These prompts get reliable answers from the triaging agent in Koodisi AI chat. Replace anything in <angle brackets> with your own application, workflow, trace ID or identifier. For how to start a conversation and read the answers, see Investigating Integration Health.
Use Ask mode. The agent is read-only: it looks things up and explains them, and it never changes your application. If your team has the Koodisi AI bot in Webex, the same prompts work there.
Tips for good answers
- Name the application, the time window and the environment. If you leave out the environment, the agent uses Production.
- Give a timezone when the window matters, for example "between 10 PM ET yesterday and 10 PM ET today".
- Ask follow-ups in the same chat. A follow-up such as "now only for yesterday" or "how many of those are validation errors?" keeps the scope of the previous answer.
- A trace ID on its own is enough. You don't need to say which environment it ran in.
- Check the evidence. Every answer links to its traces.
Health and volume
| Prompt |
|---|
| How are my applications doing in production over the last 7 days? |
Break down <application> by workflow for the last 7 days: runs, errors and p95. |
| (follow-up) now only for yesterday |
How many runs did <application> have yesterday in production? |
Failures
| Prompt |
|---|
Show me the failed events for <application> in production in the last 7 days. |
How many workflow runs of <application> failed in production between <10:00 PM ET yesterday> and <10:00 PM ET today>? |
Show me the 3 most recent failed runs of the <workflow> workflow of <application> in the last 7 days. |
| (follow-up) how many of those are validation errors? |
Business identifiers
Any value you capture as an attribute tag on your activities is searchable.
| Prompt |
|---|
Is <identifier name> <value> in error? |
What's the status of <identifier name> <value> from <date>? |
Show me the failed events for <application> where <identifier name> is <value>. |
Run and record counts
| Prompt |
|---|
How many records did the <workflow> workflow of <application> process in production in the last 7 days? |
How many times did the <workflow> workflow of <application> run in production in the last 7 days? |
One trace: show it and explain it
| Prompt |
|---|
Show me trace <trace id> |
Why did trace <trace id> fail? |
The answer names the first step that failed, and the application, workflow, step and setting to change, as deployed in the version that ran.
What called a trace
| Prompt |
|---|
<trace id> what called this? |
For trace <trace id>, show me where in the parent this trace is being called. |
| (follow-up) Yes, open and show the exact parent step |
Mapping problems
| Prompt |
|---|
In <application> I mapped <field> and it is not showing up in the output. Why? |
Why is the date mapping failing in trace <trace id>, and what expression should I use? |
Suggested expressions use only real platform functions, and show their signatures.
Where a workflow is used
| Prompt |
|---|
| What workflows in this application reference the current workflow? |
Where is the <workflow> workflow used? |
Which workflows reference <workflow>? |
| List all the workflows in this application. |
Alerts