref_map[nid]

                    X_list.append(geo[0])
                    Y_list.append(geo[1])

                    target_wfe = target_disp['disp'][i, 2] * std_extractor.wfe_factor
                    ref_wfe = ref_disp['disp'][ref_idx, 2] * std_extractor.wfe_factor
                    W_list.append(target_wfe - ref_wfe)

                    # Relative displacements (mm to µm)
                    dx_list.append((target_disp['disp'][i, 0] - ref_disp['disp'][ref_idx, 0]) * 1000.0)
                    dy_list.append((target_disp['disp'][i, 1] - ref_disp['disp'][ref_idx, 1]) * 1000.0)
                    dz_list.append((target_disp['disp'][i, 2] - ref_disp['disp'][ref_idx, 2]) * 1000.0)

                X = np.array(X_list)
                Y = np.array(Y_list)
                W = np.array(W_list)
                dx = np.array(dx_list)
                dy = np.array(dy_list)
                dz = np.array(dz_list)

                lateral_um = np.sqrt(dx**2 + dy**2)
                max_lateral = float(np.max(np.abs(lateral_um)))
                rms_lateral = float(np.sqrt(np.mean(lateral_um**2)))

                rms_global_opd = std_rel['relative_global_rms_nm']
                rms_filtered_opd = std_rel['relative_filtered_rms_nm']
                coefficients = np.array(std_rel['coefficients'])

            # Standard method RMS values
            rms_global_std = std_rel['relative_global_rms_nm']
            rms_filtered_std = std_rel['relative_filtered_rms_nm']

            # Compute residual surface
            Xc = X - np.mean(X)
            Yc = Y - np.mean(Y)
            R = float(np.max(np.hypot(Xc, Yc)))
            r = np.hypot(Xc/R, Yc/R)
            th = np.arctan2(Yc, Xc)

            Z_basis = np.column_stack([zernike_noll(j, r, th) for j in range(1, N_MODES+1)])
            W_res_filt = W - Z_basis[:, :FILTER_LOW_ORDERS].dot(coefficients[:FILTER_LOW_ORDERS])

            # Downsample for display
            n = len(X)
            if n > PLOT_DOWNSAMPLE:
                rng = np.random.default_rng(42)
                sel = rng.choice(n, size=PLOT_DOWNSAMPLE, replace=False)
                Xp, Yp = X[sel], Y[sel]
                Wp = W_res_filt[sel]
                dxp, dyp, dzp = dx[sel], dy[sel], dz[sel]
            else:
                Xp, Yp, Wp = X, Y, W_res_filt
                dxp, dyp, dzp = dx, dy, dz

            res_amp = AMP * Wp
            max_amp = float(np.max(np.abs(res_amp))) if res_amp.size else 1.0

            # Helper to build mesh trace
            def build_mesh_trace(Zp, colorscale, colorbar_title, unit):
                try:
                    tri = Triangulation(Xp, Yp)
                    if tri.triangles is not None and len(tri.triangles) > 0:
                        i_idx, j_idx, k_idx = tri.triangles.T
                        return go.Mesh3d(
                            x=Xp.tolist(), y=Yp.tolist(), z=Zp.tolist(),
                            i=i_idx.tolist(), j=j_idx.tolist(), k=k_idx.tolist(),
                            intensity=Zp.tolist(),
                            colorscale=colorscale,
                            opacity=1.0,
                            flatshading=False,
                            lighting=dict(ambient=0.4, diffuse=0.8, specular=0.3, roughness=0.5, fresnel=0.2),
                            lightposition=dict(x=100, y=200, z=300),
                            showscale=True,
                            colorbar=dict(title=dict(text=colorbar_title, side='right'), thickness=15, len=0.5),
                            hovertemplate=f"X: %{{x:.1f}}<br>Y: %{{y:.1f}}<br>{unit}: %{{z:.3f}}<extra></extra>"
                        )
                except Exception:
                    pass
                return go.Scatter3d(
                    x=Xp.tolist(), y=Yp.tolist(), z=Zp.tolist(),
                    mode='markers', marker=dict(size=2, color=Zp.tolist(), colorscale=colorscale, showscale=True)
                )

            # Build traces for each view
            trace_wfe = build_mesh_trace(res_amp, COLORSCALE, "WFE (nm)", "WFE nm")
            trace_dx = build_mesh_trace(dxp, 'RdBu_r', "ΔX (µm)", "ΔX µm")
            trace_dy = build_mesh_trace(dyp, 'RdBu_r', "ΔY (µm)", "ΔY µm")
            trace_dz = build_mesh_trace(dzp, 'RdBu_r', "ΔZ (µm)", "ΔZ µm")

