We can generate a data exhaust, which is a list of conversations between the bot and the user. When the bot and user converse, our service sends an event (AKA telemetry event) to a third-party service, which then uploads it to the blob storage.
When a message is sent to the user and the user replies to it, a telemetry event will be generated and will be sent to the third-party service. We can then download a list of these events for a specific conversation for a specific time period using the APIs mentioned in the document.
The telemetry event sent is generated according to the specification mentioned in the link.
Data Exhaust Requests
We can generate a job for two types of exhaust requests.
1. UCI Response Exhaust
This will provide a list of conversations done by any user between the time period mentioned in the exhaust request. Each event consists of the user device id, question asked, question options if any, question reply from the user, timestamp, etc. A conversation event is basically a telemetry event that we send for each conversation message. A sample for the UCI response exhaust
2. UCI Private Exhaust
This will provide a list of users that have conversed with the bot. It consists of user phone number(If consented by user), device id, username etc. The device id is the one that connects the response & private exhaust. To identify the phone number for the conversation done by the user, check the device id from the response exhaust, it should exists in the private exhaust. A sample for the UCI private exhaust
Data Exhaust Apis
1. API to submit a job request for response/private exhaust
This API will submit a job request to fetch the response/private exhaust for a specific conversation(bot) id based on the dataset config. The API will give a requestID in the response. The requestID & tag are used to fetch the job info.
URL:
{{host}}/dataset/v1/request/submi
Method: POST
Sample Request Body:
{
"id": "ekstep.analytics.job.request.submit",
"ver": "1.0",
"ts": "2022-02-15T10:07:00+05:30",
"params": {
"msgid": "06fb2660-88ba-11ec-bea0-578c52d1ac65"
},
"request": {
"dataset": "uci-response-exhaust", // dataset should be "uci-private-exhaust" or "uci-response-exhaust"
"tag": "7ed871f7-0d5b-4bb0-86a0-a00151352706", // random UUID
"requestedBy":"user_id",
"requestedChannel": "01309282781705830427", // channel id for the bot
"datasetConfig": {
"type": "uci-response-exhaust", // type should be "uci-private-exhaust" or "uci-response-exhaust"
"conversationId": "c63b6faa-11a9-4259-b941-11cd2f8446c6", // bot id
"startDate": "2022-03-01", // fetch the response data from date
"endDate": "2022-04-01" // fetch the response data to date
},
"output_format":"csv" //output format
}
}
Sample Response
{
"id": "ekstep.analytics.dataset.request.submit",
"ver": "1.0",
"ts": "2022-04-06T09:32:00.079+00:00",
"params": {
"resmsgid": "d15de3c9-875d-4fbd-912c-4daa1cee0f1b",
"status": "successful",
"client_key": null
},
"responseCode": "OK",
"result": {
"attempts": 0,
"lastUpdated": 1649237520065,
"tag": "7ed871f7-0d5b-4bb0-86a0-a00151352706:01269878797503692810", // Tag to be used in job info api, but without the channel id
"expiresAt": 1649239320077,
"datasetConfig": {
"type": "uci-response-exhaust",
"conversationId": "c63b6faa-11a9-4259-b941-11cd2f8446c6",
"startDate": "2022-03-01",
"endDate": "2022-04-01"
},
"downloadUrls": [],
"requestedBy": "user_id",
"jobStats": {
"dtJobSubmitted": 1649237520065,
"dtJobCompleted": null,
"executionTime": null
},
"status": "SUBMITTED",
"dataset": "uci-response-exhaust",
"requestId": "1E2A13DA30671FDBA10C418C9EF14A70", // Request id to be used in job info api
"requestedChannel": "01269878797503692810"
}
}
2. API to get job info of a job
This API is used to get the job info which includes the job status (Eg. Submitted, Success, Failed), download URLs, (if the job is successful), job request param etc. Tag & Request Id for the request will be used from the submitted job request.
The download url is a signed url which is valid for one day only.
The jobs will be executed via a cron job that runs daily at 02.00 AM. We should check the job info, the next day to check the job's current status.
If we need an immediate result we can also run a job from the Jenkins AnalyticsReplayJobs service. Follow the below step to execute the jobs.
Login to Jenkins & Go to the AnalyticsReplayJobs service from the Deploy->Dev->DataPipeline services list or directly hit the url.
Go to Build with Parameter.
Select run-job from the job_type drop down.
Leave the rest as it is & click build. It will run all the submitted jobs.
When it's done, run the job info api to fetch the job response. The job will have either a success/failed status. If the job is successful it will have a download url in the response.
