Documentation

KoolBox Guides

Step-by-step instructions for every feature. From connecting your first GPU node to building production RAG pipelines.

Add a GPU NodeDeploy a ModelCreate a Knowledge BaseBuild a RAG PipelineCreate a WorkflowCreate a Chatbot

Guide 01

How to Add a GPU Node

  1. 1Navigate to Infrastructure > Servers
  2. 2Click Add Server
  3. 3Enter: display name, host/IP, port (22), username
  4. 4Choose auth: password or SSH private key
  5. 5Click Test Connection to verify
  6. 6Click Add Server to save

Note: The system will auto-discover hardware (GPU, VRAM, CUDA).

How to Add a GPU Node

Guide 02

How to Deploy a Model

  1. 1Navigate to Infrastructure > AI Models
  2. 2Click Deploy Model (preset) or Custom Model (HF/custom)
  3. 3Select the model and target node
  4. 4Click Deploy — the model pulls in the background
  5. 5Check status with refresh (turns 'ready' when complete)

Note: Supports Ollama registry models (Llama, Mistral, Phi) and Hugging Face GGUF.

How to Deploy a Model

Guide 03

How to Create a Knowledge Base

  1. 1Navigate to Knowledge > Knowledge Bases
  2. 2Click Create Knowledge Base
  3. 3Set name, description, embedding model, chunk size/overlap, target node
  4. 4Add content: upload files, fetch URL, fetch API, or add text
  5. 5Images: PNG, JPG, WebP auto-described by vision model if available
  6. 6Click Deploy to Node to push chunks to the remote server

Note: Supports PDF, DOCX, TXT, MD, HTML, CSV, JSON, and image files.

How to Create a Knowledge Base

Guide 04

How to Build a RAG Pipeline

  1. 1Navigate to Knowledge > RAG Pipelines
  2. 2Click New Pipeline, give it a name
  3. 3Drag nodes from the left palette onto the canvas
  4. 4Connect nodes by dragging from output to input handles
  5. 5Configure each node (click gear icon)
  6. 6In LLM Inference node: select both model AND execution node
  7. 7Click Save, then Test Pipeline to verify

Note: Node types: Document Loader, Text Splitter, Embeddings, Vector Store, Retriever, Prompt Template, LLM Inference, Guardrail, Redaction, Conditional Router, Integration, Custom Script, Workflow, Output.

How to Build a RAG Pipeline

Guide 05

How to Create a Workflow

  1. 1Navigate to Automation > Workflows
  2. 2Click Create Workflow, enter name and description
  3. 3Drag nodes: Start, AI Task, Integration, Form, Trigger, Gateway, Wait, End
  4. 4Connect nodes by dragging between ports
  5. 5Configure each node (click gear icon)
  6. 6Click Save, then Execute to test

Note: AI Task nodes run inference on remote GPUs via SSH. Gateway XOR for if/else, Gateway AND for parallel execution.

How to Create a Workflow

Guide 06

How to Create a Chatbot

  1. 1Navigate to AI Assistants > Manage Chatbots
  2. 2Click Create Chatbot
  3. 3Configure: name, system prompt, model, node, knowledge base, access level
  4. 4Save — the chatbot is immediately available in AI Chat
  5. 5Generate API keys for programmatic access

Note: Message flow: user sends → KB search → prompt construction → SSH to node → Ollama inference → response with metadata.

How to Create a Chatbot

Ready to Try It Yourself?

Start with the free Community plan. Deploy your first node and chatbot in minutes.

Get Started Free

FAQ

Frequently Asked Questions