Prerequisites

System Requirements

Requirement	Minimum	Recommended
Docker Engine	20.10+	27.x
Docker Compose	2.0+	2.30+
RAM	4 GB	8 GB
Disk Space	5 GB	20 GB
OS	Linux, macOS, Windows (WSL2)	Linux (amd64)

External Dependencies

FalkorDB

FluidRAG requires a running FalkorDB instance for knowledge graph storage and traversal. FalkorDB is not bundled in the Docker Compose stack — you must provide your own instance.

You will need:

• Host and port of your FalkorDB instance (e.g., 10.0.70.120:6397)
• Password if authentication is enabled
• A knowledge graph name (e.g., my_kg_v1)

FalkorDB runs on port 6379 by default. If you run FalkorDB in Docker, map it to a different port to avoid conflicts with the bundled Redis container.

# Example: running FalkorDB separately
docker run -d --name falkordb \
  -p 6397:6379 \
  falkordb/falkordb:latest

LLM API Key

FluidRAG supports multiple LLM providers. You need at least one API key:

Provider	API Key	Used For
OpenAI	sk-...	Embeddings, entity extraction, synthesis, agent execution
Anthropic	sk-ant-...	Agent execution, synthesis
Google	AIza...	Agent execution, synthesis, embeddings
OpenAI-Compatible	Varies	Custom endpoints (Ollama, vLLM, etc.)

Other supported providers: Groq, OpenRouter, DeepSeek, Fireworks, AWS Bedrock.

API keys are configured through the Admin UI after deployment — do not put them in your .env file.

Python (Development Only)

If you want to run FluidRAG outside Docker for development:

• Python 3.11+ (3.12 recommended)
• Dependencies are managed via uv (see pyproject.toml)

Network Requirements

FluidRAG containers need outbound access to:

Destination	Purpose
Your FalkorDB host	Graph database queries
LLM provider APIs (e.g., api.openai.com)	Embeddings, chat completions
etcd (bundled)	Configuration management (internal)
Redis (bundled)	Job queue, caching (internal)

No inbound ports need to be exposed beyond those mapped in docker-compose.yml.