reinforcement_learning::live_model Class Reference

Interface class for the Inference API. More...

#include <live_model.h>

Public Types
using	error_fn = void(*)(const api_status &, void *)
	Error callback function. When live_model is constructed, a background error callback and a context (void*) is registered. If there is an error in the background thread, error callback will get invoked with api_status and the context (void*). More...
template<typename ErrCntxt >
using	error_fn_t = void(*)(const api_status &, ErrCntxt *)
	Error callback function. When live_model is constructed, a background error callback and a context (void*) is registered. If there is an error in the background thread, error callback will get invoked with api_status and the context (void*). This error callback is typed by the context used in the callback. More...

Public Member Functions
	live_model (const utility::configuration &config, error_fn fn=nullptr, void *err_context=nullptr, trace_logger_factory_t *trace_factory=&trace_logger_factory, data_transport_factory_t *t_factory=&data_transport_factory, model_factory_t *m_factory=&model_factory, sender_factory_t *s_factory=&sender_factory, time_provider_factory_t *time_prov_factory=&time_provider_factory)
	Construct a new live model object. More...
	live_model (const utility::configuration &config, std::function< void(const api_status &)> error_cb, trace_logger_factory_t *trace_factory=&trace_logger_factory, data_transport_factory_t *t_factory=&data_transport_factory, model_factory_t *m_factory=&model_factory, sender_factory_t *s_factory=&sender_factory, time_provider_factory_t *time_prov_factory=&time_provider_factory)
	Construct a new live model object. More...
int	init (api_status *status=nullptr)
	Initialize inference library. Initialize the library and start the background threads used for model managment and sending actions and outcomes to the online trainer. More...
int	choose_rank (const char *event_id, string_view context_json, ranking_response &resp, api_status *status=nullptr)
	Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. More...
int	choose_rank (string_view context_json, ranking_response &resp, api_status *status=nullptr)
	Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. A unique event_id will be generated and returned in the ranking_response. The same event_id should be used when reporting the outcome for this action. More...
int	choose_rank (const char *event_id, string_view context_json, unsigned int flags, ranking_response &resp, api_status *status=nullptr)
	Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. More...
int	choose_rank (string_view context_json, unsigned int flags, ranking_response &resp, api_status *status=nullptr)
	Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. A unique event_id will be generated and returned in the ranking_response. The same event_id should be used when reporting the outcome for this action. More...
int	request_continuous_action (const char *event_id, string_view context_json, unsigned int flags, continuous_action_response &response, api_status *status=nullptr)
	(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. The corresponding event_id should be used when reporting the outcome for the continuous action. More...
int	request_continuous_action (const char *event_id, string_view context_json, continuous_action_response &response, api_status *status=nullptr)
	(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. The corresponding event_id should be used when reporting the outcome for the continuous action. More...
int	request_continuous_action (string_view context_json, unsigned int flags, continuous_action_response &response, api_status *status=nullptr)
	(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. A unique event_id will be generated and returned in the continuous_action_response. The same event_id should be used when reporting the outcome for this action. More...
int	request_continuous_action (string_view context_json, continuous_action_response &response, api_status *status=nullptr)
	(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. A unique event_id will be generated and returned in the continuous_action_response. The same event_id should be used when reporting the outcome for this action. More...
int	request_decision (string_view context_json, unsigned int flags, decision_response &resp, api_status *status=nullptr)
	(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot feautres and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. A unique event_id can be supplied for each slot using the _id json field. The corresponding event_id should be used when reporting the outcome for each slot. More...
int	request_decision (string_view context_json, decision_response &resp, api_status *status=nullptr)
	(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot feautres and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. A unique event_id can be supplied for each slot using the _id json field. The corresponding event_id should be used when reporting the outcome for each slot. More...