            # Create figure with dropdown to switch views
            fig = go.Figure()

            # Add all traces (only WFE visible initially)
            trace_wfe.visible = True
            trace_dx.visible = False
            trace_dy.visible = False
            trace_dz.visible = False

            fig.add_trace(trace_wfe)
            fig.add_trace(trace_dx)
            fig.add_trace(trace_dy)
            fig.add_trace(trace_dz)

            # Dropdown menu for view selection
            fig.update_layout(
                updatemenus=[
                    dict(
                        type="buttons",
                        direction="right",
                        x=0.0, y=1.12,
                        xanchor="left",
                        showactive=True,
                        buttons=[
                            dict(label="WFE (nm)", method="update",
                                 args=[{"visible": [True, False, False, False]}]),
                            dict(label="ΔX (µm)", method="update",
                                 args=[{"visible": [False, True, False, False]}]),
                            dict(label="ΔY (µm)", method="update",
                                 args=[{"visible": [False, False, True, False]}]),
                            dict(label="ΔZ (µm)", method="update",
                                 args=[{"visible": [False, False, False, True]}]),
                        ],
                        font=dict(size=12),
                        pad=dict(r=10, t=10),
                    )
                ]
            )

            # Compute method difference
            pct_diff = 100.0 * (rms_filtered_opd - rms_filtered_std) / rms_filtered_std if rms_filtered_std > 0 else 0.0

            # Annotations for metrics
            method_label = "OPD (X,Y,Z)" if use_opd else "Standard (Z-only)"
            annotations_text = f"""
<b>Method:</b> {method_label} {'← More Accurate' if use_opd else '(BDF not found)'}
<b>RMS Metrics (Filtered J1-J4):</b>
  • OPD:      {rms_filtered_opd:.2f} nm
  • Standard: {rms_filtered_std:.2f} nm
  • Δ:        {pct_diff:+.1f}%
<b>Lateral Displacement:</b>
  • Max: {max_lateral:.3f} µm
  • RMS: {rms_lateral:.3f} µm
<b>Displacement RMS:</b>
  • ΔX: {float(np.sqrt(np.mean(dx**2))):.3f} µm
  • ΔY: {float(np.sqrt(np.mean(dy**2))):.3f} µm
  • ΔZ: {float(np.sqrt(np.mean(dz**2))):.3f} µm
"""

            # Z-axis range for different views
            max_disp = max(float(np.max(np.abs(dxp))), float(np.max(np.abs(dyp))), float(np.max(np.abs(dzp))), 0.1)

            fig.update_layout(
                scene=dict(
                    camera=dict(eye=dict(x=1.2, y=1.2, z=0.8), up=dict(x=0, y=0, z=1)),
                    xaxis=dict(title="X (mm)", showgrid=True, gridcolor='rgba(128,128,128,0.3)',
                               showbackground=True, backgroundcolor='rgba(240,240,240,0.9)'),
                    yaxis=dict(title="Y (mm)", showgrid=True, gridcolor='rgba(128,128,128,0.3)',
                               showbackground=True, backgroundcolor='rgba(240,240,240,0.9)'),
                    zaxis=dict(title="Value", showgrid=True, gridcolor='rgba(128,128,128,0.3)',
                               showbackground=True, backgroundcolor='rgba(230,230,250,0.9)'),
                    aspectmode='manual',
                    aspectratio=dict(x=1, y=1, z=0.4)
                ),
                width=1400,
                height=900,
                margin=dict(t=120, b=20, l=20, r=20),
                title=dict(
                    text=f"<b>{title}</b><br><sup>Click buttons to switch: WFE, ΔX, ΔY, ΔZ</sup>",
                    font=dict(size=18),
                    x=0.5
                ),
                paper_bgcolor='white',
                plot_bgcolor='white',
                annotations=[
                    dict(
                        text=annotations_text.replace('\n', '<br>'),
                        xref="paper", yref="paper",
                        x=1.02, y=0.98,
                        xanchor="left", yanchor="top",
                        showarrow=False,
                        font=dict(family="monospace", size=11),
                        align="left",
                        bgcolor="rgba(255,255,255,0.9)",
                        bordercolor="rgba(0,0,0,0.3)",
                        borderwidth=1,
                        borderpad=8
                    ),
                    dict(
                        text="<b>View:</b>",
                        xref="paper", yref="paper",
                        x=0.0, y=1.15,
                        xanchor="left", yanchor="top",
                        showarrow=False,
                        font=dict(size=12)
                    )
                ]
            )

            html_content = fig.to_html(include_plotlyjs='cdn', full_html=True)

            return (html_content, rms_global_opd, rms_filtered_opd, {
                'max_lateral_um': max_lateral,
                'rms_lateral_um': rms_lateral,
                'method': 'opd' if use_opd else 'standard',
                'rms_std': rms_filtered_std,
                'pct_diff': pct_diff
            })

