Lino
Docs
Try Free
← Back to Home

Planning & Background Tasks

For big objectives (like importing entire folders, scanning large code repositories, or structuring multi-step whiteboards), Lino AI does not just execute blindly. It uses a structured planning and background task system to keep you updated.

Session Plan Checklists

When Lino AI encounters a complex task (typically requiring 3 or more distinct steps):

  • Plan Generation: The AI outlines a structured Session Plan (checklist).
  • Persistent Visibility: This plan is rendered as a progress bar strip in the AI panel (SessionPlanStrip).
  • Resilient Memory: The checklist is kept outside the normal chat history. Even if the conversation grows very long and old messages are compressed to save space, the AI will never forget your objective or偏离主题.
  • Real-time Progress: As the AI works, it marks items as Pending, In Progress, or Completed, giving you full visibility.

Background Tasks

Heavy operations (such as web research, local directory syncing, codebase scanning, and running terminal commands) run inside Lino's background task runner.

  • Non-blocking: Background tasks run silently without freezing your chat window, letting you continue typing or navigating your canvas.
  • Trace Notification Bubbles: When a background task starts, finishes, or encounters an error, a floating notification bubble appears in the bottom-right corner.
  • Control Center: From these notifications, you can:
    • Cancel running tasks.
    • Retry failed steps.
    • View Logs to read stdout/stderr from terminal commands or scanner outputs.

Subagents & Task Continuation

For highly complex, nested tasks, the main AI agent spawns specialized Subagents:

  • Isolated execution: Subagents execute targeted research or scanning tasks in parallel, returning consolidated findings without cluttering your main chat history.
  • Continuation Prompts: If a background subagent requires folder access permissions, credentials, or user input mid-run, it will post a prompt. Once you grant access or click approve, the agent automatically continues execution where it left off.