int	request_multi_slot_decision (const char *event_id, string_view context_json, unsigned int flags, multi_slot_response &resp, api_status *status=nullptr)
	(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot features and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. The corresponding event_id should be used when reporting the outcome for each slot. More...
int	request_multi_slot_decision (const char *event_id, string_view context_json, multi_slot_response &resp, api_status *status=nullptr)
int	request_multi_slot_decision (string_view context_json, unsigned int flags, multi_slot_response &resp, api_status *status=nullptr)
int	request_multi_slot_decision (string_view context_json, multi_slot_response &resp, api_status *status=nullptr)
int	request_multi_slot_decision (const char *event_id, string_view context_json, unsigned int flags, multi_slot_response &resp, const int *baseline_actions, size_t baseline_actions_size, api_status *status=nullptr)
int	request_multi_slot_decision (const char *event_id, string_view context_json, unsigned int flags, multi_slot_response_detailed &resp, api_status *status=nullptr)
int	request_multi_slot_decision (const char *event_id, string_view context_json, multi_slot_response_detailed &resp, api_status *status=nullptr)
int	request_multi_slot_decision (string_view context_json, unsigned int flags, multi_slot_response_detailed &resp, api_status *status=nullptr)
int	request_multi_slot_decision (string_view context_json, multi_slot_response_detailed &resp, api_status *status=nullptr)
int	request_multi_slot_decision (const char *event_id, string_view context_json, unsigned int flags, multi_slot_response_detailed &resp, const int *baseline_actions, size_t baseline_actions_size, api_status *status=nullptr)
int	request_episodic_decision (const char *event_id, const char *previous_id, string_view context_json, ranking_response &resp, episode_state &episode, api_status *status=nullptr)
int	request_episodic_decision (const char *event_id, const char *previous_id, string_view context_json, unsigned int flags, ranking_response &resp, episode_state &episode, api_status *status=nullptr)
int	report_action_taken (const char *event_id, api_status *status=nullptr)
	Report that action was taken. More...
int	report_action_taken (const char *primary_id, const char *secondary_id, api_status *status=nullptr)
	Report that action was taken. More...
int	report_outcome (const char *event_id, const char *outcome, api_status *status=nullptr)
	Report the outcome for the top action. More...
int	report_outcome (const char *event_id, float outcome, api_status *status=nullptr)
	Report the outcome for the top action. More...
int	report_outcome (const char *primary_id, int secondary_id, float outcome, api_status *status=nullptr)
	Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot. More...
int	report_outcome (const char *primary_id, const char *secondary_id, float outcome, api_status *status=nullptr)
	Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot. More...
int	report_outcome (const char *primary_id, int secondary_id, const char *outcome, api_status *status=nullptr)
	Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot. More...
int	report_outcome (const char *primary_id, const char *secondary_id, const char *outcome, api_status *status=nullptr)
	Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot. More...
int	refresh_model (api_status *status=nullptr)
template<typename ErrCntxt >
	live_model (const utility::configuration &config, error_fn_t< ErrCntxt > fn, ErrCntxt *err_context=nullptr, trace_logger_factory_t *trace_factory=&trace_logger_factory, data_transport_factory_t *t_factory=&data_transport_factory, model_factory_t *m_factory=&model_factory, sender_factory_t *s_factory=&sender_factory, time_provider_factory_t *time_prov_factory=&time_provider_factory)
	Construct a new live model object. More...
	live_model (live_model &&other) noexcept
	Move constructor for live model object.
live_model &	operator= (live_model &&other) noexcept
	Move assignment operator swaps implementation.
	live_model (const live_model &)=delete
live_model &	operator= (live_model &)=delete
	Prevent accidental copy, since destructor will deallocate the implementation.
	~live_model ()
	Prevent accidental copy, since destructor will deallocate the implementation.

Detailed Description

Interface class for the Inference API.

• (1) Instantiate and Initialize
• (2) choose_rank() to choose an action from a list of actions
• (3) report_outcome() to provide feedback on chosen action

Member Typedef Documentation

error_fn

using reinforcement_learning::live_model::error_fn = void (*)(const api_status&, void*)

Error callback function. When live_model is constructed, a background error callback and a context (void*) is registered. If there is an error in the background thread, error callback will get invoked with api_status and the context (void*).

