mirror of
https://github.com/velocitatem/PHANTOM.git
synced 2026-05-31 08:33:36 +00:00
Airflow addition (#28)
* introducing airflow to run pipeline * chore: updating dag with upload to registry * introducing complete provider (non refactored and noisy) * chore: removing old shit * generic pricing baselines * feature: super simple model registry (to be updated maybe third party OS software) * chore: refactoring the providers docker config and requirements * chore: refactored and broke down components (braking * exporting all * local pipeline excution working * fix: fixing import structures from nonrelativistic * chore: enables cross comm pickling with fully e2e pipeline compilation * docs: what the pipeline is like now * pipelines local running and pipeline high level definition * cleaning old pipeline and vectorization * leaked but fixing, not so important * test: started with pipeline step testing * chore: cleaning up provider of prices * test: extra tests wit hsemantic meaning checks * migrating pricers * feature: introducing pricing predictors (pricers) * chore: e2e is done with new pipeline * extra session feature extraction * feature: experiemntal sessin pricer and metrics(vibe) * chore: redefined and connected pricers (#29)
This commit is contained in:
Daniel Alves Rösel
committed by
GitHub
GitHub
parent
2a0e44ab24
commit
ad9423bf59
346
experiments/airflow/dags/elasticity_pricing_dag.py
Normal file
346
experiments/airflow/dags/elasticity_pricing_dag.py
Normal file
@@ -0,0 +1,346 @@
|
||||
from airflow import DAG
|
||||
from airflow.operators.python import PythonOperator
|
||||
from airflow.utils.dates import days_ago
|
||||
from datetime import timedelta
|
||||
import pandas as pd
|
||||
import logging
|
||||
import sys
|
||||
import pickle
|
||||
import io
|
||||
|
||||
# add parent dir to path so procesing package can be imported
|
||||
sys.path.insert(0, '/opt/airflow')
|
||||
|
||||
from procesing.context import PipelineContext
|
||||
from procesing.providers import SupabaseProvider, BackendAPIProvider
|
||||
from procesing.steps import (
|
||||
FetchInteractionsStep,
|
||||
FetchPriceLogsStep,
|
||||
CreatePriceBucketsStep,
|
||||
AugmentEventNamesStep,
|
||||
ChunkByTimeWindowStep,
|
||||
ComputeDemandForChunksStep,
|
||||
AggregatePriceLogsStep,
|
||||
ComputeElasticityStep,
|
||||
BuildStateSpaceStep,
|
||||
FitPricingFunctionStep,
|
||||
PredictPricesStep,
|
||||
)
|
||||
|
||||
default_args = {
|
||||
'owner': 'phantom-research',
|
||||
'depends_on_past': False,
|
||||
'email_on_failure': False,
|
||||
'email_on_retry': False,
|
||||
'retries': 2,
|
||||
'retry_delay': timedelta(minutes=5),
|
||||
}
|
||||
|
||||
def get_provider():
|
||||
"""Factory to create composite provider"""
|
||||
class CompositeProvider(SupabaseProvider, BackendAPIProvider):
|
||||
def __init__(self):
|
||||
SupabaseProvider.__init__(self)
|
||||
BackendAPIProvider.__init__(self)
|
||||
return CompositeProvider()
|
||||
|
||||
def get_context(**kwargs):
|
||||
"""Build pipeline context from Airflow config"""
|
||||
dag_conf = kwargs.get('dag_run').conf if kwargs.get('dag_run') else {}
|
||||
return PipelineContext(
|
||||
provider=get_provider(),
|
||||
store_mode=dag_conf.get('store_mode', 'hotel'),
|
||||
window_size=dag_conf.get('window_size', '30s'),
|
||||
n_price_buckets=dag_conf.get('n_price_buckets', 5),
|
||||
elasticity_method=dag_conf.get('elasticity_method', 'point'),
|
||||
min_observations=dag_conf.get('min_observations', 2),
