Spaces:
Runtime error
Runtime error
File size: 16,892 Bytes
77a06d0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 |
"""
Orchestrator Module for DReamMachine
Manages the 7-step multi-agent dream cycle
"""
import logging
import time
from typing import Dict, List, Optional, Any
from datetime import datetime
import yaml
from prompt_manager import PromptManager
from llm_agent import LLMAgent
from data_logger import DataLogger
logger = logging.getLogger(__name__)
class DreamOrchestrator:
"""
Orchestrates the 7-step dream cycle across multiple LLM agents
Steps:
A.1. Setup - Initialize session & constraints
A.2. Dream & Generate - 3x Dreamer LLMs create ideas
A.3. Log & Narrate - Writer/Logger/Narrator refine output
A.4. Deep Think & Verify - Deep Thinker evaluates feasibility
A.5. Curate & Grade - Curator scores the concept
A.6. Data Storage - Save everything
A.7. Reforge Loop - Decide next stage or new prompt
"""
def __init__(self, config_path: str = "config.yaml", hf_token: Optional[str] = None):
"""
Initialize the Dream Orchestrator
Args:
config_path: Path to configuration file
hf_token: HuggingFace API token
"""
# Load configuration
with open(config_path, 'r') as f:
self.config = yaml.safe_load(f)
# Initialize modules
self.prompt_manager = PromptManager(config_path)
self.llm_agent = LLMAgent(config_path, hf_token)
self.data_logger = DataLogger(config_path, hf_token)
# Orchestration settings
orch_config = self.config.get('orchestration', {})
self.max_iterations = orch_config.get('max_iterations', 1000)
self.run_interval = orch_config.get('run_interval', 3600)
self.max_runtime = orch_config.get('max_runtime', 21600)
# Thresholds
thresholds = orch_config.get('auto_advance_threshold', {})
self.feasibility_min = thresholds.get('feasibility_min', 7)
self.originality_min = thresholds.get('originality_min', 5)
# Life stage progression
self.life_stages = ['init_1_25', 'mid_26_50', 'late_51_75', 'final_76_100']
self.current_stage_index = 0
# Session tracking
self.session_history = []
self.current_idea_context = None
logger.info("DreamOrchestrator initialized successfully")
def run_dream_round(
self,
stage: Optional[str] = None,
previous_context: Optional[str] = None
) -> Dict[str, Any]:
"""
Execute a complete dream round (steps A.1 through A.7)
Args:
stage: Life stage name (or None to use current)
previous_context: Context from previous stage (for reforge)
Returns:
Complete session data dictionary
"""
logger.info("=" * 80)
logger.info(f"Starting Dream Round - Stage: {stage or 'init_1_25'}")
logger.info("=" * 80)
session_start_time = time.time()
# A.1. Setup - Initialize Session & Constraints
logger.info("\n[A.1] SETUP - Initializing session...")
stage = stage or 'init_1_25'
initial_prompt = self.prompt_manager.get_life_stage_prompt(stage, previous_context)
# A.2. Dream & Generate - 3x Dreamer LLMs
logger.info("\n[A.2] DREAM & GENERATE - Running dreamer models...")
dream_outputs = self._run_dreamers(initial_prompt)
# A.3. Log & Narrate - Writer/Logger/Narrator
logger.info("\n[A.3] LOG & NARRATE - Refining outputs...")
refinement_results = self._run_refinement(dream_outputs)
pitch_narrative = refinement_results['pitch']
technical_components = refinement_results['technical']
final_presentation = refinement_results['presentation']
# A.4. Deep Think & Verify - Feasibility Check
logger.info("\n[A.4] DEEP THINK & VERIFY - Evaluating feasibility...")
feasibility_report = self._run_deep_thinker(technical_components)
# A.5. Curate & Grade - Final Evaluation
logger.info("\n[A.5] CURATE & GRADE - Final scoring...")
curator_scorecard = self._run_curator(pitch_narrative, feasibility_report)
# A.6. Data Storage - Archive Session
logger.info("\n[A.6] DATA STORAGE - Archiving session...")
session_data = {
'life_stage': stage,
'initial_prompt': initial_prompt,
'dream_outputs': dream_outputs,
'pitch_narrative': pitch_narrative,
'technical_components': technical_components,
'final_presentation': final_presentation,
'feasibility_report': feasibility_report,
'curator_scorecard': curator_scorecard,
'execution_time_seconds': time.time() - session_start_time
}
session_id = self.data_logger.log_session_data(session_data)
session_data['session_id'] = session_id
# A.7. Reforge Loop - Iteration Prep
logger.info("\n[A.7] REFORGE LOOP - Determining next action...")
