Get the AI analysis
Tool: oppex.incident.getAnalysis
Returns the Oppex AI Responder's analysis of one incident, together with the log lines it reasoned over. A coding assistant can use it to decide whether the incident is a bug it can fix in your code, and where to start looking.
What you provide
| Input | Required | Description |
|---|---|---|
| Incident ID | Yes | The incident whose analysis you want. |
Example prompts
- "What does Oppex think caused the Hyperdrive Failure incident?"
- "Get the AI analysis for incident cacgichh-efbg-jv3t1ufo and tell me whether it's a code bug."
- "Read the analysis for the checkout latency incident, find the failing code in this repository, and propose a fix."
Who can use it
Anyone, for incidents in the connection's workspace.
What you get back
| Part | Meaning |
|---|---|
| Category | CODE (an application bug), INFRA, CONFIG, DATA or UNKNOWN. Only CODE points to a change in your code. |
| Summary and likely root cause | What happened and why, in plain text. |
| Severity assessment | How serious the analysis judges the incident to be. |
| Suggested actions | Steps worth trying. |
| Confidence | How sure the analysis is. It also says when it needs more data. |
| Evidence services and search hints | The services the log lines came from, and names worth searching for in your code, such as the class and line that logged the error. |
| Related services | Other services involved in the failure. Often empty. |
| Log lines | The log lines the analysis used, up to 60,000 characters. |
Treat suggested actions as a starting point
The analysis is drawn from logs alone. Check a suggested fix against your code, and check whether it has already shipped, before you act on it.
Limits and errors
| Message | Meaning |
|---|---|
| "No AI analysis available for this incident in your current workspace" | The analysis has not finished yet, the AI Responder did not analyze this incident, or the incident is in another workspace. |