Conversational Interface
TalkOps provides a natural language interface for interacting with complex DevOps operations.
How It Works
Request Processing Flow
Intent Recognition
The system extracts structured data from natural language:
Example
| User Input | Extracted Intent |
|---|---|
| "Deploy my app to staging" | action: deploy, env: staging |
| "Scale Redis to 5 replicas" | action: scale, resource: redis, replicas: 5 |
| "Show Prometheus metrics" | action: query, source: prometheus |
| "Rollback last deployment" | action: rollback, target: last |
Multi-Intent Requests
Complex requests are decomposed:
"Deploy my app with monitoring and auto-scaling"
Intents:
├── deploy_application
├── configure_monitoring
└── configure_autoscaling
Conversation Context
TalkOps maintains context across interactions:
Context Types
| Type | Contents |
|---|---|
| Session | Current conversation history |
| User | Preferences, permissions, defaults |
| Environment | Active namespace, region, cluster |
| Workflow | Ongoing multi-step operations |
Context Usage
User: "Deploy prometheus"
Agent: Deployed to default namespace in us-east-1
User: "Now scale it to 3 replicas"
Agent: (knows "it" = prometheus from context)
Scaled prometheus to 3 replicas
Response Formats
Success Response
✅ Deployment Successful
Summary:
- App: my-service
- Environment: staging
- Replicas: 3
- Image: my-service:v1.2.3
Resources Created:
├── Deployment: my-service
├── Service: my-service-svc
└── ConfigMap: my-service-config
Next Steps:
1. View logs: "Show logs for my-service"
2. Check metrics: "Show my-service metrics"
Approval Required
⏸️ Approval Required
This operation requires review:
Changes:
├── Add: 3 x c5.xlarge instances
├── Modify: Auto-scaling policy
└── Estimated cost: +$450/month
Requested by: developer@example.com
Approvers: admin@example.com
[Approve] [Reject] [View Details]
Error Response
❌ Operation Failed
Error: Insufficient permissions
Details:
- Action: create_cluster
- Missing: eks:CreateCluster permission
Resolution:
1. Request access from administrator
2. Or use: "Deploy to staging" (allowed)
Need help? "Explain EKS permissions"
Clarification Dialogs
When requests are ambiguous:
User: "Deploy the app"
Agent: Which application would you like to deploy?
1. my-frontend (last deployed 2 days ago)
2. my-backend (last deployed 1 hour ago)
3. my-worker (never deployed)
User: "2"
Agent: Deploy my-backend. Which environment?
1. development
2. staging
3. production (requires approval)
Progressive Disclosure
Complex operations show summary first:
User: "Set up monitoring for my cluster"
Agent: Here's the monitoring plan:
📊 Metrics
├── Prometheus server
├── Node exporters
└── kube-state-metrics
📈 Dashboards
├── Cluster overview
├── Node health
└── Pod resources
🔔 Alerts
├── High CPU (>80%)
├── Memory pressure
└── Pod restarts
Estimated time: 5 minutes
Cost impact: +$25/month
[Deploy] [Customize] [Cancel]
Best Practices
For Clear Requests
- ✅ Be specific about resources
- ✅ Include environment/namespace
- ✅ Specify version when relevant
- ✅ Use consistent naming
Examples
| Less Clear | More Clear |
|---|---|
| "Deploy it" | "Deploy my-app to staging" |
| "Fix the issue" | "Rollback my-app to v1.2.3" |
| "More resources" | "Scale my-app to 5 replicas" |
| "Check things" | "Show health for production" |