        # Generate results for each comparison
        results = {}
        comparisons = [
            ('3', '2', '40_vs_20', '40 deg vs 20 deg'),
            ('4', '2', '60_vs_20', '60 deg vs 20 deg'),
            ('1', '2', '90_vs_20', '90 deg vs 20 deg (manufacturing)'),
        ]

        for target_sc, ref_sc, key, title_suffix in comparisons:
            if target_sc not in std_extractor.displacements:
                continue

            target_angle = SUBCASE_MAP.get(target_sc, target_sc)
            ref_angle = SUBCASE_MAP.get(ref_sc, ref_sc)
            is_mfg = (key == '90_vs_20')

            html_content, rms_global, rms_filtered, lateral_stats = generate_dual_method_html(
                title=f"iter{trial_number}: {target_angle}° vs {ref_angle}°",
                target_sc=target_sc,
                ref_sc=ref_sc,
                is_manufacturing=is_mfg
            )

            results[key] = {
                "html": html_content,
                "rms_global": rms_global,
                "rms_filtered": rms_filtered,
                "title": f"{target_angle}° vs {ref_angle}°",
                "method": lateral_stats['method'],
                "rms_std": lateral_stats['rms_std'],
                "pct_diff": lateral_stats['pct_diff'],
                "max_lateral_um": lateral_stats['max_lateral_um'],
                "rms_lateral_um": lateral_stats['rms_lateral_um']
            }

        if not results:
            raise HTTPException(
                status_code=500,
                detail="Failed to generate Zernike analysis. Check if subcases are available."
            )

        return {
            "study_id": study_id,
            "trial_number": trial_number,
            "comparisons": results,
            "available_comparisons": list(results.keys()),
            "method": "opd" if use_opd else "standard"
        }

    except HTTPException:
        raise
    except Exception as e:
        import traceback
        traceback.print_exc()
        raise HTTPException(status_code=500, detail=f"Failed to generate Zernike analysis: {str(e)}")


@router.get("/studies/{study_id}/export/{format}")
async def export_study_data(study_id: str, format: str):
    """Export study data in various formats: csv, json, excel"""
    try:
        study_dir = resolve_study_path(study_id)
        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study '{study_id}' not found")

        results_dir = get_results_dir(study_dir)
        db_path = results_dir / "study.db"

        if not db_path.exists():
            raise HTTPException(status_code=404, detail="No study data available")

        conn = sqlite3.connect(str(db_path))
        conn.row_factory = sqlite3.Row
        cursor = conn.cursor()

        # Get all completed trials with their params and values
        cursor.execute("""
            SELECT t.trial_id, t.number, tv.value as objective
            FROM trials t
            JOIN trial_values tv ON t.trial_id = tv.trial_id
            WHERE t.state = 'COMPLETE'
            ORDER BY t.number
        """)
        trials_data = []

        for row in cursor.fetchall():
            trial_id = row['trial_id']

            # Get params
            cursor.execute("""
                SELECT param_name, param_value
                FROM trial_params
                WHERE trial_id = ?
            """, (trial_id,))
            params = {r['param_name']: r['param_value'] for r in cursor.fetchall()}

            # Get user attrs
            cursor.execute("""
                SELECT key, value_json
                FROM trial_user_attributes
                WHERE trial_id = ?
            """, (trial_id,))
            user_attrs = {}
            for r in cursor.fetchall():
                try:
                    user_attrs[r['key']] = json.loads(r['value_json'])
                except:
                    user_attrs[r['key']] = r['value_json']

            trials_data.append({
                "trial_number": row['number'],
                "objective": row['objective'],
                "params": params,
                "user_attrs": user_attrs
            })

        conn.close()

        if format.lower() == "json":
            return JSONResponse(content={
                "study_id": study_id,
                "total_trials": len(trials_data),
                "trials": trials_data
            })

        elif format.lower() == "csv":
            import io
            import csv

            if not trials_data:
                return JSONResponse(content={"error": "No data to export"})

            # Build CSV
            output = io.StringIO()

            # Get all param names
            param_names = sorted(set(
                key for trial in trials_data
                for key in trial['params'].keys()
            ))

            fieldnames = ['trial_number', 'objective'] + param_names
            writer = csv.DictWriter(output, fieldnames=fieldnames)
            writer.writeheader()

            for trial in trials_data:
                row_data = {
                    'trial_number': trial['trial_number'],
                    'objective': trial['objective']
                }
                row_data.update(trial['params'])
                writer.writerow(row_data)

            csv_content = output.getvalue()

            return JSONResponse(content={
                "filename": f"{study_id}_data.csv",
                "content": csv_content,
                "content_type": "text/csv"
            })

        elif format.lower() == "config":
            # Export optimization config
            setup_dir = study_dir / "1_setup"
            config_path = setup_dir / "optimization_config.json"

            if config_path.exists():
                with open(config_path, 'r') as f:
                    config = json.load(f)
                return JSONResponse(content={
                    "filename": f"{study_id}_config.json",
                    "content": json.dumps(config, indent=2),
                    "content_type": "application/json"
                })
            else:
                raise HTTPException(status_code=404, detail="Config file not found")

        else:
            raise HTTPException(status_code=400, detail=f"Unsupported format: {format}")

