Spaces:
Running
Running
File size: 5,951 Bytes
09801ca | 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 | """
🚀 Deployment Agent
Exposes production-ready inference pipeline:
- Committee approval check
- Inference API preparation
- Model serialization
- Performance validation
Only deploys when ALL agents agree the model is production-ready.
"""
import numpy as np
from typing import Dict, List, Any, Optional
import logging
import pickle
import json
from datetime import datetime
from .base import BaseAgent, AgentResult, AgentStatus, Phase
logger = logging.getLogger(__name__)
class DeploymentAgent(BaseAgent):
"""
Deployment Agent
Prepares production-ready inference pipeline after committee approval.
"""
name = "deployment"
description = "Prepares production inference pipeline"
def __init__(self, memory=None):
super().__init__(memory)
self.deployment_info: Dict[str, Any] = {}
def execute(self, **kwargs) -> AgentResult:
"""Main execution: prepare deployment package"""
# Check approvals
training_validated = self.read_state("training_validated")
evaluation_approved = self.read_state("evaluation_approved")
# In fast phase, be more lenient
if self.is_deep_phase() and not (training_validated and evaluation_approved):
return AgentResult(
status=AgentStatus.FAILED,
agent_name=self.name,
phase=self.current_phase,
errors=["Model not approved by all agents"]
)
# Get model and metadata
model_artifact = self.memory.get_latest_artifact("model")
if model_artifact is None:
return AgentResult(
status=AgentStatus.FAILED,
agent_name=self.name,
phase=self.current_phase,
errors=["No trained model found"]
)
model = model_artifact.data
task_type = self.read_state("task_type")
target_col = self.read_state("target_column")
feature_names = self.read_state("feature_names_final")
if feature_names is None:
feature_names = self.read_state("feature_names")
best_score = self.read_state("best_score")
self.logger.info(f"🚀 Preparing deployment package...")
# Create deployment package
deployment_package = {
"model": model,
"model_name": model_artifact.metadata.get("name", "unknown"),
"task_type": task_type,
"target_column": target_col,
"feature_names": feature_names,
"score": best_score,
"deployed_at": datetime.now().isoformat(),
"phase": self.current_phase.value,
"approved_by": ["training_validator", "evaluation"] if evaluation_approved else ["training_validator"]
}
# Validate inference
inference_test = self._validate_inference(model, self.read_state("features_engineered"))
if not inference_test["success"]:
return AgentResult(
status=AgentStatus.FAILED,
agent_name=self.name,
phase=self.current_phase,
errors=[f"Inference validation failed: {inference_test['error']}"]
)
deployment_package["inference_latency_ms"] = inference_test["latency_ms"]
# Store deployment package
self.memory.store_artifact(
artifact_id="deployment_package",
artifact_type="deployment",
producer=self.name,
data=deployment_package,
metadata={"ready": True}
)
self.deployment_info = deployment_package
self.write_state("deployment_ready", True, self.name)
self.write_state("deployment_package", {
k: v for k, v in deployment_package.items()
if k not in ["model"] # Don't serialize model to state
}, self.name)
self.logger.info(f" ✅ Deployment package ready")
self.logger.info(f" 📊 Model: {deployment_package['model_name']}")
self.logger.info(f" 📊 Score: {best_score:.4f}")
self.logger.info(f" 📊 Latency: {inference_test['latency_ms']:.2f}ms")
return AgentResult(
status=AgentStatus.SUCCESS,
agent_name=self.name,
phase=self.current_phase,
data={
"deployed": True,
"model_name": deployment_package["model_name"],
"latency_ms": inference_test["latency_ms"]
},
metrics={
"score": best_score,
"latency_ms": inference_test["latency_ms"]
}
)
def _validate_inference(self, model, X: np.ndarray) -> Dict[str, Any]:
"""Validate model can make predictions"""
try:
import time
# Test prediction
sample = X[:10] if X is not None else None
if sample is None:
return {"success": False, "error": "No features available"}
# Measure latency
start = time.time()
_ = model.predict(sample)
latency = (time.time() - start) * 1000 / len(sample) # Per-sample ms
return {
"success": True,
"latency_ms": latency
}
except Exception as e:
return {
"success": False,
"error": str(e)[:100]
}
def get_predictor(self):
"""Get a predictor function for inference"""
deployment = self.memory.get_artifact("deployment_package")
if deployment is None:
return None
model = deployment.data["model"]
def predict(X):
return model.predict(X)
return predict
|