next_action = self._determine_next_action(curator_scorecard, stage, session_data)
session_data['next_action'] = next_action
# Log summary
self._log_session_summary(session_data)
self.session_history.append(session_data)
return session_data
def _run_dreamers(self, prompt: str) -> List[str]:
"""
Step A.2: Run multiple dreamer models in parallel
Args:
prompt: The dream prompt
Returns:
List of dream outputs
"""
logger.info("Generating creative visions from 3 dreamer models...")
try:
dreams = self.llm_agent.run_parallel_dreamers(prompt, num_dreamers=3)
for i, dream in enumerate(dreams, 1):
logger.info(f"Dreamer {i} generated {len(dream)} characters")
return dreams
except Exception as e:
logger.error(f"Error in dreamer stage: {str(e)}")
return [f"[Error: {str(e)}]"] * 3
def _run_refinement(self, dream_outputs: List[str]) -> Dict[str, str]:
"""
Step A.3: Run Writer, Logger, and Narrator to refine outputs
Args:
dream_outputs: Raw dream texts from dreamers
Returns:
Dictionary with pitch, technical, and presentation outputs
"""
logger.info("Refining dream outputs...")
try:
# Writer: Create coherent pitch
writer_prompt = self.prompt_manager.get_writer_prompt(dream_outputs)
pitch = self.llm_agent.get_writer_output(writer_prompt)
logger.info(f"Writer created pitch ({len(pitch)} characters)")
# Logger: Extract technical components
logger_prompt = self.prompt_manager.get_logger_prompt(pitch)
technical = self.llm_agent.get_logger_output(logger_prompt)
logger.info(f"Logger extracted technical components ({len(technical)} characters)")
# Narrator: Create final presentation
narrator_prompt = self.prompt_manager.get_narrator_prompt(pitch, technical)
presentation = self.llm_agent.get_narrator_output(narrator_prompt)
logger.info(f"Narrator created presentation ({len(presentation)} characters)")
return {
'pitch': pitch,
'technical': technical,
'presentation': presentation
}
except Exception as e:
logger.error(f"Error in refinement stage: {str(e)}")
return {
'pitch': f"[Error: {str(e)}]",
'technical': f"[Error: {str(e)}]",
'presentation': f"[Error: {str(e)}]"
}
def _run_deep_thinker(self, technical_components: str) -> str:
"""
Step A.4: Run deep thinker to evaluate feasibility
Args:
technical_components: Technical specification
Returns:
Feasibility report
"""
logger.info("Running feasibility analysis...")
try:
prompt = self.prompt_manager.get_deep_thinker_prompt(technical_components)
report = self.llm_agent.get_deep_thinker_output(prompt)
logger.info(f"Deep Thinker completed analysis ({len(report)} characters)")
return report
except Exception as e:
logger.error(f"Error in deep thinker stage: {str(e)}")
return f"[Error in feasibility analysis: {str(e)}]"
def _run_curator(self, pitch: str, feasibility_report: str) -> Dict[str, Any]:
"""
Step A.5: Run curator to score and evaluate
Args:
pitch: The narrative pitch
feasibility_report: Feasibility analysis
Returns:
Curator scorecard dictionary
"""
logger.info("Running final curation and scoring...")
try:
prompt = self.prompt_manager.get_curator_prompt(pitch, feasibility_report)
scorecard = self.llm_agent.get_curator_score(prompt)
logger.info(f"Curator assigned scores:")
logger.info(f" - Originality: {scorecard.get('originality', 'N/A')}")
logger.info(f" - Feasibility: {scorecard.get('feasibility', 'N/A')}")
logger.info(f" - Global Impact: {scorecard.get('global_impact', 'N/A')}")
logger.info(f" - Narrative Coherence: {scorecard.get('narrative_coherence', 'N/A')}")
logger.info(f" - Reforge Flag: {scorecard.get('reforge_flag', False)}")
return scorecard
except Exception as e:
logger.error(f"Error in curator stage: {str(e)}")
return {
'originality': 0,
'feasibility': 0,
'global_impact': 0,
'narrative_coherence': 0,
'reforge_flag': False,
'overall_assessment': f'Error: {str(e)}'
}
def _determine_next_action(
self,
scorecard: Dict[str, Any],
current_stage: str,
session_data: Dict[str, Any]
) -> Dict[str, Any]:
"""
Step A.7: Determine next action based on scorecard
Args:
scorecard: Curator scorecard
current_stage: Current life stage
session_data: Full session data
Returns:
Dictionary describing next action
"""
reforge_flag = scorecard.get('reforge_flag', False)
current_index = self.life_stages.index(current_stage)
if reforge_flag and current_index < len(self.life_stages) - 1:
# Advance to next life stage
next_stage = self.life_stages[current_index + 1]
action = {
'type': 'advance',
'next_stage': next_stage,
'reason': 'Idea meets criteria for advancement',
'context': session_data['pitch_narrative']
}
logger.info(f"β Advancing idea to next stage: {next_stage}")
elif reforge_flag and current_index == len(self.life_stages) - 1:
# Completed all stages successfully
action = {
'type': 'complete',
'reason': 'Idea successfully completed all life stages',
'final_assessment': scorecard.get('overall_assessment', '')
}
logger.info("β Idea has completed all life stages successfully!")