NOTE: Error callback will get invoked in a background thread.

error_fn_t

template<typename ErrCntxt >

using reinforcement_learning::live_model::error_fn_t = void (*)(const api_status&, ErrCntxt*)

Error callback function. When live_model is constructed, a background error callback and a context (void*) is registered. If there is an error in the background thread, error callback will get invoked with api_status and the context (void*). This error callback is typed by the context used in the callback.

NOTE: Error callback will get invoked in a background thread.

Template Parameters

ErrCntxt

Context type used when the error callback is invoked

Constructor & Destructor Documentation

live_model() [1/3]

reinforcement_learning::live_model::live_model ( const utility::configuration &  config, error_fn  fn = nullptr, void *  err_context = nullptr, trace_logger_factory_t *  trace_factory = &trace_logger_factory, data_transport_factory_t *  t_factory = &data_transport_factory, model_factory_t *  m_factory = &model_factory, sender_factory_t *  s_factory = &sender_factory, time_provider_factory_t *  time_prov_factory = &time_provider_factory  )

explicit

Construct a new live model object.

Parameters

config	Name-Value based configuration
fn	Error callback for handling errors in background thread
err_context	Context passed back during Error callback
t_factory	Transport factory. The default transport factory is initialized with a REST based transport that gets data from an Azure storage account
m_factory	Model factory. The default model factory hydrates vw models used for local inference.
sender_factory	Sender factory. The default factory provides two senders, one for interaction and the other for observation which logs to Event Hub.

live_model() [2/3]

reinforcement_learning::live_model::live_model ( const utility::configuration &  config, std::function< void(const api_status &)>  error_cb, trace_logger_factory_t *  trace_factory = &trace_logger_factory, data_transport_factory_t *  t_factory = &data_transport_factory, model_factory_t *  m_factory = &model_factory, sender_factory_t *  s_factory = &sender_factory, time_provider_factory_t *  time_prov_factory = &time_provider_factory  )

explicit

Construct a new live model object.

Parameters

config	Name-Value based configuration
error_cb	Error callback that takes no context
t_factory	Transport factory. The default transport factory is initialized with a REST based transport that gets data from an Azure storage account
m_factory	Model factory. The default model factory hydrates vw models used for local inference.
sender_factory	Sender factory. The default factory provides two senders, one for interaction and the other for observation which logs to Event Hub.

live_model() [3/3]

template<typename ErrCntxt >

reinforcement_learning::live_model::live_model ( const utility::configuration &  config, error_fn_t< ErrCntxt >  fn, ErrCntxt *  err_context = nullptr, trace_logger_factory_t *  trace_factory = &trace_logger_factory, data_transport_factory_t *  t_factory = &data_transport_factory, model_factory_t *  m_factory = &model_factory, sender_factory_t *  s_factory = &sender_factory, time_provider_factory_t *  time_prov_factory = &time_provider_factory  )

explicit

Construct a new live model object.

Template Parameters

ErrCntxt

Context type used in error callback.

Parameters

config	Name-Value based configuration
fn	Error callback for handling errors in background thread
err_context	Context passed back during Error callback
t_factory	Transport factory. The default transport factory is initialized with a REST based transport that gets data from an Azure storage account
m_factory	Model factory. The default model factory hydrates vw models used for local inference.
sender_factory	Sender factory. The default factory provides two senders, one for interaction and the other for observation which logs to Event Hub.

Template Parameters

ErrCntxt

Context type used in error callback.

Parameters

config	Name-Value based configuration
fn	Error callback for handling errors in background thread
err_context	Context passed back during Error callback
t_factory	Transport factory. The default transport factory is initialized with a REST based transport that gets data from an Azure storage account
m_factory	Model factory. The default model factory hydrates vw models used for local inference.
sender_factory	Sender factory. The default factory provides two senders, one for interaction and the other for observations which logs to Event Hub.

Member Function Documentation

choose_rank() [1/4]

int reinforcement_learning::live_model::choose_rank	(	const char *	event_id,
		string_view	context_json,
		ranking_response &	resp,
		api_status *	status = nullptr
	)

Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it.

