# Support plan

> **Audience.** Buyers and admins deciding whether to subscribe to
> Support. 4-minute read.
> ⟵ [back to docs index](../README.md)

**TL;DR.** Subscribe and you get a private Slack channel with a real
Bilbs engineer in it, a 24/7 Sev-1 pager, incident triage, and someone
to file your OEM hardware RMAs for you. Not required — skip it and you
still own everything; you just handle issues yourself.

## 01 · What's in the plan

| Coverage | Detail |
|----------|--------|
| **Private Slack channel** | Shared between your admin team and the Bilbs engineer assigned to your account. Named `#bilbs-<yourco>`. |
| **Response SLA** | < 4 business hours for Sev-3 / routine questions. |
| **Incident SLA** | < 1 hour for Sev-1 (production down), 24/7 pager. |
| **Monthly health review** | 30-minute call, admin + Bilbs engineer, walking through metrics, open issues, upcoming upgrades. |
| **OEM RMA shepherding** | We file hardware RMAs with the OEM on your behalf and track them to resolution. |
| **Post-incident reports** | Written RCA after any Sev-1 or Sev-2 incident within 5 business days. |
| **Consulting hours** | 4 hours/month of ad-hoc architecture / integration help included in the base plan (Pro + Cluster). |

## 02 · Severity definitions

| Sev | Definition | Response target |
|-----|-----------|-----------------|
| **Sev-1** | Box is down or unresponsive; production workload impacted. | < 1 h any time of day |
| **Sev-2** | Degraded but functional — e.g., elevated error rate, partial outage. | < 4 h business hours |
| **Sev-3** | Working, but a question or minor issue. | < 1 business day |
| **Sev-4** | Feature request, cosmetic, non-urgent. | Next monthly review |

You set the severity when you open a ticket. We can downgrade after
triage if we disagree; we always honour your severity until we've had
a look.

## 03 · How to reach us

### The Slack channel (primary)

- Post freely — we react within the SLA for the severity you mark.
- Paste `@bilbs-oncall` for Sev-1 (wakes the on-call engineer's pager).
- Sharing logs / screenshots is fine — the channel is scoped to your
  account and ours; no other clients see it.

### The pager (Sev-1 only)

A phone number printed on your wallet card, also accessible via
`bilbs support pager` from the CLI. Use this only when your Slack
integration is down or the situation genuinely requires voice.

### Email fallback

`support@groupebilbs.com` — accepted but slower than Slack. Use only when
Slack isn't available to you for some reason.

### Not for Support

- **Billing questions** → `admin@groupebilbs.com`.
- **End-user questions** (your employees) → must go through you first.
  We do not support end users directly by design.
- **Commercial negotiations** → your account owner, not the Support
  engineer.

## 04 · Response-time examples

- Monday 10:14 AM: "Eval pass-rate dropped overnight, not sure what
  happened." → Sev-3 → reply by 2:14 PM Monday.
- Saturday 3:30 AM: "Production UI returns 500 on every request." →
  Sev-1 → engineer acks by 4:30 AM; starts working.
- Tuesday 11:47 PM: "We want to add a new tool to the agent — how?" →
  Sev-3 → reply next business day.
- Thursday 9:03 AM: "GPU index 1 offline, degraded to one card." →
  Sev-2 → reply + RMA filed by 1:03 PM.

## 05 · What's not included

- **Feature development.** New admin-UI features, new integrations —
  not Support. Use Embedded or a new Build engagement.
- **End-user support** for your employees — by design. You support your
  employees; we support you.
- **Training / enablement for your team.** We include one 90-min
  onboarding call for new admins (free); beyond that, it's a separate
  SOW.
- **Third-party software problems.** If your IdP is broken or your
  firewall is dropping traffic, we'll help you diagnose that it's a
  third-party issue, then you fix it on your side.

## 06 · Pricing

| Tier | Monthly |
|------|--------:|
| Mini | $249 |
| Box | $399 |
| Pro | $799 |
| Cluster | custom SLA, starts $2,500 |

Volume discount on multi-Box accounts (contact us).

## 07 · Hardware RMA shepherding

What this actually involves when a GPU dies:

1. Your admin sees the alert in the Dashboard. They post the alert in
   Slack and tag `@bilbs-oncall`.
2. Engineer opens in ≤ 1 hour. Confirms hardware failure from the
   logs and IPMI.
3. Engineer pulls the `diag collect` tarball from your Box.
4. Engineer files the OEM RMA (Dell / Supermicro / NVIDIA — whoever
   made the chassis) and provides the case number.
5. OEM tech dispatch is scheduled; engineer shares tracking with you.
6. When the replacement arrives, engineer walks you through the swap
   (or attends on-site for Pro + Cluster clients in major metros).
7. After restore, engineer runs the eval suite to confirm recovery
   and writes the post-incident report.

Without the Support plan, you do steps 3 through 7 yourself. The
troubleshooting guide covers the mechanics.

## 08 · On-site visits (Pro + Cluster)

- **Pro:** one on-site install day included at Build time; additional
  on-site visits quoted.
- **Cluster:** up to two on-site visits per year included in the base
  Support fee (major NA / UK / EU metros). Additional visits quoted.

On-site engineer arrives with spare parts kit, runs pre-arranged
diagnostic work (e.g., for an upgrade window), and produces a written
report.

## 09 · Scope of the engineer relationship

- The same engineer stays on your account for the life of the
  subscription (unless they leave the company).
- Scheduled quarterly 1:1 with the engineer's backup so there's always
  a second person who knows you.
- Engineer reviews your metrics weekly, even outside active tickets;
  proactive flags land in Slack.
- If the engineer isn't a fit, ask for a swap — no question.

## 10 · Consulting hours

Included on Pro and Cluster Support plans (Mini / Box can add this for
+$150/h, prepaid 10 hours minimum).

Typical uses:
- Architecture review of a new agent / workflow.
- Cost-tuning session.
- Debug help for a custom tool you're integrating.
- Eval-suite design for a new domain you're expanding into.

Track hours in the Support Slack channel; monthly cap resets on the
calendar month.

## 11 · Opting in & out

**Subscribe:** `/admin/system/plans/support → Subscribe` — or via
`bilbs plan subscribe support`.

**Cancel:** same place. Effective end of current billing month.

Your Slack channel is archived 30 days after cancellation. You can
export the transcript before then. If you re-subscribe within 6 months,
the channel is restored with history intact.

---

See also: [updates-plan.md](./updates-plan.md) ·
[troubleshooting.md](../operations/troubleshooting.md) ·
[bilbs-box.md](../bilbs-box.md).