AI Assistant

The AI Assistant is an intelligent chat-based interface that helps you interact with the Cloud. Ask questions about your configurations, security events, troubleshooting, and platform operations using natural language.

Quick Start

Your First Query

Ask a simple question to get started:

xcsh ai query "What load balancers do I have?"

Example Output:

Found 3 HTTP load balancers:
- lb-prod-frontend (production) - ACTIVE
- lb-staging-api (staging) - ACTIVE
- lb-dev-test (development) - INACTIVE

Follow-up Questions:
  1. Show details for lb-prod-frontend
  2. Which load balancers are inactive?
  3. Create a new load balancer

Starting an Interactive Chat

For multi-turn conversations, use chat mode:

xcsh ai chat

=== F5 XC AI Assistant Chat ===
Namespace: default
Type /help for commands, /exit to quit.

ai> What security events occurred today?
...

ai> 1
Following up: Show me the most severe events...

Submitting Feedback

Help improve the AI by providing feedback:

# Positive feedback on last response
xcsh ai feedback --positive

# Negative feedback with reason
xcsh ai feedback --negative inaccurate --comment "Data was outdated"

---

Command Reference

ai query - Single Questions

Ask one-off questions without entering interactive mode.

Syntax:

xcsh ai query "<question>" [options]

Options:

Flag	Short	Description
--namespace	-ns	Target namespace (default: session namespace)
--output	-o	Output format: json, yaml, table, text, tsv, none
--spec	-	Output command specification as JSON

Aliases: ask, q

Examples:

# Basic query
xcsh ai query "How many sites are active?"

# Query with JSON output
xcsh ai query "List all origin pools" -o json

# Query in specific namespace
xcsh ai query "What WAF policies exist?" -ns production

# Short alias
xcsh ai q "Show my tenant configuration"

---

ai chat - Interactive Conversations

Enter a multi-turn conversation with context preservation.

Syntax:

xcsh ai chat [options]

Options:

Flag	Short	Description
--namespace	-ns	Default namespace for the session
--spec	-	Output command specification as JSON

Aliases: interactive, i

Built-in Chat Commands:

Command	Aliases	Description
/help	/h	Show available commands
/clear	/c	Clear conversation context
/exit	/quit, /q	Exit chat mode
/feedback <type>	-	Submit feedback on last response
1, 2, 3…	-	Select follow-up question by number

Examples:

# Start chat in default namespace
xcsh ai chat

# Start chat in production namespace
xcsh ai chat -ns production

Chat Session Example:

=== AI Assistant Chat ===
Namespace: production
Type /help for commands, /exit to quit.

ai> What's the status of my load balancers?

Load Balancer Status Summary
Active: 5
Degraded: 1
Inactive: 2

Follow-up Questions:
  1. Which load balancer is degraded?
  2. Show details for inactive load balancers
  3. How do I troubleshoot degraded status?

ai> 1
Following up: Which load balancer is degraded?

lb-api-gateway is showing degraded status:
- Health check failures on 2 of 5 origin servers
- Origin pool: api-backend-pool
- Last healthy: 2026-01-02 09:45:00 UTC

Follow-up Questions:
  1. Show health check details
  2. How do I fix origin server issues?
  3. Remove unhealthy origins temporarily

ai> /feedback positive
Positive feedback submitted. Thank you!

ai> /clear
Conversation context cleared.

ai> /exit
Exiting chat mode.

---

ai feedback - Improving Responses

Submit feedback to help improve AI response quality.