else:
# Archive and start new idea
action = {
'type': 'new_idea',
'reason': f'Scores did not meet threshold (F:{scorecard.get("feasibility")}, O:{scorecard.get("originality")})',
'next_stage': 'init_1_25'
}
logger.info("β Archiving idea and starting fresh with new prompt")
return action
def _log_session_summary(self, session_data: Dict[str, Any]) -> None:
"""Log a human-readable summary of the session"""
logger.info("\n" + "=" * 80)
logger.info("SESSION SUMMARY")
logger.info("=" * 80)
logger.info(f"Session ID: {session_data.get('session_id', 'N/A')}")
logger.info(f"Life Stage: {session_data.get('life_stage', 'N/A')}")
logger.info(f"Execution Time: {session_data.get('execution_time_seconds', 0):.2f}s")
scorecard = session_data.get('curator_scorecard', {})
logger.info(f"\nScores:")
logger.info(f" Originality: {scorecard.get('originality', 'N/A')}/10")
logger.info(f" Feasibility: {scorecard.get('feasibility', 'N/A')}/10")
logger.info(f" Global Impact: {scorecard.get('global_impact', 'N/A')}/10")
logger.info(f" Narrative Coherence: {scorecard.get('narrative_coherence', 'N/A')}/10")
logger.info(f"\nReforge Flag: {scorecard.get('reforge_flag', False)}")
next_action = session_data.get('next_action', {})
logger.info(f"\nNext Action: {next_action.get('type', 'unknown')}")
logger.info(f"Reason: {next_action.get('reason', 'N/A')}")
logger.info("=" * 80 + "\n")
def run_batch_mode(self, num_rounds: int = 10, sleep_between: int = 0) -> List[Dict[str, Any]]:
"""
Run multiple dream rounds in batch mode
Args:
num_rounds: Number of rounds to run
sleep_between: Seconds to sleep between rounds
Returns:
List of all session data
"""
logger.info(f"Starting batch mode: {num_rounds} rounds")
results = []
for i in range(num_rounds):
logger.info(f"\n### BATCH ROUND {i + 1}/{num_rounds} ###\n")
try:
# Determine stage and context based on previous session
if self.session_history and self.session_history[-1]['next_action']['type'] == 'advance':
last_session = self.session_history[-1]
stage = last_session['next_action']['next_stage']
context = last_session['next_action']['context']
else:
stage = 'init_1_25'
context = None
# Run dream round
session_data = self.run_dream_round(stage=stage, previous_context=context)
results.append(session_data)
# Sleep between rounds if configured
if sleep_between > 0 and i < num_rounds - 1:
logger.info(f"Sleeping {sleep_between} seconds before next round...")
time.sleep(sleep_between)
except Exception as e:
logger.error(f"Error in batch round {i + 1}: {str(e)}")
continue
logger.info(f"\nBatch mode completed: {len(results)} successful rounds")
return results
def run_scheduled_mode(self) -> None:
"""
Run in scheduled mode (for HuggingFace Spaces)
Runs until max_runtime is reached
"""
logger.info(f"Starting scheduled mode (max runtime: {self.max_runtime}s)")
start_time = time.time()
round_count = 0
while True:
elapsed = time.time() - start_time
if elapsed >= self.max_runtime:
logger.info(f"Max runtime reached ({self.max_runtime}s). Stopping.")
break
if round_count >= self.max_iterations:
logger.info(f"Max iterations reached ({self.max_iterations}). Stopping.")
break
try:
logger.info(f"\n### SCHEDULED ROUND {round_count + 1} (Elapsed: {elapsed:.0f}s) ###\n")
# Determine stage and context
if self.session_history and self.session_history[-1]['next_action']['type'] == 'advance':
last_session = self.session_history[-1]
stage = last_session['next_action']['next_stage']
context = last_session['next_action']['context']
else:
stage = 'init_1_25'
context = None
# Run dream round
self.run_dream_round(stage=stage, previous_context=context)
round_count += 1
# Wait before next iteration
if self.run_interval > 0:
logger.info(f"Waiting {self.run_interval}s before next round...")
time.sleep(self.run_interval)
except Exception as e:
logger.error(f"Error in scheduled round: {str(e)}")
time.sleep(60) # Wait 1 minute on error before retrying
logger.info(f"\nScheduled mode completed: {round_count} rounds in {time.time() - start_time:.0f}s")
# Convenience function
def create_orchestrator(config_path: str = "config.yaml", hf_token: Optional[str] = None) -> DreamOrchestrator:
"""Create and return a configured DreamOrchestrator"""
return DreamOrchestrator(config_path, hf_token)
|