Store LLM output and update events timeline #59
Conversation
benakansara
commented
Oct 28, 2024
benakansara
commented
- Stores LLM output into investigation Saved Object once analysis is completed
- If there is a stored LLM output on investigation SO, the events timeline will show significant events found by LLM
benakansara
changed the title
| @@ -24,6 +24,7 @@ const updateInvestigationParamsSchema = z.object({ | |||
| }), | |||
| tags: z.array(z.string()), | |||
| externalIncidentUrl: z.string().nullable(), | |||
| automatedRcaAnalysis: z.array(z.any()).optional(), | |||
There was a problem hiding this comment.
let's make this an object: { events: z.array(z.any()) } so we can easily add more top-level properties later
| import { AnnotationEvent } from './annotation_event'; | ||
| import { TIME_LINE_THEME } from './timeline_theme'; | ||
| import { useFetchEvents } from '../../../../hooks/use_fetch_events'; | ||
| import { useInvestigation } from '../../contexts/investigation_context'; | ||
| import { useKibana } from '../../../../hooks/use_kibana'; | ||
| import { AlertEvent } from './alert_event'; | ||
| import { LLMEvent } from './llm_event'; |
There was a problem hiding this comment.
this should probably be more specific. maybe something like RootCauseAnalysisEventAnnotation? also, if you want the root cause analysis event types, those types are exported from service_rca/index.ts iirc.
There was a problem hiding this comment.
I have renamed LLMEvent to RootCauseAnalysisEvent, and event type from llm to rca. I am mapping RCA significant event here to existing one we are using for events timeline.
| const [loading, setLoading] = useState(false); | ||
|
|
||
| useEffect(() => { |
There was a problem hiding this comment.
It looks like I need to keep useEffect that I previously reverted. If I update investigation SO in complete function, it doesn't get updated properly every time.
| 'response' in event && 'report' in event.response && 'timeline' in event.response | ||
| ) | ||
| ) { | ||
| await updateInvestigation({ |
There was a problem hiding this comment.
Sorry, I missed this, but we need to do it on the server, not on the client. Otherwise the user needs to keep the browser tab open.
There was a problem hiding this comment.
I didn't get why the user needs to keep browser tab open? It needs to be open only first time when storing the results to SO, afterwards it will just fetch events from SO unless you run the analysis again in which case it will rewrite results to SO.
There was a problem hiding this comment.
Yes that is the whole point, if the user runs it, they shouldn't be required to keep the browser tab open.
There was a problem hiding this comment.
The reason for storing LLM result was that when you visit the page again, you don't have to run the analysis and wait for few minutes. But when you run the analysis for the first time, you would want to see the result, isn't it? So imo it's ok to keep the browser tab open whenever new analysis is being run. Another aspect is that if user closes the tab, that would mean they don't want to continue with the analysis, right? We should not carry on with the requests to LLM in that case.
There was a problem hiding this comment.
No, as I said before, the point is that the user needs to be able to navigate away from the page and the results will be ready when they get back. Why ask them to keep a browser tab open for minutes and make sure they stay on the same actual page?
