The in-product AI Assistant at AI → AI Assistant — an F5-expert chat grounded in a live snapshot of your own estate, plus an AI report generator and an iRule builder. What it reads, what it can and can't do, how it differs from AutoPilot and Support Copilot, and the AI-provider prerequisite.
AI Assistant is a full-page, conversational expert for the people who run your application delivery estate. Ask it how to stage an F5 upgrade, why a WAF policy is throwing false positives, or which of your certificates expire this month — it answers as an F5 / ADC engineer, and every chat is grounded in a live snapshot of your own Versus instance.
It lives in the common sidebar under AI → AI Assistant (/ai) and requires administrator access.
The left rail switches between three tools that share one expert persona:
| Mode | What it does |
|---|---|
| Chat | A multi-turn conversation with the "F5 Expert AI" — BIG-IP LTM/GTM configuration, iRules, WAF/ASM, HA and failover, upgrades, SSL/TLS, SNAT/NAT, iControl REST, performance tuning, plus NetScaler and Alteon. Quick prompts (Upgrade checklist, WAF tuning, Performance tuning, iRule example) seed the box; every answer has a copy button, and New Chat clears the thread. |
| Reports | Pick Capacity Planning, Security Posture, Upgrade Readiness, or Health Report and click Generate Report for a structured document — utilization and growth projections, threat landscape and hardening steps, pre-checks and rollback plan, or an executive summary with recommended actions. |
| iRule Builder | Describe what an iRule should do ("redirect all /api/* traffic to pool B, add CORS headers, log client IP…") and get the complete iRule, an explanation of what it does, performance considerations, and testing guidance. |
Answers render as markdown — tables, code blocks, and status pills — so tmsh commands and iControl examples come out ready to paste.
Every Chat request carries a live platform snapshot assembled from your database at that moment — real, tenant-scoped data from this instance, not a training-time memory:
| Section | What the model sees |
|---|---|
| Devices | Total, count by status, how many are offline, and up to 20 offline device names |
| Virtual servers and pools | Totals, and the virtual-server count by status |
| Certificates | Total, already expired, expiring within 30 days, and the ten soonest with days remaining |
| Collectors | Total and count by status |
| Workflow runs | Of the last 50 runs, how many failed and how many are still running |
The assistant is instructed to use that data, cite exact names and counts, and never claim it lacks platform data. Each section is best-effort — if a module isn't present on your instance, that section is simply absent from the snapshot.
The Reports mode is grounded in the same live snapshot: pick a report type (Capacity, Security, Upgrade or Health) and the report is written from your real devices, virtual servers, certificates, collectors and workflow runs, naming what it found and saying plainly where the platform has no data for a section. The iRule Builder uses the F5 expert persona alone — it needs your requirement, not your estate.
Two lighter surfaces share the same backend:
| AI Assistant | Diagnostics AutoPilot | Support Copilot | |
|---|---|---|---|
| Audience | Engineers inside the product | Engineers inside the product | Customers on the Versus support site |
| Trigger | You ask a question | New incidents in Versus Desk / ServiceNow | You ask a question |
| Grounding | Live estate snapshot plus F5 expertise | Iterative read-only probes against live devices and integrations, through the collector | Docs, knowledge base, platform status, and Sentinel known issues — with citations |
| Can act? | No — read-only, no device access | Yes — generates remediation workflows, held for approval or run autonomously per policy | Limited — open or track a Desk ticket, escalate to a human |
| Output | A conversational answer, report, or iRule | An evidence-backed root cause plus a workflow | A cited answer or a ticket |
See Diagnostics AutoPilot and Versus Support Copilot.
Every mode on this page uses the platform's default AI provider. Providers are registered under Versus Platform → Integrations → AI Providers (/integrations/ai-providers): add an Anthropic Claude or OpenAI-compatible endpoint (Azure OpenAI, Ollama, vLLM, LM Studio), optionally pin a model, Test it, and Set default — that single default drives Chat, Reports, iRule Builder, the floating panel and screen help, as well as script generation, Quick Python workflows and vulnerability scanning. An Anthropic provider may leave its key blank to inherit the backend's ANTHROPIC_API_KEY environment variable, which is also the fallback when no default provider exists (and when the default fails); with neither configured, requests from every mode fail.
| Item | Detail |
|---|---|
| Location | AI → AI Assistant (/ai), administrator access |
| Modes | Chat, Reports (Capacity / Security / Upgrade / Health), iRule Builder |
| Grounding | Live snapshot (Chat, Reports, panel, screen help): devices, virtual servers, pools, certificates, collectors, recent workflow runs |
| Actions | None — read-only, never touches devices or workflows |
| Provider | The default under Integrations → AI Providers (Anthropic or OpenAI-compatible); ANTHROPIC_API_KEY as fallback |
| Related surfaces | Floating Versus AI panel (bottom-right) and screen help |