Canvas — Agent Builder

The Canvas is Queria's visual builder for creating AI agent pipelines. With a drag-and-drop interface you can connect functional components — document retrieval, AI reasoning, logical routing, user interaction — to build automated workflows without writing a single line of code.

The result is a custom AI agent that you can publish as a dedicated assistant for your organization or a specific team.

Canvas vs Direct Chat

Not everything requires a Canvas. Direct chat is fast and flexible; use the Canvas when you need structured, repeatable logic.

Situation	Use Chat	Use Canvas
Ad-hoc question about a document	Yes	No
Guided analysis with predefined steps	No	Yes
Simple natural language response	Yes	No
Automatic routing by request type	No	Yes
Flow with structured input form	No	Yes
On-the-fly document comparison	Yes	No
Repeatable business process (e.g. claim analysis)	No	Yes
Automated multi-source integration	No	Yes
Iteration over lists of items	No	Yes
Generic conversation with memory	Yes	No

Core Concepts

Node (Component)

A node is a functional block with a precise role: execute an LLM call, retrieve documents, evaluate a condition, collect input from the user. Each node has configurable parameters and produces one or more output values.

Edge

An edge is the connection between two nodes. It defines the execution order and data path: the output of one node becomes available to subsequent nodes through variable references.

DAG (Directed Acyclic Graph)

The Canvas executes nodes following a DAG: each node starts only when all its predecessors have completed. The engine automatically handles parallelization where possible.

Variables

Variables are references to values produced by nodes, using the syntax {{component.key}}. For example, {{retrieval_1.formalized_content}} contains the text retrieved by the first Retrieval node. Variables can be used in prompts, conditions, and configuration parameters of any subsequent node.

Streaming Execution

The Canvas streams results in real time via SSE (Server-Sent Events). Users see responses token by token, exactly as in direct chat, even within complex flows with multiple LLM components.

Session and Suspension

Some components (such as UserFillUp) can suspend execution while waiting for additional user input. The session resumes transparently when the user provides the required information.

Available Components

The Canvas provides 15 components divided into six categories:

Category	Components
Basic Flow	Begin, Message, Switch
Artificial Intelligence	LLM, Retrieval, Categorize, Agent
User Interaction	UserFillUp
Data Manipulation	VariableAssigner, VariableAggregator, DataOperations
Loops	Iteration, Loop, ExitLoop
Composition	Invoke

For complete documentation of each component, see the Components page.

Sector Templates

Queria includes 6 ready-to-use templates, one for each supported sector. Each template is a starting point you can customize:

Sector	Template	Main Components
Insurance	Claim Analysis	Begin, Categorize, Retrieval, LLM, Switch, UserFillUp, Message
Legal	Assisted Regulatory Search	Begin, Retrieval, LLM, Message
Food	Product Compliance Check	Begin, Retrieval, LLM, Message
Chemical	Safety Data Sheet Analysis	Begin, Retrieval, LLM, Message
Pharmaceutical	Clinical Trial Research	Begin, Retrieval, DataOperations, LLM, Message
Tax	Tax Compliance Assistant	Begin, Categorize, Retrieval, LLM, Message

For details on each template, see the Sector Templates page.

Canvas Guide Sections

• First Workflow -- Create your first Canvas in minutes: from the editor interface to the basic RAG flow.
• Components -- Complete documentation of all 15 components with parameters, outputs, and use cases.
• Variables and References -- How to use the {{namespace.path}} syntax to connect nodes and pass data between components.
• Practical Examples -- Three progressive examples: simple RAG, Smart Router, and a complete Insurance Claim Analysis.
• Sector Templates -- The 6 predefined templates with flow schema, data sources, and customization tips.

Start from templates

If you are taking your first steps, start from a sector template rather than from scratch. In a few clicks you have a working flow to examine and adapt to your needs.

---

Queria v3.1.2 -- Canvas Agent Builder