> ## Documentation Index
> Fetch the complete documentation index at: https://docs.quadrillion.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Tasks

> Organize work into a network of tasks with dependencies and assignments.

When you're working on complex research with multiple steps and dependencies, Qualia's task system helps you organize the work. Tasks form a directed acyclic graph (DAG) where each task can depend on others and be assigned to different agents.

## Understanding tasks

A task represents a unit of work:

* **Title**: What needs to be done
* **Objective**: Detailed description of the goal
* **Status**: Where the task is in its lifecycle
* **Dependencies**: Other tasks that must complete first
* **Assignment**: Which agent is working on it

Tasks are identified by IDs like `T-1`, `T-2`, etc. within a project.

## Task statuses

Tasks move through a defined lifecycle:

```mermaid theme={null}
flowchart LR
    planned --> in_progress
    in_progress --> completed
    in_progress --> failed
    in_progress --> cancelled
    in_progress <--> paused_by_user
```

| Status               | Meaning                             |
| -------------------- | ----------------------------------- |
| **planned**          | Task is defined but not yet started |
| **in\_progress**     | An agent is actively working on it  |
| **completed**        | Task finished successfully          |
| **failed**           | Task encountered an error           |
| **cancelled**        | Task was stopped before completion  |
| **paused\_by\_user** | You paused the task; can be resumed |

### Blocking

Tasks can be **blocked** when they depend on upstream tasks that haven't completed yet. Blocking is tracked separately from status — a planned task may be blocked or unblocked depending on whether its dependencies are satisfied. When upstream tasks complete, blocked tasks automatically become unblocked and available to start.

## The Tasks sidebar

Open the **Tasks** sidebar from the activity bar (far left) to see your project's task graph.

### Views

Toggle between two views:

* **Gantt view**: Timeline showing tasks by assignee, with dependency lines connecting related tasks
* **DAG view**: Graph visualization of task dependencies

### Creating tasks

Click **New tasks**, choose which agent should handle the request, and describe the work in natural language:

> *"Analyze this dataset: clean the data, run three different models, compare their performance, and summarize the best approach."*

The request is handed to the agent you picked, which plans the work into linked tasks with appropriate dependencies and assigns each one to an executor.

### Modifying and deleting tasks

You can edit or remove **planned** tasks before they start:

* **Edit**: Click a planned task to modify its title, objective, or dependencies
* **Delete**: Remove tasks that are no longer needed

Once a task moves to **in\_progress**, it can no longer be edited or deleted — pause it first if you need to make changes.

### Reassigning tasks

In Gantt view, drag a task bar to reassign it:

* Drop on another agent's row to reassign to that agent
* Drop on **New agent** to create a new agent for the task

Terminal tasks (completed, failed, cancelled) cannot be reassigned.

## Tasks and agents

Tasks and [agents](/agents) work together:

* Agents can have multiple tasks assigned to them
* An agent works through its tasks, respecting dependencies
* When you create parallel agents, each can take on different tasks

The agent confirmation setting in [Settings > AI](/settings/ai) controls whether new agent creation requires your approval.

## Tasks and knowledge

When an agent completes a task, it typically produces **claims** in the [Knowledge System](/knowledge). This claim captures what the agent learned or concluded, with links back to the evidence (code cells, files, other claims).

In notebook mode, tasks can capture values and figures from their notebook kernels. In script mode, tasks preserve durable outputs as workspace files, which can be captured as file evidence. A notebook-only evidence requirement cannot be delegated around in script mode; attach a notebook for the workflow or skip the requirement with an explanation.

## Viewing task progress

### Inline summary

In the Chat sidebar, each agent shows a collapsible task summary:

* Count of tasks by status (completed, in progress, planned, etc.)
* Click to expand and see individual tasks
* Link to open the full Tasks view

### Project selector

If you have multiple projects, use the project selector at the top of the Tasks sidebar to switch between them.