AI API¶

API endpoints for the AI assistant.

Chat¶

Send Message¶

POST /api/v1/ai/chat

Request Body:

{
  "message": "How many pods are running?",
  "context": "optional additional context"
}

Response:

{
  "response": "Currently there are 47 pods running across 8 namespaces...",
  "sources": ["kubernetes"],
  "timestamp": "2024-01-15T10:30:00Z"
}

Example Queries¶

Kubernetes¶

{"message": "Show me all failed pods"}
{"message": "What's the CPU usage across nodes?"}
{"message": "List deployments in production namespace"}

Security¶

{"message": "What's my security score?"}
{"message": "Show critical vulnerabilities"}
{"message": "Are there privileged containers?"}

Health¶

Check AI Service Health¶

GET /api/v1/ai/health

Response:

{
  "status": "available",
  "model": "gemini-1.5-flash",
  "services": {
    "kubernetes": "connected",
    "security": "connected",
    "jenkins": "disconnected",
    "helm": "connected"
  }
}

Query Types¶

The AI automatically detects and fetches data for:

Type	Keywords
pods	pod, pods, running pods
deployments	deployment, replicas, rollout
services	service, svc, endpoints
nodes	node, worker, master
security	security, vulnerability, CVE
jenkins	build, pipeline, job
helm	chart, release, helm

Rate Limiting¶

AI endpoints have stricter rate limits:

20 requests per minute per user
429 Too Many Requests when exceeded

Error Responses¶

Model Not Available¶

{
  "detail": "AI model not available",
  "status_code": 503
}

Invalid Request¶

{
  "detail": "Message cannot be empty",
  "status_code": 400
}

Best Practices¶

Be Specific - Include namespace or resource names
Ask for Actions - Request kubectl commands
Use Context - Provide additional context when helpful
Follow Up - Ask clarifying questions