Logging Interface Access

ChatGPT has no clue how Appian works.

Did you check whether the records usage log meets your requirements?

ChatGPT's approach isn't "wrong", it's just incomplete and can be misleading. Even 2025's latest LLMs are like the over-confident interns we're constantly correcting and rolling our eyes at. Yet in this case, its approach is possible.

For this to work

• You need a what's colloquially known as a "loopback" API call, where an Appian API is called via an Appian integration that uses an API key from the admin console. In theory it's possible to have the logged in user authenticate to the Appian API as themselves, but in practice it's a little more nuanced than that. Depending on your environment's security requirements, loopback API's may not be a good idea to do in production.
• API trace logging must be enabled in the admin console. Enabling this logging can cause a bit of overhead in production, and the logs will be overwritten every few days (or ~10MB) if there is a large volume of API calls.
• View perflogs/web_api_trace.csv

According to your prompt, you don't need to know who loaded the page, just that the page was loaded. If you need to know who loaded the page, there's a bit more development / complexity there.

If you absolutely must have this, it's pretty easy to create an Appian Web API that accepts key info (i.e. the username of the current user, and the name or some sort of identifier for the interface in question), which you can then call from the interface at load time in a local variable declaraion calling an Integration.  The API would simply launch a process model that quickly makes an entry for that username and that identifier in a DB table you have previously created, as well as a "loaded at" timestamp. 

I've implemented something like this for myself and decided to limit the DB table to 1 row per user per day per interface, so when it launches for a particular user, if there's already a row for that date/user/interface, it updates the existing row (incrementing an "updated" timestamp, as well as a "counter" that starts at 1 and increments for every unique load by that user).  This is slightly extra work of course but has been pretty useful to me and our management team in tracking overall site usage.

Thank you all for your advice.

Although I initially lacked some information, ultimately, I would like to calculate usage rates for each application from the log data. Therefore, I prefer to save the logs as records rather than text files. Additionally, since I need to calculate usage rates by organization, user information is also necessary.

Hence, I believe the approach Mike has already implemented is the closest to what I need.

I implemented the Web API as shown below. While it returns an image file as the response, it seems that neither a!startProcess, a!writeToDataStoreEntity, nor a!writeRecords functions are executing.

I implemented it using the GET method to return the image file. Is there anything else I should do?

a!localVariables(

  local!codea: 200,
  local!codeb: 200,
  local!codec: 200,

  local!a: a!writeToDataStoreEntity(
    dataStoreEntity: cons!XLOG_EVENT,
    valueToStore: 'type!{urn:com:appian:types:XLOG}XLOG_RT_EVENT'(
      timestamp: now(),
      appName: "Application Name",
      functionName: "Function Name",
      user: tostring(loggedInUser())
    ),
    onSuccess: a!save(local!codea, 201),
    onError: a!save(local!codea, 202)
  ),

  local!b: a!writeRecords(
    records: 'recordType!{}XLOG rtEvent'(
      'recordType!{}XLOG rtEvent.fields.{}timestamp': now(),
      'recordType!{}XLOG rtEvent.fields.{}appName': "Application Name RT",
      'recordType!{}XLOG rtEvent.fields.{}functionName': "Function Name RT",
      'recordType!{}XLOG rtEvent.fields.{}user': tostring(loggedInUser())
    ),
    onSuccess: a!save(local!codeb, 201),
    onError: a!save(local!codeb, 202)
  ),

  local!c: a!startProcess(
    processModel: cons!XLOG_pmTest,
    onSuccess: a!save(local!codec, 201),
    onError: a!save(local!codec, 202)
  ),

  local!d: a!richTextDisplayField(
    value: "code: " & local!codea & " " & local!codeb& " " & local!codec
  ),

  a!httpResponse(
    statusCode: local!codec,
    headers: {
      a!httpHeader(name: "Content-Type", value: "image/png"),
      a!httpHeader(name: "Cache-Control", value: "no-store, no-cache, must-revalidate, max-age=0"),
      a!httpHeader(name: "Pragma", value: "no-cache")
    },
    body: cons!XLOG_PNG
  )
)

NOTE:In this sample code, these local variables are only for debbuging.

You web API structure as per this sample code doesnt seem right! the smart servcies should not be related to a local variable. 

Please check the documentation for steps/considerations. In you use case you can create a web API that has a!startProcess() in it. Whatever you are trying to accomplish in a!writeRecord() and a!writeToDataStoreEntity() should be moved inside the process. 

The https Response configuration would also come within the onSuccess or onError of a!startProcess() configuration.

"Usage Rate" ... isn't this about users actually doing something vs. just staring at a page? User activity in Appian is closely tied to starting processes. When writing records, you can also write record events which then should correlate with "user activity".

Dear Stephan,

I generally agree with your opinion.
However, we are building a portal site for users and are considering displaying various information for them on the top page.
Since there are quite a few users for whom simply viewing the portal site is sufficient, we would like to also track who has accessed the portal site.

Dear Harsha,

Thank you for your advice.

I reviewed the documentation again, modified my code and confirmed that a!startProcess() and a!writeRecord can be executed.

However, as mentioned in my initial post,
(1) Create a Web API on Appian that returns an image as the response and simultaneously saves a log.
(2) Place an image on the target interface and set the endpoint of the Web API created in (1) as the image source.
(3) As a result, when the interface is opened, the Web API in (1) is called, and the log is recorded.

Since we intend to implement this behavior, the Web API must be implemented with the GET method.

As a result, we encountered the error:
"Smart Services cannot be executed in Web APIs that have a method of 'GET'."

and cannot proceed further.

Is there any way to work around this?

It is my understanding that portal users do not 'log in' to Appian, and therefore loggedInUser() will return the service account that is called from the portal rather than the person accessing the portal. You can probably add some metadata (headers, path, query param, etc) to your API call to get some sort of user info.

Hi Keizo,

It is not a hard requirement to write records from a Process Model. For APIs that are intended to be called frequently, it is better in practice to call a!writeRecords() from the API object itself to remove the overhead and latency of starting a process.

Keep in mind that, if enabled in the record types, you can write multiple related records at once if they are sync'ed records and have a relationship defined. For me, this has worked for 99% of my use cases. For the other 1%, I did something like the following: 

a!writeRecords(
    records: {Record A and its embedded related records}
    onSuccess: a!writeRecords(
        records: {Record not related to record A}
        onSuccess: a!httpResponse()
        onError: a!httpResponse()
    ),
    onError: a!httpResponse()
)