Quick Setup: the one-screen flow

QUICK SETUP COVERS ~70% of autoscaler use cases:
  Single metric (CPU typically)
  Target tracking model
  Smart defaults accepted as-is or with minor tune

ADVANCED MODE for the other 30%:
  Custom metrics (queue depth, request count)
  Step scaling
  Schedule-triggered scaling
  Multi-tier or resource group targeting

NEW POLICY (Quick Setup)
─────────────────────────────────────────────────────────
Cloud account:    [prod-aws-us-east-1 ▾]
Target:           [asg-prod-payments-api ▾]
                  (filtered list: only autoscalable targets)

[Use smart defaults]   [Customize ▾]

When you click "Use smart defaults":
  ZopNight queries metrics; computes recommendations
  Shows the suggested policy

  Suggested Policy:
    Min capacity:    3
    Max capacity:   15
    Target:         75% CPU
    Cooldown:      300s
    Mode:           autopilot (based on credential permission level)

[Edit settings]   [Apply policy]

1. Customer picks cloud account + target (10 seconds)
2. ZopNight reads the aggregator's CPU stats for the target
3. Welford computes the standard deviation (sub-second)
4. Smart defaults presented (instant)
5. Customer accepts or edits, and saves
6. Customer clicks Apply: ZopNight queues an apply intent
7. The executor writes the cloud-side policy and reports back
8. Policy active

END-TO-END: ~30-60 seconds from "I want autoscaling"
                to "policy is active and managing capacity"

SETTING                          SET TO
─────────────────────────────────────────────────────────
Min capacity                     from cpu.Avg (from L2)
Max capacity                     max(count+2, ceil(count×P99/target×1.3)) (from L2)
Target value                     3-way tier on P95: 70 / 75 / 80 (from L2)
Cooldown                         180 / 300 / 600s by stddev (300s, no metrics)
Metric                           CPUUtilization (default for compute)
Scaling model                     Target tracking
Mode                              Auto-derived from credential permissions
                                  (autopilot, or recommend if read-only)

CREDENTIAL TYPE                       AUTO-MODE
─────────────────────────────────────────────────────────
Read-only (Reader role)                recommend (shown as "Manual")
Read-write                             autopilot (if you did not pick one)

QUICK SETUP FITS:
  Production web tier with target tracking on CPU
  Worker pool with target tracking on queue depth (if queue depth
    is a standard metric)
  ECS service with target tracking
  ASG with predictable load pattern
  Single workload type per policy

ADVANCED MODE NEEDED FOR:
  Multi-metric policies (CPU + latency + queue depth combined)
  Step scaling (specific scale-out steps for spikes)
  Schedule-triggered scaling (time-of-day patterns)
  Custom metrics (application-specific)
  Resource group targeting (scale a group, not single target)
  Multi-tier app coordination

ADVANCED MODE
─────────────────────────────────────────────────────────
TARGET SETTINGS (same as Quick Setup, plus):
  Resource group target           ✓ if scaling a group of resources
  Custom metric                    ✓ if not CPU
  Memory threshold guard            ✓ if memory-aware

SCALING RULES:
  [Target tracking ▾] OR [Step scaling ▾]
  Multiple metrics?               ✓ define each
  Specific scale-out steps         (for step scaling)

SCHEDULE TRIGGERS:
  ✓ if time-based scaling
  Preset:                          [Business Hours / Peak / Weekend ▾]
  Cron:                            custom expressions

MODE:
  Autopilot / Manual (recommend)
  (auto-derived but customizable; read-only accounts locked to Manual)

SCENARIO: New service deployed yesterday

ZopNight's response:
  Any CPU data at all: the normal L2 formulas run on whatever
    data exists (no warning, no special limited-data branch)
  NO CPU metrics yet: fixed fallback defaults
    Min: 1
    Max: max(2 × count, 5)
    Target: 70% (default; not data-derived)
    Cooldown: 300s
    Mode: autopilot (recommend if the account is read-only)

PRACTICAL RAMP-UP:
  Day 1-7: short window, noisy suggestions; sanity-check min/max
  Day 7-14: re-run smart defaults; edit the policy with the new values

WHAT MODE IS:
  autopilot ("Autopilot" in the form): default for read-write accounts
  recommend ("Manual" in the form): forced for read-only accounts
  monitor: accepted by the API and shown as a label; not in the form

WHAT MODE DOES NOT CHANGE:
  Writing the policy to the cloud is always a user action
    (the Apply button, or saving an edit that changes cloud settings)
  Once applied, the cloud's own autoscaler does the scaling
  ZopNight runs no scaling loop of its own in any mode

WHEN YOU CLICK APPLY:

  1. ZopNight pushes an apply intent onto a queue and marks the
     policy "applying" (the call returns at once)
  2. The executor picks it up and calls the cloud provider API:
     AWS: PutScalingPolicy + UpdateASG
     GCP: insertAutoscaler
     Azure: PUT autoscaleSettings
  3. Cloud creates the policy
  4. The executor reports success/failure back to ZopNight
  5. UI shows "Policy active"

PROVISIONER HANDLES:
  Authentication (using stored cloud credentials)
  Error mapping (cloud error → ZopNight error class)
  Retry on transient errors
  Audit logging

T+0      Service deployed yesterday: 6 EC2 instances behind an ASG
T+5 sec  Navigate to Automate → Auto Scaling → + Create Autoscaling Policy
T+10 sec Quick Setup mode (default)
T+15 sec Pick prod-aws-us-east-1 → asg-new-orders-service
T+20 sec Click "Use smart defaults"
T+30 sec ZopNight reads metrics: no CPU datapoints yet for the group

         Fallback defaults shown (no warning banner):
           Min: 1
           Max: 5 (max(2 × count, 5); the endpoint assumes a
                   count of 2 today, so 5 rather than 12)
           Target: 70% CPU
           Cooldown: 300s
           Mode: autopilot (read-write account)

T+40 sec Operator raises Min to 6 (current capacity) and Max to 12,
         saves, then clicks Apply
T+50 sec ZopNight queues the apply; the executor writes the ASG policy
T+60 sec Policy active; the ASG's own autoscaler now scales

FOLLOW-UP over next 2 weeks:
  Week 1: CPU data accumulates
  Week 2: re-run smart defaults; edit the policy with data-driven
          values (saving the edit re-applies it to the cloud)

RESULT after 1 month:
  Smart defaults refined to data-driven values
  Engineer rarely touches the policy

□ STEP 1: Navigate to Automate → Auto Scaling → + Create Autoscaling Policy
  Mode: Quick Setup (default)

□ STEP 2: Pick a target
  Cloud account: __________
  Target: __________

□ STEP 3: Click "Use smart defaults"
  Observe: smart defaults appear
  Note timing: how fast?

□ STEP 4: Review the defaults
  Min: _____
  Max: _____
  Target: ___%
  Cooldown: ___ s
  Mode: __________

□ STEP 5: Cancel (don't commit if sandbox)
  Or apply if you're ready to commit

BEFORE  A ZopNight account with one cloud connected. A scalable group: an AWS Auto Scaling Group, a GCP Managed Instance Group or an Azure VM Scale Set.
ASK     "Set up an autoscaler on this group using the smart defaults, then show me its events."
CHECK   the first few events. A policy with no events is not scaling, it is decoration.