Private · System Requirements

Minimum and recommended hardware, software, and container configuration for production.

Hardware

Deployment	CPU	Memory	Disk	Notes
Minimum (POC)	4 cores	16 GB	200 GB SSD	Single instance
Recommended (production)	8 cores × 2 instances	32 GB × 2	500 GB SSD × 2	HA
Large scale (> 1000 users)	16 cores × 3+	64 GB × 3+	NVMe	Horizontal scale

Databases

Database	Use	Minimum	Recommended
MySQL 8.0+	Business data (users/org/workspace/Agent/Workflow metadata/billing/audit)	4 cores / 16 GB	16 cores / 64 GB / SSD
PostgreSQL 14+ + pgvector	Knowledge base RAG vectors (HNSW / IVFFlat)	4 cores / 16 GB	8 cores / 32 GB (memory-sensitive)
Redis 7+	Cache / session / queue	1 core / 4 GB	2 cores / 8 GB

File Storage

Option	Description
Local	Suitable for small scale
OSS / S3	Recommended; can mount Aliyun OSS / AWS S3 / MinIO

Containers

Item	Version
Docker	20.10+ (24+ recommended)
Docker Compose	v2
Kubernetes (optional)	1.24+
Helm (optional)	3.x

Operating System

Supported:

Ubuntu 20.04+ / 22.04 LTS (recommended)
CentOS 7+ / Rocky Linux 8+
Debian 11+
Domestic: OpenEuler / Kylin / UOS (via OpenAI-compatible path)

Performance Targets

Metric	Target
API response	< 500ms P95
Concurrent users	1000+
Availability	99.9%

→ Numbers under Private depend on hardware and operations maturity.

Network

Port	Use
443	HTTPS (external)
80	HTTP → HTTPS redirect
Internal ports	Managed by docker compose / k8s

External access:

Target	Use
Model APIs	Only when using SaaS LLMs (OpenAI / Anthropic, etc.)
System updates	Container image registries (can mirror to intranet)
Company SSO	LDAP / AD / OAuth IDP

Fully air-gapped deployment

All dependent images can be delivered with the package. Fully offline operation is possible (requires self-hosted models).

Standard 5-Phase Implementation (2–4 weeks)

Phase 1 · Requirements (2-3 days)
  Capacity planning · Security · Integration scope

Phase 2 · Environment prep (3-5 days)
  Servers · Network · Databases · Storage

Phase 3 · Deploy & configure (5-10 days)
  Docker / K8s · SSO · Model integration

Phase 4 · Testing & validation (3-5 days)
  Functional · Load · Security tests

Phase 5 · Go-live (1-2 days)
  Data migration · User training · Cutover

Next Steps

Docker deployment → Docker
K8s deployment → Kubernetes
High availability → HA
Upgrade procedure → Upgrade

Private · System Requirements

On this page