Syntax:

xcsh ai feedback <--positive | --negative <type>> [options]

Options:

Flag	Short	Description
--positive	-p	Submit positive feedback
--negative	-n	Submit negative feedback with reason type
--comment	-c	Add optional comment text
--query-id	-q	Target specific query (default: last query)
--namespace	-ns	Namespace context
--output	-o	Output format

Negative Feedback Types:

Type	When to Use
other	General feedback not fitting other categories
inaccurate	Response contained incorrect information
irrelevant	Response didn’t address the question
poor_format	Response was hard to read or understand
slow	Response took too long

Aliases: fb, rate

Examples:

# Quick positive feedback
xcsh ai feedback -p

# Negative with type
xcsh ai feedback --negative inaccurate

# Detailed negative feedback
xcsh ai feedback -n irrelevant -c "I asked about WAF but got DNS info"

# Feedback for specific query
xcsh ai feedback -p --query-id "qid-abc123"

---

ai eval - RBAC Testing Mode

Test AI queries under different access permissions for RBAC validation.

eval query

xcsh ai eval query "<question>" [options]

Sends queries through the eval endpoint with [EVAL MODE] context. Used for testing what responses would be returned for different permission levels.

eval feedback

xcsh ai eval feedback <--positive | --negative <type>> [options]

Submit feedback for eval queries on a separate analytics track.

Examples:

# Test query as different user context
xcsh ai eval query "List all namespaces"

# Feedback on eval response
xcsh ai eval feedback --positive

---

Response Types Explained

The AI assistant returns different response types based on your question:

Security Event Explanations

Detailed analysis of WAF, bot, and security events.

xcsh ai query "Why was request xyz blocked?"

Security Event Analysis
Summary: WAF blocked request due to SQL injection attempt

Request Details:
  Method: GET
  Path: /api/users?id=1 OR 1=1
  Source IP: 192.168.1.100
  Time: 2026-01-02 10:30:00 UTC

Violations:
  - SQL_INJECTION: UNION SELECT pattern detected in query parameter
  - ANOMALY_SCORE: Score 15 exceeded threshold of 5

Action: BLOCK
Confidence: HIGH

Threat Campaigns:
  - SQLi-Campaign-2024-001

List Responses

Resource listings with counts and summaries.

xcsh ai query "List my origin pools"

Found 4 origin pools:

  NAME              NAMESPACE    SERVERS    STATUS
  api-backend       production   3          HEALTHY
  web-frontend      production   5          HEALTHY
  staging-pool      staging      2          DEGRADED
  dev-pool          development  1          INACTIVE

Total: 4 pools (2 healthy, 1 degraded, 1 inactive)

Site Analysis Reports

Health analysis with metrics and recommendations.

xcsh ai query "Analyze site-1 health"

Site Analysis: site-1
Health Status: DEGRADED

Metrics:
  Availability:     98.5%
  Latency (P99):    245ms
  Error Rate:       1.2%
  Throughput:       1.5K rps

Recommendations:
  1. Consider scaling up to handle current traffic levels
  2. Investigate high latency - check origin server response times
  3. Review error logs for the 1.2% error rate

Dashboard Filters

Filter expressions for use in dashboards.

xcsh ai query "Create a filter for blocked requests from today"

Dashboard Filter Generated:

  Expression: waf_action:BLOCK AND timestamp:[now-24h TO now]
  Context: Security Events Dashboard

You can apply this filter in the F5 XC Console under:
  Dashboards > Security > Events

Widget Data Tables

Tabular data for dashboard visualization.

xcsh ai query "Show traffic summary by load balancer"

Traffic Summary
  LOAD BALANCER     REQUESTS    AVAILABILITY
  lb-prod           1.2M        99.9%
  lb-staging        45K         98.5%
  lb-dev            2.3K        100.0%

---

Working with Context

Multi-Turn Conversations

In chat mode, the AI maintains context between messages:

ai> Show me production load balancers
[Lists 3 production LBs]

ai> Which one has the most traffic?
[Knows you're asking about the 3 production LBs just listed]

ai> Show its origin pool configuration
[Knows "its" refers to the highest-traffic LB]

Follow-Up Suggestions

After each response, the AI suggests relevant follow-up questions:

Follow-up Questions:
  1. Show details for this load balancer
  2. View health check configuration
  3. How do I update the origin pool?

Select by typing the number:

ai> 2
Following up: View health check configuration...

Clearing Context

If the conversation becomes confusing, clear and start fresh:

ai> /clear
Conversation context cleared.