Parameters

event_id	The unique identifier for this interaction. The same event_id should be used when reporting the outcome for this action.
context_json	Contains action, action features and context features in json format
resp	Ranking response contains the chosen action, probability distribution used for sampling actions and ranked actions
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

choose_rank() [2/4]

int reinforcement_learning::live_model::choose_rank	(	const char *	event_id,
		string_view	context_json,
		unsigned int	flags,
		ranking_response &	resp,
		api_status *	status = nullptr
	)

Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it.

Parameters

event_id	The unique identifier for this interaction. The same event_id should be used when reporting the outcome for this action.
context_json	Contains action, action features and context features in json format
flags	Action flags (see action_flags.h)
resp	Ranking response contains the chosen action, probability distribution used for sampling actions and ranked actions
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

choose_rank() [3/4]

int reinforcement_learning::live_model::choose_rank	(	string_view	context_json,
		ranking_response &	resp,
		api_status *	status = nullptr
	)

Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. A unique event_id will be generated and returned in the ranking_response. The same event_id should be used when reporting the outcome for this action.

Parameters

context_json	Contains action, action features and context features in json format
resp	Ranking response contains the chosen action, probability distribution used for sampling actions and ranked actions
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

choose_rank() [4/4]

int reinforcement_learning::live_model::choose_rank	(	string_view	context_json,
		unsigned int	flags,
		ranking_response &	resp,
		api_status *	status = nullptr
	)

Choose an action, given a list of actions, action features and context features. The inference library chooses an action by creating a probability distribution over the actions and then sampling from it. A unique event_id will be generated and returned in the ranking_response. The same event_id should be used when reporting the outcome for this action.

Parameters

context_json	Contains action, action features and context features in json format
flags	Action flags (see action_flags.h)
resp	Ranking response contains the chosen action, probability distribution used for sampling actions and ranked actions
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

init()

int reinforcement_learning::live_model::init

(

api_status *

status = nullptr

)

Initialize inference library. Initialize the library and start the background threads used for model managment and sending actions and outcomes to the online trainer.

Parameters

status

Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_action_taken() [1/2]

int reinforcement_learning::live_model::report_action_taken	(	const char *	event_id,
		api_status *	status = nullptr
	)

Report that action was taken.

Parameters

event_id	The unique event_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_action_taken() [2/2]

int reinforcement_learning::live_model::report_action_taken	(	const char *	primary_id,
		const char *	secondary_id,
		api_status *	status = nullptr
	)

Report that action was taken.

Parameters

primary_id	The unique primary_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
secondary_id	Index of the partial outcome.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [1/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	event_id,
		const char *	outcome,
		api_status *	status = nullptr
	)

Report the outcome for the top action.

Parameters

event_id	The unique event_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
outcome	Outcome serialized as a string
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [2/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	event_id,
		float	outcome,
		api_status *	status = nullptr
	)

Report the outcome for the top action.

Parameters

event_id	The unique event_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
outcome	Outcome as float
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [3/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	primary_id,
		const char *	secondary_id,
		const char *	outcome,
		api_status *	status = nullptr
	)

Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot.

Parameters

primary_id	The unique primary_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
secondary_id	Index of the partial outcome.
outcome	Outcome as float.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [4/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	primary_id,
		const char *	secondary_id,
		float	outcome,
		api_status *	status = nullptr
	)

Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot.

Parameters

primary_id	The unique primary_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
secondary_id	Index of the partial outcome.
outcome	Outcome as float.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [5/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	primary_id,
		int	secondary_id,
		const char *	outcome,
		api_status *	status = nullptr
	)

Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot.

Parameters

primary_id	The unique primary_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
secondary_id	Index of the partial outcome.
outcome	Outcome as float.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

report_outcome() [6/6]

int reinforcement_learning::live_model::report_outcome	(	const char *	primary_id,
		int	secondary_id,
		float	outcome,
		api_status *	status = nullptr
	)

Report outcome of a decision based on a pair of primary and secondary indentifiers. This identifier pair is problem specific. For CCB, the primary is the event id and the secondary is the index of the slot.

