This article provides system requirements for installing and running BRIX AI.
BRIX AI server
- Operating system.
Parameter |
Value |
OS |
Ubuntu Server 22.04 LTS* or 24.04 LTS |
Architecture |
x86_64 (amd64) |
*LTS (Long Term Support) refers to Ubuntu releases with extended security updates for five years.
- Minimum hardware requirements.
Resource |
Minimum |
Recommended |
CPU |
4 cores, 2 GHz |
8 cores |
RAM |
16 GB |
32 GB |
System disk |
50 GB SSD |
60 GB SSD |
Data disk |
50 GB |
Depends on data volume |
- Network requirements.
Bandwidth |
Minimum 100 Mbps |
Inbound ports |
443/tcp (HTTPS), configured during installation |
Outbound connections |
Access to Large Language Model (LLM) Provider API (OpenAI, YandexGPT, GigaChat, etc., or a locally deployed provider) |
- Additional software.
Docker Engine |
Recommended: 29.0+ Supported: 24.0+ |
Docker Compose |
2.20+ (included with Docker Engine) |
Docker Swarm |
Built into Docker Engine. Required for installing BRIX AI in a cluster. |
BRIX integration requirements
To integrate BRIX AI with the BRIX platform, make sure the following requirements are met:
Domain name |
A dedicated DNS name for the BRIX AI server, for example, cortex.company.ru. |
TLS certificate |
A valid SSL or TLS certificate for the specified domain. |
Network accessibility |
The BRIX AI server must be accessible from the BRIX server via HTTPS. |
Supported certificate types
To secure the connection between the user and the server, you can use:
- Certificates from public Certificate Authorities (CAs). Browsers trust these by default. We recommend using this option.
- Certificates from your corporate Certificate Authority. You need to add this CA to the list of trusted CAs on the BRIX side.
- Certificates issued via Let's Encrypt. This option requires a public DNS and internet access to the BRIX AI server.
Disk space requirements
Disk space is used for two main purposes: installing server software and storing user data.
Server
Component |
Size |
Docker images for BRIX AI |
~3 GB |
Docker images for dependencies (MongoDB, Redis, RabbitMQ, MinIO, Qdrant) |
~5 GB |
System data and logs |
~2 GB |
Total |
~10 GB |
BRIX AI data
The required storage space depends on your data volume:
Data type |
Estimated calculation |
Chats and messages (MongoDB) |
~16 KB per message |
Vector indexes (Qdrant) |
~8–16 KB per document fragment |
Uploaded files (MinIO) |
Depends on file size |
Scaling
Depending on data volume and the number of users working with BRIX AI components, you may need to increase system resources.
начало внимание
Resource scaling under increased load is available only in specific BRIX AI editions.
конец внимание
Parameter |
Low load |
Medium load |
High load |
Concurrent users |
1–10 |
10–30 |
30+ |
CPU |
4–8 cores |
8–16 cores |
16+ cores |
RAM |
16–32 GB |
32–64 GB |
64+ GB |
Data disk |
50–100 GB |
100–300 GB |
300+ GB |
Network |
100 Mbps |
1 Gbps |
1+ Gbps |
Important: Exact requirements depend on your use cases, the number of configured AI agents and AI operations, the volume of indexed data, and the frequency of queries sent to the language model.
To increase throughput, you can configure performance parameters in the environment variables file.
Using external services
BRIX AI includes built-in services (MongoDB, Redis, RabbitMQ, MinIO, Qdrant). You can replace them with external managed services already configured in your infrastructure. In this case, exclude these components when calculating resources for installing BRIX AI.
To learn how to enable external services, see Configure BRIX AI environment variables.
Outbound network connections
The BRIX AI server requires access to external services.
LLM Provider API |
Large Language Model Provider API (OpenAI, Anthropic, YandexGPT, GigaChat, etc.) |
Embedding API |
Vector representation API (may share the same provider as the LLM) |
MCP servers |
External MCP tools, if used |
Data sources |
Web pages and APIs for knowledge base indexing |
Important: On-premises installations in an isolated or air-gapped network require local LLM providers with an OpenAI-compatible API.