Introduction
The Parameter Access and Additional Corpora components of a Deployment can be configured to get their data from a different runtime which is configured to get the data from MongoDB, Cassandra or Presto. This enables two use cases:
- If multiple runtime cases are configured to connect to the same tables, passing the data requests through a single runtime will reduce the number of connections opened to the database.
- Each runtime can connect to a single instance of MongoDB, Cassandra or Presto. If data from multiple runtimes is required in a single runtime, then that data can be accessed through the runtime that is connected to desired instance.
Below we outline the implementation of each of these approaches to highlight to advantages and disadvantages of each approach.
🔥
Note
🔥
Note
Configure a runtime to pass data
The first step is to configure a runtime to get the data from the data source of interest. Both Parameter Access and Additional Copora can be configured. PostScoreProvideData should be configured as the post scoring logic in order to generate an API response containing the data in the expected format. Below we show an example of how the runtime can be configured using the python package:
from prediction.apis import deployment_management as dm
from prediction.apis import ecosystem_generation_engine as ge
# The name of the project and deployment on which you will be working.
project_id = "Example Project"
deployment_id_lookup = "external_data_lookup"
runtime_path_lookup="http://ecosystem-runtime:8091"
# The database from which the data will be pulled
db = "recommender_demos"
#Configure the lookup to the customer feature store
parameter_access_lookup = dm.define_deployment_parameter_access(
auth,
lookup_key="customer",
lookup_type="int",
database=db,
table_collection="offer_feature_store",
datasource="mongodb"
)
#Configure the additional corpora
additional_corpora = [
{"name":"location_details","database":"mongodb","db":"recommender_demos","table":"location_information","type":"dynamic","key":"location"}
,{"name":"device_details","database":"mongodb","db":"recommender_demos","table":"device_info","type":"static","key":"device"}
]
#Create a deployment
version = "001"
deployment_step = dm.create_deployment(
auth,
project_id=project_id,
deployment_id=deployment_id_lookup,
description="External data lookup",
version=version,
plugin_post_score_class="PostScoreProvideData.java",
scoring_engine_path_dev=runtime_path_lookup,
mongo_connect=f"mongodb://ecosystem_user:{mongo_password}@ecosystem-server:54445/?authSource=admin",
parameter_access=parameter_access_lookup,
corpora=additional_corpora,
)
#Push deployment and make a test call to validate the results
deployment_step = dm.get_deployment_step(auth, project_id, deployment_id_lookup, version,project_status="experiment")
push_result = ge.process_push(auth,deployment_step)
if "ErrorMessage" in push_result:
print(push_result["ErrorMessage"])
raise ValueError("Push failed")
#Test your deployment
post_invocations_input = {
"campaign": deployment_id_lookup
, "subcampaign": "none"
, "channel": "notebooks"
, "customer": 793
, "userid": "test"
, "numberoffers": 1
, "params": "{additional_corpora:[location_details,device_details]}"
}
offer_response = o.invocations(auth_lookup, post_invocations_input)
print(offer_response)Configure a runtime to use to passed data
To use data from another runtime, Parameter Access and Additional Corpora should be configured to use runtime as a Data Source and the URL of the runtime to get the data from should be configured. Below we show an example of how the runtime can be configured using the python package:
from prediction.apis import deployment_management as dm
from prediction.apis import ecosystem_generation_engine as ge
from prediction.apis import online_learning_management as ol
#The name of the project and deployment on which you will be working.
deployment_id = "offer_recommend_dynamic"
runtime_path="http://ecosystem-runtime2:8092"
#Get a dynamic interaction configuration to add to the deployment
dynamic_interaction_uuid = ol.get_dynamic_interaction_uuid(auth,deployment_id)
dynamic_interaction = dm.define_deployment_multi_armed_bandit(epsilon=0, dynamic_interaction_uuid=dynamic_interaction_uuid)
#Configure the lookup to the customer feature store
parameter_access = dm.define_deployment_parameter_access(
auth,
lookup_key="customer",
lookup_type="int",
datasource="runtime",
url=runtime_path_lookup,
lookup_fields=parameter_access_lookup["lookup_fields"]
)
#Configure the additional corpora
additional_corpora = [
{"name":"location_details","database":"runtime","url":"http://ecosystem-runtime:8091","type":"dynamic","key":"location"}
,{"name":"device_details","url":"http://ecosystem-runtime:8091","database":"runtime","type":"static","key":"device"}
]
#Create your deployment
version = "001"
deployment_step = dm.create_deployment(
auth,
project_id=project_id,
deployment_id=deployment_id,
description="Ecosystem Rewards algorithm getting data from external source",
version=version,
plugin_post_score_class="PlatformDynamicEngagement.java",
plugin_pre_score_class="PreScoreDynamic.java",
scoring_engine_path_dev=runtime_path,
mongo_connect=f"mongodb://ecosystem_user:{mongo_password}@ecosystem-server:54445/?authSource=admin",
parameter_access=parameter_access,
multi_armed_bandit=dynamic_interaction,
corpora=additional_corpora
)
#Push deployment and make a test call to validate the results
deployment_step = dm.get_deployment_step(auth, project_id, deployment_id, version,project_status="experiment")
push_result = ge.process_push(auth,deployment_step)
if "ErrorMessage" in push_result:
print(push_result["ErrorMessage"])
raise ValueError("Push failed")
#Test your deployment
post_invocations_input = {
"campaign": deployment_id_lookup
, "subcampaign": "none"
, "channel": "notebooks"
, "customer": 793
, "userid": "test"
, "numberoffers": 1
, "params": "{}"
}
offer_response = o.invocations(auth_runtime, post_invocations_input)
pp.pprint(offer_response)The properties file that is generated by pushing the configured Deployment will contain the following entries:
predictor.param.lookup={predictor:'offer_recommend_dynamic',mojo:1,database:'runtime',db:'default',table:'default',url:'http://ecosystem-runtime:8091',lookup:{"value":123,"key":"customer"},result:{parm1:'field1', parm2:'field2'}}
predictor.corpora=[{"database":"runtime","name":"location_details","type":"dynamic","url":"http://ecosystem-runtime:8091","key":"location"},{"database":"runtime","name":"device_details","type":"static","url":"http://ecosystem-runtime:8091","key":"device"},{"database":"mongodb","name":"dynamic_engagement","update":true,"type":"dynamic_engagement","uuid":"1db3c3b2-a140-4af5-a26e-91ca2bc59022","db":"ecosystem_meta","table":"dynamic_engagement"},{"database":"mongodb","name":"dynamic_engagement","update":true,"type":"dynamic_engagement_options","uuid":"1db3c3b2-a140-4af5-a26e-91ca2bc59022","db":"recommender_demos","table":"dynamic_set_up_feature_store_options"}]