Skip to main content

Hardware Requirements

Chatty AI hardware requirements vary based on deployment size and usage patterns. Choose the configuration that matches your needs.

Quick Reference​

ConfigurationCPURAMStorageUsersUse Case
Minimum4 cores16 GB100 GB SSD1-5Development, testing, POC
Recommended8 cores32 GB500 GB NVMe10-50Production, standard workload
High Load16+ cores64+ GB2+ TB NVMe50-100+Enterprise, heavy workload

Minimum Configuration​

Suitable for: Development, testing, proof of concept, small teams (1-5 users)

⚠️ Not recommended for production deployments

CPU​

  • 4 cores minimum
  • x86_64 or ARM64
  • 2.0 GHz or higher

Memory​

  • 16 GB RAM minimum
  • Memory allocation:
    • PostgreSQL: 2 GB
    • Qdrant: 2-4 GB
    • Chatty AI: 4-8 GB
    • Databases services: 2-4 GB
    • n8n: 1-2 GB
    • System: 2-4 GB

Storage​

  • 100 GB SSD minimum
  • Breakdown:
    • OS and Docker: 20 GB
    • Images: 15-20 GB
    • PostgreSQL: 10-20 GB
    • Qdrant: 10-30 GB
    • Application: 10-20 GB
    • Logs: 5-10 GB
    • Backups: 20-30 GB

Performance​

  • Response times: 2-5 seconds
  • Concurrent users: 1-3
  • Documents: < 1,000

Suitable for: Production deployments, medium teams (10-50 users), standard workloads

✅ Recommended for most production deployments

CPU​

  • 8 cores recommended
  • Modern x86_64 (Intel Xeon, AMD EPYC) or ARM64
  • 2.5 GHz or higher
  • Hyper-threading enabled

Memory​

  • 32 GB RAM recommended
  • Memory allocation:
    • PostgreSQL: 4 GB
    • Qdrant: 4-8 GB
    • Chatty AI: 8-12 GB
    • Databases services: 4-6 GB
    • n8n: 2-4 GB
    • Nginx: 1 GB
    • System: 4-6 GB

Storage​

  • 500 GB NVMe SSD recommended
  • Breakdown:
    • OS and Docker: 30 GB
    • Images: 20-30 GB
    • PostgreSQL: 50-100 GB
    • Qdrant: 100-200 GB
    • Application: 50-100 GB
    • Logs: 20-30 GB
    • Backups: 100-150 GB

Storage Type:

  • NVMe SSD: Strongly recommended for Qdrant performance
  • RAID 10: Optional for redundancy
  • Separate Volumes: Consider for data and backups

Network​

  • Bandwidth: 1 Gbps
  • Latency: < 50ms to LLM endpoints
  • Redundancy: Optional backup connections

Performance​

  • Response times: 1-2 seconds
  • Concurrent users: 10-30
  • Documents: 5,000-10,000
  • Query throughput: Medium-high

High Load Configuration (Enterprise)​

Suitable for: Large enterprises (50-100+ users), heavy workloads, extensive document collections

CPU​

  • 16+ cores for high load
  • Enterprise processors (Intel Xeon Scalable, AMD EPYC)
  • 3.0 GHz or higher
  • Multiple sockets supported

Memory​

  • 64+ GB RAM for high load
  • Memory allocation:
    • PostgreSQL: 8-16 GB
    • Qdrant: 16-32 GB
    • Chatty AI: 16-24 GB
    • Databases services: 8-12 GB
    • n8n: 4-8 GB
    • Nginx: 2 GB
    • System: 8-12 GB

Consider 128 GB for very large deployments

Storage​

  • 2+ TB NVMe SSD for high load
  • Breakdown:
    • OS and Docker: 50 GB
    • Images: 30-50 GB
    • PostgreSQL: 200-500 GB
    • Qdrant: 500 GB - 1 TB
    • Application: 200-400 GB
    • Logs: 50-100 GB
    • Backups: 500 GB - 1 TB

Storage Architecture:

  • NVMe SSD: Required
  • RAID 10: Recommended
  • Separate Volumes:
    • OS and applications
    • PostgreSQL data
    • Qdrant data
    • Backups (can be slower storage)