Parameters

primary_id	The unique primary_id used when choosing an action should be presented here. This is so that the action taken can be matched with feedback received.
secondary_id	Index of the partial outcome.
outcome	Outcome as float.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_continuous_action() [1/4]

int reinforcement_learning::live_model::request_continuous_action	(	const char *	event_id,
		string_view	context_json,
		continuous_action_response &	response,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. The corresponding event_id should be used when reporting the outcome for the continuous action.

Parameters

event_id	The unique identifier for this interaction. The same event_id should be used when reporting the outcome for this action.
context_json	Contains context features in json format
response	Continuous action response contains the chosen action and the probability density value of the chosen action location from the continuous range.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_continuous_action() [2/4]

int reinforcement_learning::live_model::request_continuous_action	(	const char *	event_id,
		string_view	context_json,
		unsigned int	flags,
		continuous_action_response &	response,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. The corresponding event_id should be used when reporting the outcome for the continuous action.

Parameters

event_id	The unique identifier for this interaction. The same event_id should be used when reporting the outcome for this action.
context_json	Contains context features in json format
flags	Action flags (see action_flags.h)
response	Continuous action response contains the chosen action and the probability density value of the chosen action location from the continuous range.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_continuous_action() [3/4]

int reinforcement_learning::live_model::request_continuous_action	(	string_view	context_json,
		continuous_action_response &	response,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. A unique event_id will be generated and returned in the continuous_action_response. The same event_id should be used when reporting the outcome for this action.

Parameters

context_json	Contains context features in json format
response	Continuous action response contains the chosen action and the probability density value of the chosen action location from the continuous range.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_continuous_action() [4/4]

int reinforcement_learning::live_model::request_continuous_action	(	string_view	context_json,
		unsigned int	flags,
		continuous_action_response &	response,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from a continuous range, given a list of context features The inference library chooses an action by sampling the probability density function produced per continuous action range. A unique event_id will be generated and returned in the continuous_action_response. The same event_id should be used when reporting the outcome for this action.

Parameters

context_json	Contains context features in json format
flags	Action flags (see action_flags.h)
response	Continuous action response contains the chosen action and the probability density value of the chosen action location from the continuous range.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_decision() [1/2]

int reinforcement_learning::live_model::request_decision	(	string_view	context_json,
		decision_response &	resp,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot feautres and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. A unique event_id can be supplied for each slot using the _id json field. The corresponding event_id should be used when reporting the outcome for each slot.

Parameters

context_json	Contains slots, slot_features, slot ids, actions, action features and context features in json format
resp	Decision response contains the chosen action per slot, probability distribution used for sampling actions and ranked actions.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_decision() [2/2]

int reinforcement_learning::live_model::request_decision	(	string_view	context_json,
		unsigned int	flags,
		decision_response &	resp,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot feautres and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. A unique event_id can be supplied for each slot using the _id json field. The corresponding event_id should be used when reporting the outcome for each slot.

Parameters

context_json	Contains slots, slot_features, slot ids, actions, action features and context features in json format
flags	Action flags (see action_flags.h)
resp	Decision response contains the chosen action per slot, probability distribution used for sampling actions and ranked actions.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

request_multi_slot_decision()

int reinforcement_learning::live_model::request_multi_slot_decision	(	const char *	event_id,
		string_view	context_json,
		unsigned int	flags,
		multi_slot_response &	resp,
		api_status *	status = nullptr
	)

(DEPRECATED) Choose an action from the given set for each slot, given a list of actions, slots, action features, slot features and context features. The inference library chooses an action per slot by sampling the probability distribution produced per slot. The corresponding event_id should be used when reporting the outcome for each slot.

Parameters

event_id	The unique identifier for this interaction. The same event_id should be used when reporting the outcome for this action.
context_json	Contains slots, slot_features, slot ids, actions, action features and context features in json format
flags	Action flags (see action_flags.h)
resp	Decision response contains the chosen action per slot, probability distribution used for sampling actions and ranked actions.
status	Optional field with detailed string description if there is an error

Returns

int Return error code. This will also be returned in the api_status object

---

The documentation for this class was generated from the following file:

• /home/runner/work/reinforcement_learning/reinforcement_learning/include/live_model.h