Skip to main content

Use the AI Assistant

Open the project's AI assistant to build teams and workflows through natural language conversation, instead of configuring them manually.

Prerequisites​

  • An existing project (see Create a Project)
  • At least one LLM configuration in your organization (see Configure an LLM)
  • The Python engine must be running and accessible

Steps​

1. Open the AI Assistant​

Navigate to your project in the sidebar. Click the AI Assistant button (chat icon) in the project header or sidebar. This opens the assistant chat panel.

AI Assistant chat panel

2. Start a conversation​

Type a message describing what you want to build. The assistant understands the full ORQO data model -- projects, teams, agents, workflows, stages, assignments, triggers, tools, and skills.

Example prompts:

  • "Create a research team with two agents: a researcher and a writer"
  • "Build a workflow that scrapes a URL, summarizes the content, and writes a blog post"
  • "Add a cron trigger to the content pipeline workflow that runs every Monday at 9am"
tip

Be specific about what you want. Include agent roles, tool requirements, and workflow structure. The assistant uses a knowledge base with workflow patterns, agent design guidelines, and tool documentation to give you informed suggestions.

3. Review the assistant's plan​

The assistant typically responds in two phases:

  1. Informational updates -- Progress messages explaining what it is creating (these appear as status indicators)
  2. Questions -- If it needs clarification, it asks before proceeding (e.g., "Which LLM should I use for the team?")

The assistant calls tools behind the scenes to create teams, agents, workflows, stages, and assignments. You see real-time status updates for each tool call.

Assistant creating resources

4. Provide guidance when asked​

The assistant may ask you to choose between options:

  • Which LLM configuration to assign to a team
  • Which tools or skills an agent needs
  • How stages should be ordered or connected
  • Whether a workflow should await user input

Answer naturally. The assistant interprets your response and continues building.

5. Verify the results​

After the assistant finishes, navigate to the Teams or Workflows section of your project to review what was created. You can:

  • Open the Team Builder to see the visual graph of agents, LLM assignments, and skills
  • Open the Workflow Builder to see the stage flow, assignments, and routing
  • Edit any entity directly if you want to adjust the assistant's output

Reviewing created workflow

6. Validate and run​

Ask the assistant to validate your workflow before running it:

  • "Validate the content pipeline workflow"
  • "Preview the workflow execution plan"
  • "Run the workflow"

The assistant uses the workflow_run tool to validate, preview, and start workflow runs directly from the chat.

7. Manage conversations​

Your conversations with the assistant are persisted. You can:

  • Return to a previous conversation from the conversation list
  • Start a new conversation for a different task
  • Delete old conversations you no longer need

What's next​

Learn more​