Documentation · Use Cases
Use Case Patterns
Arion Flow focuses on a small set of powerful patterns: brand content, RAG, model fine-tuning, AI agents, and agentic automation. Each can run in isolation or be combined into end-to-end flows.
Arion Node (AI-Ready Compute)
Private, isolated compute environments for running AI workloads within your sovereignty boundary.
- Input: Docker containers, Python scripts, or Jupyter notebooks.
- Output: GPU-accelerated inference, training, or batch processing.
- Best for: data science teams, custom model serving, and secure research.
RAG (Retrieval-Augmented Generation)
Ground model outputs in your own documents, knowledge bases or product content.
- Input: PDFs, docs, web content or structured data.
- Output: grounded answers, summaries, and explanations.
- Best for: support, sales enablement, policy and internal search.
Model fine-tuning
Adapt models to your domain, datasets and style inside controlled workspaces.
- Input: curated training examples and datasets.
- Output: improved performance on your specific tasks.
- Best for: strong brand tone, domain-heavy content, classifiers.
AI agents
Build agents that can call tools, use your data and take guided actions within your environment.
- Input: a goal plus access to tools and context.
- Output: multi-step reasoning, tool calls and proposed actions.
- Best for: research tasks, internal copilots, structured workflows.
Agentic workflow automation
Combine models, agents, data sources and human approvals into end-to-end flows.
- Input: events, triggers, and business rules.
- Output: actions taken across tools with auditability.
- Best for: recurring workflows that justify automation, not just ad-hoc prompts.
Next steps
Once you know which patterns you want to start with, the next step is to consider data, access and governance.
Continue to security & governance →