|
||||
)
|
||||
|
||||
# atomic task functions (each wraps one sklearn step)
|
||||
def fetch_interactions(**kwargs):
|
||||
"""Task: Fetch interaction data from Kafka"""
|
||||
context = get_context(**kwargs)
|
||||
step = FetchInteractionsStep(context)
|
||||
df = step.transform(None)
|
||||
|
||||
kwargs['ti'].xcom_push(key='interactions_raw', value=pickle.dumps(df))
|
||||
logging.info(f"Fetched {len(df)} interaction records")
|
||||
return len(df)
|
||||
|
||||
def fetch_price_logs(**kwargs):
|
||||
"""Task: Fetch price logs from Kafka"""
|
||||
context = get_context(**kwargs)
|
||||
step = FetchPriceLogsStep(context)
|
||||
df = step.transform(None)
|
||||
|
||||
kwargs['ti'].xcom_push(key='price_logs_raw', value=pickle.dumps(df))
|
||||
logging.info(f"Fetched {len(df)} price records")
|
||||
return len(df)
|
||||
|
||||
def create_price_buckets(**kwargs):
|
||||
"""Task: Create price buckets for interactions"""
|
||||
ti = kwargs['ti']
|
||||
df = pickle.loads(ti.xcom_pull(key='interactions_raw'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = CreatePriceBucketsStep(context)
|
||||
df = step.transform(df)
|
||||
|
||||
ti.xcom_push(key='interactions_bucketed', value=pickle.dumps(df))
|
||||
logging.info(f"Created price buckets for {len(df)} interactions")
|
||||
return len(df)
|
||||
|
||||
def augment_event_names(**kwargs):
|
||||
"""Task: Augment event names with product and price schema"""
|
||||
ti = kwargs['ti']
|
||||
df = pickle.loads(ti.xcom_pull(key='interactions_bucketed'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = AugmentEventNamesStep(context)
|
||||
df = step.transform(df)
|
||||
|
||||
ti.xcom_push(key='interactions_final', value=pickle.dumps(df))
|
||||
logging.info(f"Augmented event names for {len(df)} interactions")
|
||||
return len(df)
|
||||
|
||||
def chunk_interactions(**kwargs):
|
||||
"""Task: Chunk interactions into time windows"""
|
||||
ti = kwargs['ti']
|
||||
df = pickle.loads(ti.xcom_pull(key='interactions_final'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = ChunkByTimeWindowStep(context)
|
||||
chunks = step.transform(df)
|
||||
|
||||
ti.xcom_push(key='interaction_chunks', value=pickle.dumps(chunks))
|
||||
logging.info(f"Generated {len(chunks)} interaction chunks")
|
||||
return len(chunks)
|
||||
|
||||
def compute_demand(**kwargs):
|
||||
"""Task: Compute demand vectors for all chunks"""
|
||||
ti = kwargs['ti']
|
||||
chunks = pickle.loads(ti.xcom_pull(key='interaction_chunks'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = ComputeDemandForChunksStep(context)
|
||||
demand_chunks = step.transform(chunks)
|
||||
|
||||
ti.xcom_push(key='demand_chunks', value=pickle.dumps(demand_chunks))
|
||||
logging.info(f"Computed demand for {len(demand_chunks)} chunks")
|
||||
return len(demand_chunks)
|
||||
|
||||
def aggregate_price_logs(**kwargs):
|
||||
"""Task: Aggregate price logs into time windows """
|
||||
ti = kwargs['ti']
|
||||
df = pickle.loads(ti.xcom_pull(key='price_logs_raw'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = AggregatePriceLogsStep(context)
|
||||
price_chunks = step.transform(df)
|
||||
|
||||
ti.xcom_push(key='price_chunks', value=pickle.dumps(price_chunks))
|
||||
logging.info(f"Aggregated {len(price_chunks)} price chunks")
|
||||
return len(price_chunks)
|
||||
|
||||
def compute_elasticity(**kwargs):
|
||||
"""Task: Compute price elasticity from demand and price chunks"""
|
||||
ti = kwargs['ti']
|
||||
