feat: Implement Agentic Architecture for robust session workflows
Phase 1 - Session Bootstrap: - Add .claude/ATOMIZER_CONTEXT.md as single entry point for new sessions - Add study state detection and task routing Phase 2 - Code Deduplication: - Add optimization_engine/base_runner.py (ConfigDrivenRunner) - Add optimization_engine/generic_surrogate.py (ConfigDrivenSurrogate) - Add optimization_engine/study_state.py for study detection - Add optimization_engine/templates/ with registry and templates - Studies now require ~50 lines instead of ~300 Phase 3 - Skill Consolidation: - Add YAML frontmatter metadata to all skills (versioning, dependencies) - Consolidate create-study.md into core/study-creation-core.md - Update 00_BOOTSTRAP.md, 01_CHEATSHEET.md, 02_CONTEXT_LOADER.md Phase 4 - Self-Expanding Knowledge: - Add optimization_engine/auto_doc.py for auto-generating documentation - Generate docs/generated/EXTRACTORS.md (27 extractors documented) - Generate docs/generated/TEMPLATES.md (6 templates) - Generate docs/generated/EXTRACTOR_CHEATSHEET.md Phase 5 - Subagent Implementation: - Add .claude/commands/study-builder.md (create studies) - Add .claude/commands/nx-expert.md (NX Open API) - Add .claude/commands/protocol-auditor.md (config validation) - Add .claude/commands/results-analyzer.md (results analysis) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Antoine
322
optimization_engine/study_state.py
Normal file
322
optimization_engine/study_state.py
Normal file
@@ -0,0 +1,322 @@
|
||||
"""
|
||||
Study State Detector for Atomizer
|
||||
|
||||
This module provides utilities to detect and summarize the state of an optimization study.
|
||||
Used by Claude sessions to quickly understand study context on initialization.
|
||||
"""
|
||||
|
||||
import json
|
||||
import sqlite3
|
||||
from pathlib import Path
|
||||
from typing import Dict, Any, Optional, List
|
||||
from datetime import datetime
|
||||
|
||||
|
||||
def detect_study_state(study_dir: Path) -> Dict[str, Any]:
|
||||
"""
|
||||
Detect the current state of an optimization study.
|
||||
|
||||
Args:
|
||||
study_dir: Path to the study directory
|
||||
|
||||
Returns:
|
||||
Dictionary with study state information
|
||||
"""
|
||||
study_dir = Path(study_dir)
|
||||
state = {
|
||||
"is_study": False,
|
||||
"study_name": study_dir.name,
|
||||
"status": "unknown",
|
||||
"config": None,
|
||||
"fea_trials": 0,
|
||||
"nn_trials": 0,
|
||||
"pareto_solutions": 0,
|
||||
"best_trial": None,
|
||||
"last_activity": None,
|
||||
"has_turbo_report": False,
|
||||
"has_surrogate": False,
|
||||
"warnings": [],
|
||||
"next_actions": []
|
||||
}
|
||||
|
||||
# Check if this is a valid study directory
|
||||
config_path = study_dir / "optimization_config.json"
|
||||
if not config_path.exists():
|
||||
# Try 1_setup subdirectory
|
||||
config_path = study_dir / "1_setup" / "optimization_config.json"
|
||||
|
||||
if not config_path.exists():
|
||||
state["warnings"].append("No optimization_config.json found")
|
||||
return state
|
||||
|
||||
state["is_study"] = True
|
||||
|
||||
# Load config
|
||||
try:
|
||||
with open(config_path, 'r') as f:
|
||||
config = json.load(f)
|
||||
state["config"] = _summarize_config(config)
|
||||
except Exception as e:
|
||||
state["warnings"].append(f"Failed to parse config: {e}")
|
||||
|
||||
# Check results directory
|
||||
results_dir = study_dir / "2_results"
|
||||
if not results_dir.exists():
|
||||
state["status"] = "not_started"
|
||||
state["next_actions"].append("Run: python run_optimization.py --discover")
|
||||
return state
|
||||
|
||||
# Check study.db for FEA trials
|
||||
db_path = results_dir / "study.db"
|
||||
if db_path.exists():
|
||||
fea_stats = _query_study_db(db_path)
|
||||
state.update(fea_stats)
|
||||
|
||||
# Check nn_study.db for NN trials
|
||||
nn_db_path = results_dir / "nn_study.db"
|
||||
if nn_db_path.exists():
|
||||
nn_stats = _query_study_db(nn_db_path, prefix="nn_")
|
||||
state["nn_trials"] = nn_stats.get("nn_fea_trials", 0)
|
||||
|
||||
# Check for turbo report
|
||||
turbo_report_path = results_dir / "turbo_report.json"
|
||||
if turbo_report_path.exists():
|
||||
state["has_turbo_report"] = True
|
||||
try:
|
||||
with open(turbo_report_path, 'r') as f:
|
||||
turbo = json.load(f)
|
||||
state["turbo_summary"] = {
|
||||
"mode": turbo.get("mode"),
|
||||
"nn_trials": turbo.get("total_nn_trials", 0),
|
||||
"fea_validations": turbo.get("fea_validations", 0),
|
||||
"time_minutes": round(turbo.get("time_minutes", 0), 1)
|
||||
}
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
# Check for trained surrogate
|
||||
surrogate_path = results_dir / "surrogate.pt"
|
||||
state["has_surrogate"] = surrogate_path.exists()
|
||||
|
||||
# Determine overall status
|
||||
state["status"] = _determine_status(state)
|
||||
|
||||
# Suggest next actions
|
||||
state["next_actions"] = _suggest_next_actions(state)
|
||||
|
||||
return state
|
||||
|
||||
|
||||
def _summarize_config(config: Dict) -> Dict[str, Any]:
|
||||
"""Extract key information from config."""
