Platform For AI:A/B testing service integration

A/B testing integration

1. In the diagram, "Client" refers to your client-side service. You can configure experiments in the PAI-Rec console.

2. After you integrate the PAI-Rec engine, it automatically fetches experiment configurations on startup. When a recommendation request arrives, the SDK matches the request to an experiment and retrieves the corresponding parameters. The engine then adjusts its behavior accordingly, such as using different recall strategies or calling different ranking models.

3. PAI-Rec provides predefined experiment parameters. You must use the exact parameter names as specified. Otherwise, the engine cannot find the parameters even if an experiment is matched.

Parameter	Type	Description	Example
default.RecallNames	json array	All recall strategies to use.	"default.RecallNames":[ "HomepageEtrecRecall", "HomepageDssmRecall"]
"recall." + The name of the specific recall	json object	Configuration for a new recall.	{"recall.MyRecall":{"version":"v2"}}
filterNames	json array	Filter processes to apply.	{"filterNames":["UniqueFilter", "UserExposureFilter"]}
default.SortNames	json array	Re-ranking processes to apply.	{"default.SortNames":["ItemRankScore"]}
rankconf	recconf.RankConfig	Ranking algorithm configuration.	"rankconf":{"RankAlgoList":["pai_homepage_fm"],"RankScore":"${pai_homepage_fm}"}
features.scene.name	string	Recommendation scenario name for feature loading.	"homepagetf"
generalRankConf	recconf.GeneralRankConfig	Coarse ranking configuration, including user feature retrieval and the RankConf algorithm. For details, see Configure coarse ranking.	{"generalRankConf":{"FeatureLoadConfs":[{"FeatureDaoConf":{}}],"RankConf":{},"ActionConfs":[]}}
coldStartGeneralRankConf	recconf.ColdStartGeneralRankConfig	Coarse ranking configuration for cold start scenarios. For details, see Configure coarse ranking.	{"coldStartGeneralRankConf":{"FeatureLoadConfs":[{"FeatureDaoConf":{}}],"RankConf":{},"ActionConfs":[]}}
coldStartRankConf	recconf.ColdStartRankConfig	Ranking algorithm for cold start scenarios.	{"coldStartRankConf":{"RecallName":"ColdStartRecall", "AlgoName":"linucb"}}

        func (c *HomeFeedController) makeRecommendContext() {
        c.context = context.NewRecommendContext()
        c.context.Size = c.param.Limit
        c.context.Param = &c.param
        c.context.RecommendId = c.RequestId
        c.context.Config = recconf.Config
        c.context.Debug = c.param.Debug
        abcontext := model.ExperimentContext{
                Uid:         c.param.DistinctId,
                RequestId:   c.RequestId,
                FilterParams: map[string]interface{}{},
        }
        if abtest.GetExperimentClient() != nil {
                c.context.ExperimentResult = abtest.GetExperimentClient().MatchExperiment(c.param.SceneId, &abcontext)
                log.Info(c.context.ExperimentResult.Info())
        }
}

5. To adjust experiment parameters in your code, retrieve them from the context.ExperimentResult object within a RecommendContext object.

        count := r.recallCount
        if context.ExperimentResult != nil {
                count = context.ExperimentResult.GetLayerParams("").GetInt("recall.base.count", count)
        }
        fmt.Println("test count", count)

GetLayerParams retrieves experiment parameters from a specific layer. Use methods such as Get, GetInt, GetFloat, or GetInt64 to access a specific parameter. The first argument is the parameter name, and the second is a default value returned if the parameter is not found.

Use dynamic parameters

You can use custom parameters to change the engine's behavior at runtime without restarting the service, similar to experiment parameters.

In the left navigation pane, click system configuration > parameter management. Select the current development environment, such as Daily, and the recommendation scenario, such as HomePage. Then, click Create Parameter to add a custom parameter. For example, create a parameter with the name custom_param and the value custom.

// Get the name of the recommendation scenario.
scene := context.GetParameter("scene").(string)
// Retrieve the list of parameters for the scenario, and then use a Get* function to obtain a specific parameter value.
count := abtest.GetParams(scene).GetInt("count", 100)
fmt.Println("recall count:", count)