demand_chunks = pickle.loads(ti.xcom_pull(key='demand_chunks'))
|
||||
price_chunks = pickle.loads(ti.xcom_pull(key='price_chunks'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = ComputeElasticityStep(context)
|
||||
elasticity_df = step.transform((demand_chunks, price_chunks))
|
||||
|
||||
ti.xcom_push(key='elasticity_results', value=pickle.dumps(elasticity_df))
|
||||
logging.info(f"Computed elasticity for {len(elasticity_df)} products")
|
||||
|
||||
return {
|
||||
'n_products': len(elasticity_df),
|
||||
'mean_elasticity': float(elasticity_df['elasticity'].mean()),
|
||||
'median_elasticity': float(elasticity_df['elasticity'].median())
|
||||
}
|
||||
|
||||
def build_state_space(**kwargs):
|
||||
"""Task: Build state space from elasticity"""
|
||||
ti = kwargs['ti']
|
||||
elasticity_df = pickle.loads(ti.xcom_pull(key='elasticity_results'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = BuildStateSpaceStep(context)
|
||||
state_space = step.transform(elasticity_df)
|
||||
|
||||
ti.xcom_push(key='state_space', value=pickle.dumps(state_space))
|
||||
logging.info("Built state space for pricing")
|
||||
return True
|
||||
|
||||
def fit_pricing_function(**kwargs):
|
||||
"""Task: Fit pricing function using elasticity"""
|
||||
ti = kwargs['ti']
|
||||
elasticity_df = pickle.loads(ti.xcom_pull(key='elasticity_results'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = FitPricingFunctionStep(context)
|
||||
pricer = step.transform(elasticity_df)
|
||||
|
||||
ti.xcom_push(key='pricer', value=pickle.dumps(pricer))
|
||||
logging.info("Fitted pricing function")
|
||||
return True
|
||||
|
||||
def predict_prices(**kwargs):
|
||||
"""Task: Predict optimal prices"""
|
||||
ti = kwargs['ti']
|
||||
pricer = pickle.loads(ti.xcom_pull(key='pricer'))
|
||||
state_space = pickle.loads(ti.xcom_pull(key='state_space'))
|
||||
|
||||
context = get_context(**kwargs)
|
||||
step = PredictPricesStep(context)
|
||||
prices_df = step.transform((pricer, state_space))
|
||||
|
||||
ti.xcom_push(key='predicted_prices', value=pickle.dumps(prices_df))
|
||||
logging.info(f"Predicted prices for {len(prices_df)} products")
|
||||
return len(prices_df)
|
||||
|
||||
def publish_results(**kwargs):
|
||||
"""Task: Publish elasticity and pricing results to model registry"""
|
||||
ti = kwargs['ti']
|
||||
elasticity_df = pickle.loads(ti.xcom_pull(key='elasticity_results'))
|
||||
prices_df = pickle.loads(ti.xcom_pull(key='predicted_prices'))
|
||||
|
||||
sys.path.insert(0, '/opt/airflow')
|
||||
from lib.model_registry import ModelRegistry
|
||||
|
||||
registry = ModelRegistry()
|
||||
dag_conf = kwargs.get('dag_run').conf if kwargs.get('dag_run') else {}
|
||||
|
||||
metadata = {
|
||||
'timestamp': pd.Timestamp.now().isoformat(),
|
||||
'window_size': dag_conf.get('window_size', '30s'),
|
||||
'store_mode': dag_conf.get('store_mode', 'hotel'),
|
||||
'dag_run_id': kwargs['dag_run'].run_id if kwargs.get('dag_run') else 'manual'
|
||||
}
|
||||
|
||||
registry.publish_elasticity(elasticity_df, model_name='latest', metadata=metadata)
|
||||
|
||||
# get fitted pricer from XCom
|
||||
pricer = pickle.loads(ti.xcom_pull(key='pricer'))
|
||||
registry.publish_pricing_model(
|
||||
pricer,
|
||||
model_name='latest',
|
||||
metadata={**metadata, 'model_type': type(pricer).__name__}
|
||||
)
|
||||
|
||||
logging.info(f"Published elasticity + pricing for {len(elasticity_df)} products")
|
||||
|
||||
return {
|
||||
'n_products': len(elasticity_df),
|
||||
