Prerequisites & Hardware Requirements

Infrastructure Requirements

Central Processing Unit (CPU): 4 Physical Cores (x86_64 or ARM64)
Random Access Memory (RAM): 16 GB Minimum (32 GB recommended for full local embedding generation and simultaneous multi-agent execution)
Disk Storage: 50 GB available SSD space (database volumes, Docker container layers, and RustFS S3 storage)
Operating System: Linux (Ubuntu 22.04 LTS+, Debian 12, RHEL 9), macOS (Sonoma+), or Windows 11 with WSL2

The entire platform is orchestrated through containerization, minimizing host-level software requirements:

Docker Engine: Version 26.0 or higher
Docker Compose: Version 2.24 or higher
Node.js (Optional for host-level development): Version 20 LTS or Version 22 LTS
Python (Optional for host-level scripting): Version 3.11 or 3.12

The platform supports local inference, cloud foundation models, or hybrid configurations.

Ollama Endpoint: http://host.docker.internal:11434
Dense Embedding Model: qwen3-embedding:8b (4096-dimensional vectors) or nomic-embed-text:latest
Language Models: deepseek-r1:8b, qwen2.5:14b, or llama3.3:70b
Installation: Ensure Ollama is running on the host system with CORS enabled:
Terminal window
```
OLLAMA_ORIGINS="*" ollama serve
```

DeepSeek API: deepseek-v4-flash or deepseek-chat (Recommended for high-throughput reasoning and cost efficiency). Set DEEPSEEK_API_KEY.
OpenAI API: gpt-4o, text-embedding-3-large. Set OPENAI_API_KEY.
Vercel AI Gateway: Configured via AI_GATEWAY_API_KEY in gxpchat.

To avoid conflicts with existing services, review the host port bindings configured in docker-compose.yml:

Service	Container Port	Host Port (Default)	Environment Override Variable	Purpose
`wiki`	`80`	`4321`	`WIKI_HOST_PORT`	Astro Documentation Wiki (HTTP)
`gxpchat`	`3000`	`3010` / `3002`	`CHATBOT_HOST_PORT`	GxPChat Unified Next.js Web Suite
`api`	`8000`	`8090`	`APP_HOST_PORT`	FastAPI Backend & FastMCP SSE Server
`postgres`	`5432`	`5434`	`POSTGRES_HOST_PORT`	PostgreSQL Relational System of Record
`qdrant`	`6333`	`6335`	`QDRANT_HOST_PORT`	Qdrant Vector REST API
`qdrant`	`6334`	`6336`	`QDRANT_GRPC_HOST_PORT`	Qdrant gRPC Interface
`memgraph`	`7687`	`7688`	`MEMGRAPH_HOST_PORT`	Memgraph Bolt Protocol
`memgraph-lab`	`3000`	`3006`	`MEMGRAPH_LAB_HOST_PORT`	Memgraph In-Memory Graph Web UI
`rustfs`	`9000`	`9120`	`RUSTFS_HOST_API_PORT`	RustFS S3-Compatible Object API
`rustfs`	`9001`	`9121`	`RUSTFS_HOST_CONSOLE_PORT`	RustFS Storage Administration Console
`langfuse`	`3000`	`3001`	`LANGFUSE_HOST_PORT`	Langfuse Observability Web Console
`clickhouse`	`8123`	`8124`	(Hardcoded Host Port)	ClickHouse OLAP HTTP Protocol
`clickhouse`	`9000`	`9001`	(Hardcoded Host Port)	ClickHouse OLAP Native Protocol
`redis`	`6379`	`6389`	(Hardcoded Host Port)	Langfuse Redis Queue & Cache
`minio`	`9000`	`9092`	(Hardcoded Host Port)	Langfuse Internal S3 Storage API
`minio`	`9001`	`9093`	(Hardcoded Host Port)	Langfuse Internal S3 Console
`docx-service`	`5050`	`5055`	`DOCX_HOST_PORT`	ASP.NET Core Regulatory Document API

When launched with --profile standalone: