A Picture Is Worth A Thousand Words: Dashboards To The Rescue

Building an application in the age of AI assisted development is only half the battle.

With an APEX application hosted on OCI, monitoring resource consumption is vital, because with the free tier you cannot exceed certain volumes of requests, whether they are REST requests or emails or notifications.

A well-crafted dashboard, beyond just eye-catching aesthetics, can help you spot problems that might otherwise go unnoticed until the application runs out of resources and stops working. 

The position of the big blue circle shows there is something wrong with the embedding calls 

Spotting anomalies at a glance:

A prime example of a dashboard's utility is how it highlights hidden inefficiencies and runaway processes.
In the API requests visualization, in the lower left portion of the image, a massive blue circle was dominating the chart, immediately drawing my attention.

• The big blue bubble instantly signals an anomaly in the system's traffic pattern. 
• The position of its centre on the Y axis compared to the position of other types of requests over the week reveals that the application is currently executing a disproportionate number of embedding requests to the LLM. 
• Looking at the chart I could immediately start investigating the rogue document processing procedure.

SELECT 
  -- Chronological X: 6 = 6 days ago, 0 = Today
  to_char(w.request_date,'Day') as dy,
  (trunc(w.request_date) - trunc(sysdate)) as x_value,
  nvl(regexp_substr(w.url,'(http(s)?:/)(/([^/]+))+',1,1,'c',4), 'UNKNOWN') as series,
  count(*)                                     as requests,    -- Y-axis
  sum(elapsed_sec)                             as elapsed,     -- Bubble radius
  to_char(w.request_date,'Day') || '
Requests: ' || count(*) || '
Time: ' || sum(elapsed_sec) as custom_tooltip
FROM apex_webservice_log w
WHERE w.request_date >= trunc(sysdate) - 6 and w.request_date < trunc(sysdate) + 1
GROUP BY to_char(w.request_date,'Day'),
         (trunc(w.request_date) - trunc(sysdate)), 
         nvl(regexp_substr(w.url,'(http(s)?:/)(/([^/]+))+',1,1,'c',4), 'UNKNOWN')
ORDER BY x_value desc, series;

Dashboards can bridge the gap between hundreds of log entries in an APEX internal view and the need to extract and present the significant information in an efficient and concise way. 

Last but not least, AI can suggest the critical resources to monitor and how to best represent the information with a chart if you are short of ideas.