Network​

  • Bandwidth: 10 Gbps
  • Latency: < 20ms to LLM endpoints
  • Redundancy: Required backup connections
  • Internal: 10 Gbps for storage access

Performance​

  • Response times: < 1 second
  • Concurrent users: 50-100+
  • Documents: 10,000-100,000+
  • Query throughput: High
  • Uptime: 99.9%+

Storage Performance Requirements​

Critical: Qdrant Performance​

Qdrant performance heavily depends on storage speed:

  • NVMe SSD: Required for production
  • Latency: < 1ms
  • IOPS: > 10,000

Impact of slow storage:

  • ❌ Slow vector search
  • ❌ Degraded RAG performance
  • ❌ Timeout errors

Important: PostgreSQL Performance​

  • SSD: Minimum (NVMe preferred)
  • IOPS: Adequate for transaction load

General Storage​

Other services can use standard SSD.

Resource Limits (Docker)​

Configured in docker-compose.yaml:

Qdrant​

  • Memory Limit: 4 GB
  • Memory Reservation: 2 GB
  • CPUs: 1 core

Chatty AI​

  • Memory Limit: 8 GB
  • CPUs: 1 core
  • Node.js Heap: 8192 MB (configurable via NODE_MAX_MEMORY)

Scaling Indicators​

When to Upgrade​

Upgrade from Minimum → Recommended:

  • Moving to production
  • More than 5 users
  • Document collection growing

Upgrade from Recommended → High Load:

  • CPU usage consistently > 70%
  • Memory usage > 80%
  • Disk I/O bottlenecks
  • More than 50 concurrent users
  • Document collection > 10,000
  • Response times degrading

High Availability (HA) Considerations​

For production and enterprise deployments:

Infrastructure​

  • Redundant power supplies
  • RAID 10 storage
  • Backup network connections
  • Load balancers (for Nginx)

Monitoring​

  • Real-time performance monitoring
  • Automated alerting
  • Capacity planning
  • Predictive analytics

Backup​

  • Automated daily backups
  • Off-site backup storage
  • Regular restore testing
  • Point-in-time recovery

Platform Support​

Operating Systems​

  • Linux: Ubuntu 22.04 LTS (recommended), Debian, RHEL, Rocky Linux
  • Architecture: x86_64 (amd64) or ARM64

Docker Requirements​

  • Docker Engine 20.10+
  • Docker Compose 2.0+

See Software Requirements for details.

Cost Considerations​

Minimum Configuration​

  • Server: $2,000-$5,000
  • Suitable for: Testing, POC
  • Server: $5,000-$15,000
  • Storage: $1,000-$3,000
  • Network: $500-$2,000
  • Total: ~$6,500-$20,000

High Load Configuration​

  • Server: $10,000-$50,000+
  • Storage: $5,000-$20,000+
  • Network: $2,000-$10,000+
  • Redundancy/HA: 2x base costs
  • Total: ~$34,000-$160,000+

Plus ongoing costs: power, cooling, maintenance.

Monitoring Resources​

Monitor these metrics to ensure adequate resources:

CPU​

  • Average utilization < 60%
  • Peak utilization < 85%
  • Per-container CPU usage

Memory​

  • Average utilization < 70%
  • Swap usage minimal
  • Per-container memory usage

Storage​

  • Disk utilization < 80%
  • I/O wait times
  • Read/write throughput
  • IOPS

Network​

  • Bandwidth utilization
  • Latency to external services
  • Packet loss
  • Connection counts

Recommendations by Use Case​

Use CaseConfigurationNotes
DevelopmentMinimumLocal testing only
POC/DemoMinimumShort-term evaluation
Small Team (< 10)RecommendedStandard production
Medium Team (10-50)RecommendedStandard production
Large Team (50-100)High LoadEnterprise deployment
Enterprise (100+)High LoadConsider clustering
24/7 CriticalHigh Load + HAMission-critical

Next Steps​

  1. Choose configuration based on your use case
  2. Review Software Requirements
  3. Prepare infrastructure using Prerequisites Checklist