|
||||
# Handle different config formats
|
||||
variables = config.get("design_variables", config.get("variables", []))
|
||||
objectives = config.get("objectives", [])
|
||||
constraints = config.get("constraints", [])
|
||||
|
||||
# Get variable names (handle different key names)
|
||||
var_names = []
|
||||
for v in variables:
|
||||
name = v.get("parameter") or v.get("name") or v.get("expression_name", "unknown")
|
||||
var_names.append(name)
|
||||
|
||||
# Get objective names
|
||||
obj_names = []
|
||||
for o in objectives:
|
||||
name = o.get("name") or o.get("metric", "unknown")
|
||||
direction = o.get("goal") or o.get("direction", "minimize")
|
||||
obj_names.append(f"{name} ({direction})")
|
||||
|
||||
return {
|
||||
"n_variables": len(variables),
|
||||
"n_objectives": len(objectives),
|
||||
"n_constraints": len(constraints),
|
||||
"variable_names": var_names[:5], # First 5 only
|
||||
"objective_names": obj_names,
|
||||
"study_type": "multi_objective" if len(objectives) > 1 else "single_objective"
|
||||
}
|
||||
|
||||
|
||||
def _query_study_db(db_path: Path, prefix: str = "") -> Dict[str, Any]:
|
||||
"""Query Optuna study database for statistics."""
|
||||
stats = {
|
||||
f"{prefix}fea_trials": 0,
|
||||
f"{prefix}completed_trials": 0,
|
||||
f"{prefix}failed_trials": 0,
|
||||
f"{prefix}pareto_solutions": 0,
|
||||
"best_trial": None,
|
||||
"last_activity": None
|
||||
}
|
||||
|
||||
try:
|
||||
conn = sqlite3.connect(str(db_path))
|
||||
cursor = conn.cursor()
|
||||
|
||||
# Count trials by state
|
||||
cursor.execute("""
|
||||
SELECT state, COUNT(*) FROM trials
|
||||
GROUP BY state
|
||||
""")
|
||||
for state, count in cursor.fetchall():
|
||||
if state == "COMPLETE":
|
||||
stats[f"{prefix}completed_trials"] = count
|
||||
stats[f"{prefix}fea_trials"] = count
|
||||
elif state == "FAIL":
|
||||
stats[f"{prefix}failed_trials"] = count
|
||||
|
||||
# Get last activity time
|
||||
cursor.execute("""
|
||||
SELECT MAX(datetime_complete) FROM trials
|
||||
WHERE datetime_complete IS NOT NULL
|
||||
""")
|
||||
result = cursor.fetchone()
|
||||
if result and result[0]:
|
||||
stats["last_activity"] = result[0]
|
||||
|
||||
# Get best trial (for single objective)
|
||||
cursor.execute("""
|
||||
SELECT trial_id, value FROM trial_values
|
||||
WHERE objective_id = 0
|
||||
ORDER BY value ASC
|
||||
LIMIT 1
|
||||
""")
|
||||
result = cursor.fetchone()
|
||||
if result:
|
||||
stats["best_trial"] = {"trial_id": result[0], "value": result[1]}
|
||||
|
||||
# Count Pareto solutions (trials with user_attr pareto=True or non-dominated)
|
||||
# Simplified: count distinct trials in trial_values
|
||||
cursor.execute("""
|
||||
SELECT COUNT(DISTINCT trial_id) FROM trial_values
|
||||
""")
|
||||
result = cursor.fetchone()
|
||||
if result:
|
||||
# For multi-objective, this is a rough estimate
|
||||
stats[f"{prefix}pareto_solutions"] = min(result[0], 50) # Cap at 50
|
||||
|
||||
conn.close()
|
||||
except Exception as e:
|
||||
stats["db_error"] = str(e)
|
||||
|
||||
return stats
|
||||
|
||||
|
||||
def _determine_status(state: Dict) -> str:
|
||||
"""Determine overall study status."""
