File size: 12,431 Bytes
fcf8749 | 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 | """
Admin Learning API endpoints for Phase 8.
Provides monitoring, control, and debugging for the Learning Agent.
"""
from datetime import datetime, timedelta
from typing import Optional
from uuid import UUID
from fastapi import APIRouter, Depends, HTTPException, Query, status
from sqlalchemy import select, func
from sqlalchemy.ext.asyncio import AsyncSession
from app.database import get_db
from app.models import (
LearningEpisode,
DriverEffortModel,
FairnessConfig,
Driver,
)
from app.services.learning_agent import LearningAgent, hash_config
from app.schemas.learning_schemas import (
LearningStatusResponse,
LearningEpisodeResponse,
LearningEpisodesListResponse,
DriverModelStatusResponse,
ForceConfigRequest,
ForceConfigResponse,
TriggerLearningRequest,
TriggerLearningResponse,
DriverModelUpdateRequest,
DriverModelUpdateResponse,
AllDriverModelsListResponse,
BanditStatistics,
)
router = APIRouter(prefix="/admin/learning", tags=["Admin - Learning Agent"])
@router.get(
"/status",
response_model=LearningStatusResponse,
summary="Get learning agent status",
description="Returns overall learning status including bandit statistics, "
"active driver models, and top performing configurations.",
)
async def get_learning_status(
db: AsyncSession = Depends(get_db),
) -> LearningStatusResponse:
"""Get current learning agent status."""
agent = LearningAgent(db)
status_dict = await agent.get_learning_status()
return LearningStatusResponse(
current_config=status_dict.get("current_config"),
top_performing_configs=status_dict.get("top_performing_configs", []),
driver_models_active=status_dict.get("driver_models_active", 0),
avg_prediction_mse=status_dict.get("avg_prediction_mse", 0.0),
recent_episodes_7d=status_dict.get("recent_episodes_7d", 0),
total_arms=status_dict.get("total_arms", 81),
bandit_statistics=BanditStatistics(
total_samples=status_dict.get("bandit_statistics", {}).get("total_samples", 0),
explored_arms=status_dict.get("bandit_statistics", {}).get("explored_arms", 0),
total_arms=status_dict.get("total_arms", 81),
),
)
@router.get(
"/episodes",
response_model=LearningEpisodesListResponse,
summary="List learning episodes",
description="Returns paginated list of learning episodes with rewards.",
)
async def list_episodes(
page: int = Query(1, ge=1),
page_size: int = Query(20, ge=1, le=100),
has_reward: Optional[bool] = Query(None),
is_experimental: Optional[bool] = Query(None),
db: AsyncSession = Depends(get_db),
) -> LearningEpisodesListResponse:
"""List learning episodes with optional filters."""
# Build query
query = select(LearningEpisode).order_by(LearningEpisode.created_at.desc())
count_query = select(func.count(LearningEpisode.id))
# Apply filters
if has_reward is True:
query = query.where(LearningEpisode.episode_reward.isnot(None))
count_query = count_query.where(LearningEpisode.episode_reward.isnot(None))
elif has_reward is False:
query = query.where(LearningEpisode.episode_reward.is_(None))
count_query = count_query.where(LearningEpisode.episode_reward.is_(None))
if is_experimental is not None:
query = query.where(LearningEpisode.is_experimental == is_experimental)
count_query = count_query.where(LearningEpisode.is_experimental == is_experimental)
# Get total count
total_result = await db.execute(count_query)
total = total_result.scalar() or 0
# Apply pagination
offset = (page - 1) * page_size
query = query.offset(offset).limit(page_size)
result = await db.execute(query)
episodes = result.scalars().all()
return LearningEpisodesListResponse(
episodes=[LearningEpisodeResponse.model_validate(ep) for ep in episodes],
total=total,
page=page,
page_size=page_size,
has_next=(offset + len(episodes)) < total,
)
@router.get(
"/episodes/{episode_id}",
response_model=LearningEpisodeResponse,
summary="Get learning episode details",
)
async def get_episode(
episode_id: UUID,
db: AsyncSession = Depends(get_db),
) -> LearningEpisodeResponse:
"""Get details of a specific learning episode."""
result = await db.execute(
select(LearningEpisode).where(LearningEpisode.id == episode_id)
)
episode = result.scalar_one_or_none()
if not episode:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=f"Episode {episode_id} not found",
)
return LearningEpisodeResponse.model_validate(episode)
@router.post(
"/force_config",
response_model=ForceConfigResponse,
summary="Force a specific fairness config",
description="Override bandit selection with a specific config. "
"Use for emergency rollbacks or testing.",
)
async def force_config(
request: ForceConfigRequest,
db: AsyncSession = Depends(get_db),
) -> ForceConfigResponse:
"""Force a specific fairness configuration."""
