Auto-Scaling

# Configure auto-scaling for an agent
$ hyperfold agent scale sales-negotiator \
  --min=2 \
  --max=50 \
  --target-concurrency=100
 
Scaling configuration updated:
  Agent:              sales-negotiator
  Min Instances:      2
  Max Instances:      50
  Target Concurrency: 100 requests/instance
 
# View current scaling status
$ hyperfold agent scale sales-negotiator --status
 
SCALING STATUS: sales-negotiator
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
CONFIGURATION
  Min Instances:      2
  Max Instances:      50
  Target Concurrency: 100
 
CURRENT STATE
  Active Instances:   8
  Current Load:       720 req/min
  Avg Concurrency:    90 req/instance
  CPU Utilization:    65%
  Memory Usage:       412MB/512MB
 
SCALING ACTIVITY (last hour)
  10:15    Scale up    6 → 8 instances    (traffic spike)
  09:45    Scale up    4 → 6 instances    (approaching target)
  09:00    Stable      4 instances
  08:30    Scale down  6 → 4 instances    (traffic decrease)

# Advanced scaling configuration
$ hyperfold agent scale sales-negotiator \
  --min=2 \
  --max=100 \
  --target-concurrency=80 \
  --scale-up-threshold=0.7 \
  --scale-down-threshold=0.3 \
  --cooldown=60
 
# Scale based on multiple metrics
$ hyperfold agent scale sales-negotiator \
  --metrics='[
    {"metric": "concurrency", "target": 80},
    {"metric": "cpu", "target": 70},
    {"metric": "latency_p95", "max": 500}
  ]'
 
# Scheduled scaling for predictable traffic
$ hyperfold agent scale sales-negotiator \
  --schedule='[
    {"cron": "0 9 * * 1-5", "min": 10, "max": 100},
    {"cron": "0 18 * * 1-5", "min": 5, "max": 50},
    {"cron": "0 0 * * 6-7", "min": 2, "max": 20}
  ]'
 
Scheduled scaling configured:
  Weekdays 9 AM:    10-100 instances (business hours)
  Weekdays 6 PM:    5-50 instances (evening)
  Weekends:         2-20 instances (low traffic)
 
# Configure scaling per agent type
$ hyperfold agent scale fulfillment-agent \
  --min=1 \
  --max=20 \
  --target-concurrency=50 \
  --scale-up-rate=2 \
  --scale-down-rate=1

# Scaling strategies comparison
 
TARGET CONCURRENCY (recommended)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Scales based on requests per instance.
Best for: API-heavy workloads, negotiation agents
 
$ hyperfold agent scale myagent \
  --strategy=target_concurrency \
  --target-concurrency=100
 
How it works:
  desired = ceil(current_requests / target_concurrency)
  If 800 requests and target=100: need 8 instances
 
 
CPU UTILIZATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Scales based on CPU usage percentage.
Best for: Compute-intensive workloads
 
$ hyperfold agent scale myagent \
  --strategy=cpu \
  --target-cpu=70
 
How it works:
  Scale up when avg CPU > 70%
  Scale down when avg CPU < 40%
 
 
QUEUE DEPTH
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Scales based on pending requests.
Best for: Batch processing, async workflows
 
$ hyperfold agent scale myagent \
  --strategy=queue_depth \
  --target-queue=10
 
How it works:
  desired = ceil(queue_size / target_queue)
  Scale up quickly when queue grows
 
 
HYBRID
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Combines multiple metrics with weights.
Best for: Complex workloads
 
$ hyperfold agent scale myagent \
  --strategy=hybrid \
  --metrics='[
    {"metric": "concurrency", "target": 80, "weight": 0.5},
    {"metric": "cpu", "target": 70, "weight": 0.3},
    {"metric": "memory", "target": 80, "weight": 0.2}
  ]'

# Monitor scaling activity
$ hyperfold agent scale-history sales-negotiator --since=24h
 
SCALING HISTORY: sales-negotiator (24 hours)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
TIME          ACTION      FROM    TO      REASON
Jan 20 10:15  Scale Up    6       8       Concurrency at 95%
Jan 20 09:45  Scale Up    4       6       Concurrency at 92%
Jan 20 08:30  Scale Down  6       4       Concurrency at 25%
Jan 20 02:00  Scale Down  8       6       Scheduled (night)
Jan 19 22:00  Scale Down  10      8       Traffic decrease
Jan 19 18:30  Scale Up    8       10      Traffic spike
Jan 19 14:00  Scale Up    6       8       Concurrency at 88%
 
SUMMARY
  Scale Up Events:    4
  Scale Down Events:  3
  Avg Instances:      6.2
  Peak Instances:     10
  Min Instances:      4
 
# View real-time scaling metrics
$ hyperfold metrics agents sales-negotiator --live
 
LIVE METRICS: sales-negotiator
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
[▓▓▓▓▓▓▓▓░░] 8/10 instances
 
Requests:     842/min    [▓▓▓▓▓▓▓▓▓░]
Concurrency:  95/100     [▓▓▓▓▓▓▓▓▓▓]
CPU:          68%        [▓▓▓▓▓▓▓░░░]
Memory:       420MB      [▓▓▓▓▓▓▓▓░░]
Latency p95:  340ms      [▓▓▓▓░░░░░░]
 
Scaling: ⚡ Scale up likely in ~2 minutes
 
(Press q to quit, r to refresh)

# Optimize scaling configuration
$ hyperfold agent scale-analyze sales-negotiator --since=7d
 
SCALING ANALYSIS: sales-negotiator (7 days)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
CURRENT CONFIGURATION
  Min: 2, Max: 50, Target Concurrency: 100
 
OBSERVATIONS
  ⚠ Over-provisioned during off-hours
    Avg utilization 2-6 AM: 15%
    Recommendation: Reduce min to 1 during night
 
  ⚠ Scale-up lag during morning spike
    Avg scale-up time: 3.2 minutes
    Peak missed capacity: 12%
    Recommendation: Increase min to 4 at 8 AM
 
  ✓ Target concurrency well-tuned
    Avg utilization: 72%
    Recommendation: Keep at 100
 
  ⚠ Frequent scale oscillation 2-4 PM
    Scale events: 8 in 2 hours
    Recommendation: Increase cooldown to 90s
 
RECOMMENDED CONFIGURATION
  --min=1
  --max=50
  --target-concurrency=100
  --cooldown=90
  --schedule='[
    {"cron": "0 8 * * 1-5", "min": 4},
    {"cron": "0 20 * * 1-5", "min": 2},
    {"cron": "0 2 * * *", "min": 1}
  ]'
 
ESTIMATED SAVINGS
  Current monthly cost:    $2,450
  Optimized monthly cost:  $1,890
  Savings:                 $560 (23%)
 
Apply recommendations? [Y/n]