|
||||
if state["fea_trials"] == 0:
|
||||
return "not_started"
|
||||
elif state["fea_trials"] < 3:
|
||||
return "discovery"
|
||||
elif state["fea_trials"] < 10:
|
||||
return "validation"
|
||||
elif state["has_turbo_report"]:
|
||||
return "turbo_complete"
|
||||
elif state["has_surrogate"]:
|
||||
return "training_complete"
|
||||
elif state["fea_trials"] >= 50:
|
||||
return "fea_complete"
|
||||
else:
|
||||
return "in_progress"
|
||||
|
||||
|
||||
def _suggest_next_actions(state: Dict) -> List[str]:
|
||||
"""Suggest next actions based on study state."""
|
||||
actions = []
|
||||
|
||||
if state["status"] == "not_started":
|
||||
actions.append("Run: python run_optimization.py --discover")
|
||||
elif state["status"] == "discovery":
|
||||
actions.append("Run: python run_optimization.py --validate")
|
||||
elif state["status"] == "validation":
|
||||
actions.append("Run: python run_optimization.py --test")
|
||||
actions.append("Or run full: python run_optimization.py --run --trials 50")
|
||||
elif state["status"] == "in_progress":
|
||||
actions.append("Continue: python run_optimization.py --resume")
|
||||
elif state["status"] == "fea_complete":
|
||||
actions.append("Analyze: python -m optimization_engine.method_selector optimization_config.json 2_results/study.db")
|
||||
actions.append("Or run turbo: python run_nn_optimization.py --turbo")
|
||||
elif state["status"] == "turbo_complete":
|
||||
actions.append("View results in dashboard: cd atomizer-dashboard && npm run dev")
|
||||
actions.append("Generate report: python generate_report.py")
|
||||
|
||||
return actions
|
||||
|
||||
|
||||
def format_study_summary(state: Dict) -> str:
|
||||
"""Format study state as a human-readable summary."""
|
||||
if not state["is_study"]:
|
||||
return f"❌ Not a valid study directory: {state['study_name']}"
|
||||
|
||||
lines = [
|
||||
f"📊 **Study: {state['study_name']}**",
|
||||
f"Status: {state['status'].replace('_', ' ').title()}",
|
||||
""
|
||||
]
|
||||
|
||||
if state["config"]:
|
||||
cfg = state["config"]
|
||||
lines.append(f"**Configuration:**")
|
||||
lines.append(f"- Variables: {cfg['n_variables']} ({', '.join(cfg['variable_names'][:3])}{'...' if cfg['n_variables'] > 3 else ''})")
|
||||
lines.append(f"- Objectives: {cfg['n_objectives']} ({', '.join(cfg['objective_names'])})")
|
||||
lines.append(f"- Constraints: {cfg['n_constraints']}")
|
||||
lines.append(f"- Type: {cfg['study_type']}")
|
||||
lines.append("")
|
||||
|
||||
lines.append("**Progress:**")
|
||||
lines.append(f"- FEA trials: {state['fea_trials']}")
|
||||
if state["nn_trials"] > 0:
|
||||
lines.append(f"- NN trials: {state['nn_trials']}")
|
||||
if state["has_turbo_report"] and "turbo_summary" in state:
|
||||
ts = state["turbo_summary"]
|
||||
lines.append(f"- Turbo mode: {ts['nn_trials']} NN + {ts['fea_validations']} FEA validations ({ts['time_minutes']} min)")
|
||||
if state["last_activity"]:
|
||||
lines.append(f"- Last activity: {state['last_activity']}")
|
||||
lines.append("")
|
||||
|
||||
if state["next_actions"]:
|
||||
lines.append("**Suggested Next Actions:**")
|
||||
for action in state["next_actions"]:
|
||||
lines.append(f" → {action}")
|
||||
|
||||
if state["warnings"]:
|
||||
lines.append("")
|
||||
lines.append("**Warnings:**")
|
||||
for warning in state["warnings"]:
|
||||
lines.append(f" ⚠️ {warning}")
|
||||
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def get_all_studies(atomizer_root: Path) -> List[Dict[str, Any]]:
|
||||
"""Get state of all studies in the Atomizer studies directory."""
|
||||
studies_dir = atomizer_root / "studies"
|
||||
if not studies_dir.exists():
|
||||
return []
|
||||
|
||||
studies = []
|
||||
for study_path in studies_dir.iterdir():
|
||||
if study_path.is_dir() and not study_path.name.startswith("."):
|
||||
state = detect_study_state(study_path)
|
||||
if state["is_study"]:
|
||||
studies.append(state)
|
||||
|
||||
# Sort by last activity (most recent first)
|
||||
studies.sort(
|
||||
key=lambda s: s.get("last_activity") or "1970-01-01",
|
||||
reverse=True
|
||||
)
|
||||
|
||||
return studies
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
import sys
|
||||
|
||||
if len(sys.argv) > 1:
|
||||
study_path = Path(sys.argv[1])
|
||||
else:
|
||||
# Default to current directory
|
||||
study_path = Path.cwd()
|
||||
|
||||
state = detect_study_state(study_path)
|
||||
print(format_study_summary(state))
|
||||
Reference in New Issue
Block a user