# Get current active config
result = await db.execute(
select(FairnessConfig).where(FairnessConfig.is_active == True).limit(1)
)
current = result.scalar_one_or_none()
previous_config = None
if current:
previous_config = {
"gini_threshold": current.gini_threshold,
"stddev_threshold": current.stddev_threshold,
"recovery_lightening_factor": current.recovery_lightening_factor,
"ev_charging_penalty_weight": current.ev_charging_penalty_weight,
}
current.is_active = False
# Create new config
new_config = FairnessConfig(
is_active=True,
gini_threshold=request.gini_threshold,
stddev_threshold=request.stddev_threshold,
recovery_lightening_factor=request.recovery_lightening_factor,
ev_charging_penalty_weight=request.ev_charging_penalty_weight,
max_gap_threshold=request.max_gap_threshold,
)
db.add(new_config)
await db.commit()
return ForceConfigResponse(
status="success",
message=f"Config forced. Reason: {request.reason}",
config_applied={
"gini_threshold": request.gini_threshold,
"stddev_threshold": request.stddev_threshold,
"recovery_lightening_factor": request.recovery_lightening_factor,
"ev_charging_penalty_weight": request.ev_charging_penalty_weight,
"max_gap_threshold": request.max_gap_threshold,
},
previous_config=previous_config,
)
@router.post(
"/trigger",
response_model=TriggerLearningResponse,
summary="Manually trigger learning pipeline",
description="Trigger the daily learning pipeline manually. "
"Useful for testing or forcing immediate updates.",
)
async def trigger_learning(
request: TriggerLearningRequest,
db: AsyncSession = Depends(get_db),
) -> TriggerLearningResponse:
"""Manually trigger learning pipeline."""
from cron.daily_learning import DailyLearningPipeline
start_time = datetime.utcnow()
metrics = {
"episodes_processed": 0,
"rewards_computed": 0,
"models_updated": 0,
"config_selection": None,
"errors": [],
}
try:
pipeline = DailyLearningPipeline(db)
if request.process_episodes:
await pipeline._process_pending_episodes()
metrics["episodes_processed"] = pipeline.metrics.get("episodes_processed", 0)
metrics["rewards_computed"] = pipeline.metrics.get("rewards_computed", 0)
if request.select_config:
await pipeline._select_todays_config()
metrics["config_selection"] = pipeline.metrics.get("config_selection")
if request.update_models:
await pipeline._update_driver_models()
metrics["models_updated"] = pipeline.metrics.get("models_updated", 0)
await db.commit()
return TriggerLearningResponse(
status="success",
episodes_processed=metrics["episodes_processed"],
rewards_computed=metrics["rewards_computed"],
models_updated=metrics["models_updated"],
config_selection=metrics["config_selection"],
duration_seconds=(datetime.utcnow() - start_time).total_seconds(),
errors=metrics.get("errors", []),
)
except Exception as e:
return TriggerLearningResponse(
status="error",
duration_seconds=(datetime.utcnow() - start_time).total_seconds(),
errors=[str(e)],
)
@router.get(
"/models",
response_model=AllDriverModelsListResponse,
summary="List all driver effort models",
)
async def list_driver_models(
active_only: bool = Query(True),
db: AsyncSession = Depends(get_db),
) -> AllDriverModelsListResponse:
"""List all driver effort models."""
query = select(DriverEffortModel)
if active_only:
query = query.where(DriverEffortModel.active == True)
query = query.order_by(DriverEffortModel.last_trained_at.desc().nullsfirst())
result = await db.execute(query)
models = result.scalars().all()
# Calculate stats
active_count = sum(1 for m in models if m.active)
mse_values = [m.current_mse for m in models if m.current_mse is not None]
avg_mse = sum(mse_values) / len(mse_values) if mse_values else None
return AllDriverModelsListResponse(
models=[
DriverModelStatusResponse(
driver_id=m.driver_id,
model_version=m.model_version,
training_samples=m.training_samples,
current_mse=m.current_mse,
r2_score=m.r2_score,
mse_history=m.mse_history,
active=m.active,
last_trained_at=m.last_trained_at,
)
for m in models
],
total=len(models),
active_count=active_count,
avg_mse=avg_mse,
)
@router.get(
"/models/{driver_id}",
response_model=DriverModelStatusResponse,
summary="Get driver model status",
)
async def get_driver_model(
driver_id: UUID,
db: AsyncSession = Depends(get_db),
) -> DriverModelStatusResponse:
"""Get status of a specific driver's effort model."""
result = await db.execute(
select(DriverEffortModel).where(DriverEffortModel.driver_id == driver_id)
)
model = result.scalar_one_or_none()
if not model:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=f"No model found for driver {driver_id}",
)
return DriverModelStatusResponse(
driver_id=model.driver_id,
model_version=model.model_version,
training_samples=model.training_samples,
current_mse=model.current_mse,
r2_score=model.r2_score,
mse_history=model.mse_history,
active=model.active,
last_trained_at=model.last_trained_at,
)
@router.post(
"/models/{driver_id}/retrain",
response_model=DriverModelUpdateResponse,
summary="Retrain a driver's model",
)
async def retrain_driver_model(
driver_id: UUID,
db: AsyncSession = Depends(get_db),
) -> DriverModelUpdateResponse:
"""Manually trigger retraining of a driver's effort model."""
# Check driver exists
result = await db.execute(
select(Driver).where(Driver.id == driver_id)
)
driver = result.scalar_one_or_none()
if not driver:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=f"Driver {driver_id} not found",
)
agent = LearningAgent(db)
result = await agent.effort_learner.update_model(driver_id)
await db.commit()
return DriverModelUpdateResponse(
status=result.get("status", "unknown"),
driver_id=driver_id,
model_version=result.get("model_version"),
training_samples=result.get("training_samples"),
mse=result.get("mse"),
r2_score=result.get("r2_score"),
reason=result.get("reason"),
)
|