'registry_status': 'success',
|
||||
'elasticity_mean': float(elasticity_df['elasticity'].mean())
|
||||
}
|
||||
|
||||
|
||||
# DAG definition
|
||||
with DAG(
|
||||
'elasticity_pricing_pipeline',
|
||||
default_args=default_args,
|
||||
description='E2E refactored pipeline: atomic steps with proper separation',
|
||||
schedule_interval='*/15 * * * *',
|
||||
start_date=days_ago(1),
|
||||
catchup=False,
|
||||
max_active_runs=1,
|
||||
tags=['pricing', 'elasticity', 'research', 'refactored'],
|
||||
) as dag:
|
||||
|
||||
# parallel data fetching
|
||||
t_fetch_interactions = PythonOperator(
|
||||
task_id='fetch_interactions',
|
||||
python_callable=fetch_interactions,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_fetch_price_logs = PythonOperator(
|
||||
task_id='fetch_price_logs',
|
||||
python_callable=fetch_price_logs,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# interaction processing branch
|
||||
t_create_buckets = PythonOperator(
|
||||
task_id='create_price_buckets',
|
||||
python_callable=create_price_buckets,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_augment_events = PythonOperator(
|
||||
task_id='augment_event_names',
|
||||
python_callable=augment_event_names,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_chunk_interactions = PythonOperator(
|
||||
task_id='chunk_interactions',
|
||||
python_callable=chunk_interactions,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_compute_demand = PythonOperator(
|
||||
task_id='compute_demand',
|
||||
python_callable=compute_demand,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# price processing branch (VECTORIZED)
|
||||
t_aggregate_prices = PythonOperator(
|
||||
task_id='aggregate_price_logs',
|
||||
python_callable=aggregate_price_logs,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# convergence: compute elasticity
|
||||
t_compute_elasticity = PythonOperator(
|
||||
task_id='compute_elasticity',
|
||||
python_callable=compute_elasticity,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# pricing tasks
|
||||
t_build_state = PythonOperator(
|
||||
task_id='build_state_space',
|
||||
python_callable=build_state_space,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_fit_pricer = PythonOperator(
|
||||
task_id='fit_pricing_function',
|
||||
python_callable=fit_pricing_function,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
t_predict_prices = PythonOperator(
|
||||
task_id='predict_prices',
|
||||
python_callable=predict_prices,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# publish to registry
|
||||
t_publish = PythonOperator(
|
||||
task_id='publish_results',
|
||||
python_callable=publish_results,
|
||||
provide_context=True,
|
||||
)
|
||||
|
||||
# dependency graph (clear atomic flow)
|
||||
# parallel fetches
|
||||
[t_fetch_interactions, t_fetch_price_logs]
|
||||
|
||||
# interaction branch: fetch -> bucket -> augment -> chunk -> demand
|
||||
t_fetch_interactions >> t_create_buckets >> t_augment_events >> t_chunk_interactions >> t_compute_demand
|
||||
|
||||
# price branch: fetch -> aggregate (vectorized)
|
||||
t_fetch_price_logs >> t_aggregate_prices
|
||||
|
||||
# convergence: both branches -> elasticity
|
||||
[t_compute_demand, t_aggregate_prices] >> t_compute_elasticity
|
||||
|
||||
# pricing: elasticity -> state + fit -> predict -> publish
|
||||
t_compute_elasticity >> [t_build_state, t_fit_pricer] >> t_predict_prices >> t_publish
|
||||
Reference in New Issue
Block a user