Namespace Scoping

Queries are scoped to a namespace for relevant results:

# Set namespace at start
xcsh ai chat -ns production

# Or use flag per query
xcsh ai query "List resources" -ns staging

---

Best Practices

Crafting Effective Queries

Be Specific:

# Good - specific and actionable
xcsh ai query "Why is lb-prod-api returning 502 errors since 10am?"

# Less effective - too vague
xcsh ai query "Why errors?"

Provide Context:

# Good - includes relevant details
xcsh ai query "Explain WAF event with request ID req-abc123 from today"

# Less effective - missing specifics
xcsh ai query "Explain security event"

Ask One Thing at a Time:

# Good - single focused question
xcsh ai query "How do I add a new origin server to my-pool?"

# Less effective - multiple questions
xcsh ai query "How do I add origins and configure health checks and set up routing?"

When to Use Chat vs Query

Use ai query when:

• Asking a single question
• Scripting or automation
• Piping output to other commands
• Quick lookups

Use ai chat when:

• Exploring a topic
• Troubleshooting (needs back-and-forth)
• Learning about features
• Working through a complex task

Providing Useful Feedback

Good feedback helps improve the AI:

# Specific negative feedback
xcsh ai feedback -n inaccurate -c "The API endpoint mentioned doesn't exist in v2"

# Always provide feedback for helpful responses
xcsh ai feedback -p -c "Clear step-by-step instructions"

---

Troubleshooting

”Not connected to API”

Cause: No active connection to F5 XC API.

Solution:

# Set environment variables
export F5XC_API_URL="https://your-tenant.console.ves.volterra.io"
export F5XC_API_TOKEN="your-api-token"

# Or start xcsh and configure
xcsh
> login

“Not authenticated”

Cause: API token is invalid or expired.

Solution:

1. Generate a new API token in F5 XC Console

2. Update your credentials:

   export F5XC_API_TOKEN="new-token"

“Chat mode requires an interactive terminal”

Cause: Running chat mode in a non-interactive environment (pipe, script, etc.)

Solution: Use ai query instead for non-interactive use:

# In scripts, use query mode
xcsh ai query "Your question" -o json

Empty or Unexpected Responses

Possible Causes:

1. Wrong namespace: Check you’re querying the right namespace
2. No matching resources: The query might be correct but there’s nothing to show
3. Permission issues: Your token might lack access to certain resources

Solution:

# Verify namespace
xcsh ai query "What is my current namespace?"

# Try with explicit namespace
xcsh ai query "List resources" -ns system

# Check for errors in JSON output
xcsh ai query "List resources" -o json

No Follow-Up Suggestions

Cause: Some queries don’t generate follow-ups if they’re self-contained.

This is normal - not all responses will have follow-up suggestions.

---

Output Format Examples

JSON Output

xcsh ai query "What namespace am I in?" -o json

\{
  "query_id": "qid-abc123",
  "current_query": "What namespace am I in?",
  "generic_response": \{
    "text": "You are currently in the 'default' namespace."
  \},
  "follow_up_queries": [
    "List all namespaces",
    "Switch to a different namespace"
  ]
\}

YAML Output

xcsh ai query "What namespace am I in?" -o yaml

query_id: qid-abc123
current_query: What namespace am I in?
generic_response:
  text: You are currently in the 'default' namespace.
follow_up_queries:
  - List all namespaces
  - Switch to a different namespace

TSV Output (for parsing)

xcsh ai query "List load balancers" -o tsv

NAME NAMESPACE STATUS
lb-prod production ACTIVE
lb-staging staging ACTIVE

---

Command Aliases Summary

Full Command	Aliases
xcsh ai_services	ai, genai, assistant
ai query	q, ask
ai chat	i, interactive
ai feedback	fb, rate
ai eval query	ai eval q
ai eval feedback	ai eval fb

---

See Also

• AI Services Command Reference
• Query Command Details
• Chat Command Details
• Feedback Command Details