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to export data: {str(e)}")


# ============================================================================
# NX Model Introspection Endpoints
# ============================================================================

@router.get("/studies/{study_id}/nx/introspect")
async def introspect_nx_model(study_id: str, force: bool = False):
    """
    Introspect NX model to extract expressions and properties.

    Uses the optimization_engine.extractors.introspect_part module to run
    NX journal that extracts expressions, mass properties, materials, etc.

    Args:
        study_id: Study identifier
        force: Force re-introspection even if cached results exist

    Returns:
        JSON with introspection results including expressions list
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Find model directory
        model_dir = study_dir / "1_model"
        if not model_dir.exists():
            model_dir = study_dir / "0_model"
        if not model_dir.exists():
            raise HTTPException(status_code=404, detail=f"Model directory not found for {study_id}")

        # Find .prt file
        prt_files = list(model_dir.glob("*.prt"))
        # Exclude idealized parts
        prt_files = [f for f in prt_files if '_i.prt' not in f.name.lower()]

        if not prt_files:
            raise HTTPException(status_code=404, detail=f"No .prt files found in {model_dir}")

        prt_file = prt_files[0]  # Use first non-idealized part

        # Check for cached results
        cache_file = model_dir / "_introspection_cache.json"
        if cache_file.exists() and not force:
            try:
                with open(cache_file, 'r') as f:
                    cached = json.load(f)
                # Check if cache is for same file
                if cached.get('part_file') == str(prt_file):
                    return {
                        "study_id": study_id,
                        "cached": True,
                        "introspection": cached
                    }
            except:
                pass  # Invalid cache, re-run

        # Run introspection
        try:
            from optimization_engine.extractors.introspect_part import introspect_part
            result = introspect_part(str(prt_file), str(model_dir), verbose=False)

            # Cache results
            with open(cache_file, 'w') as f:
                json.dump(result, f, indent=2)

            return {
                "study_id": study_id,
                "cached": False,
                "introspection": result
            }

        except ImportError:
            raise HTTPException(
                status_code=500,
                detail="introspect_part module not available"
            )
        except Exception as e:
            raise HTTPException(
                status_code=500,
                detail=f"Introspection failed: {str(e)}"
            )

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to introspect: {str(e)}")


@router.get("/studies/{study_id}/nx/expressions")
async def get_nx_expressions(study_id: str):
    """
    Get just the expressions from NX model (simplified endpoint).

    Returns list of user expressions suitable for design variable selection.

    Args:
        study_id: Study identifier

    Returns:
        JSON with expressions array containing name, value, units, formula
    """
    try:
        study_dir = resolve_study_path(study_id)

        if not study_dir.exists():
            raise HTTPException(status_code=404, detail=f"Study {study_id} not found")

        # Check cache first
        model_dir = study_dir / "1_model"
        if not model_dir.exists():
            model_dir = study_dir / "0_model"

        cache_file = model_dir / "_introspection_cache.json" if model_dir.exists() else None

        if cache_file and cache_file.exists():
            with open(cache_file, 'r') as f:
                cached = json.load(f)

            expressions = cached.get('expressions', {}).get('user', [])
            return {
                "study_id": study_id,
                "expressions": expressions,
                "count": len(expressions)
            }

        # No cache, need to run introspection
        # Find .prt file
        if not model_dir or not model_dir.exists():
            raise HTTPException(status_code=404, detail=f"Model directory not found for {study_id}")

        prt_files = list(model_dir.glob("*.prt"))
        prt_files = [f for f in prt_files if '_i.prt' not in f.name.lower()]

        if not prt_files:
            raise HTTPException(status_code=404, detail=f"No .prt files found")

        prt_file = prt_files[0]

        try:
            from optimization_engine.extractors.introspect_part import introspect_part
            result = introspect_part(str(prt_file), str(model_dir), verbose=False)

            # Cache for future
            with open(cache_file, 'w') as f:
                json.dump(result, f, indent=2)

            expressions = result.get('expressions', {}).get('user', [])
            return {
                "study_id": study_id,
                "expressions": expressions,
                "count": len(expressions)
            }

        except Exception as e:
            raise HTTPException(status_code=500, detail=f"Failed to get expressions: {str(e)}")

    except HTTPException:
        raise
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Failed to get expressions: {str(e)}")