Skip to main content

Deployment and System Requirements

Client​

Minimum configuration:

4-core 1.8 GHz or faster CPU, 8 GB RAM (minimum 7.5 GB when running in a VM), and 100 GB free HDD/SSD space

Recommended configuration:

8-core 2.4 GHz or faster CPU, 16 GB RAM, and 500 GB or more free HDD/SSD space

Supported systems:

  • Windows 10 Home (x86)
  • Windows 10 Pro (x86)
  • Windows 10 Education (x86)
  • Windows 11 Home (x86)
  • Windows 11 Pro (x86)
  • Windows Server 2019 (x64)
  • Kylin-Desktop-V10-SP1 (x86 and arm64)
  • uniontechos-desktop-20-professional (x86 and arm64)
  • Ubuntu 18.04 and later (x86 and arm64)
  • macOS 15 Sequoia and later (Apple silicon and Intel)

Private Deployment​

Server Requirements​

Minimum configuration (single-machine deployment):

8-core 2.4 GHz or faster CPU, 16 GB RAM, and 200 GB free SSD space

Recommended configuration:

  • Gateway servers: 2 machines, each 1 core / 2 GB RAM / 10 GB disk
  • Application servers: 2 machines, each 16 cores / 32 GB RAM / 300 GB disk
  • MySQL primary/replica: 2 machines, each 4 cores / 8 GB RAM / 300 GB disk
  • Redis primary/replica: 2 machines, each 4 cores / 8 GB RAM / 100 GB disk
  • MongoDB cluster: 2 machines, each 4 cores / 8 GB RAM / 500 GB disk

Software requirements:

CategoryRequirement
Operating systemCentOS 7.4 or later, and Linux distributions with kernel 4.0 or above, including Ubuntu, Debian, OpenSUSE, and major domestic systems such as Kylin, UOS, OpenEuler, and Anolis
Other dependenciesDocker-ce 20.10.21 or later, Nginx 1.22.1 or later
Permission requirementRoot privileges are required throughout deployment and operations; deployment is based on Docker Swarm
Network portsApplication service: 80, 443, 9090; MySQL: 3306; Redis: 6379; MongoDB: 27017

Model Service Requirements​

The model service must support an OpenAI-compatible interface. Supported models include:

  • Qwen3.5-397b-a17b (recommended)
  • GLM-5
  • DeepSeek-V3.2

Other models require additional testing. If machine resources are limited, we recommend the open-source Qwen3.5 series, although the overall effect may be reduced.

The model service must also support:

  • streaming responses
  • function calling
  • at least a 64K context window, with 128K or above recommended

Other Requirements​

ServiceRequirement
Unified authentication integrationOptional. Integrates with the customer's identity provider instead of using bit-Agent's built-in authentication service
Speech service integrationOptional. Integrates with an existing enterprise speech service. Without it, voice features are unavailable
Image analysis model integrationOptional. Integrates with an enterprise image-analysis LLM service. Either this or OCR is required for image handling
OCR service integrationOptional. Integrates with an enterprise OCR service. Either this or an image-analysis model is required for image handling