Skip to main content

Composer controls

Overview The composer separates the agent choice, AI model, execution mode and target application. Check each control before sending a request. Ask four questions: which agent handles this, which model reasons about it, which execution surface can it use, and which application may it access? Thinking level is a further model setting. In Ideatura
  1. Identify each selector above the message box.
  2. Choose the task context before sending.
  3. Check the model and thinking level.
  4. Review mode and target access together.
Check the result You can explain each selection without treating them as one setting.

AI models

Overview The model selector names the model used for the task. Available models and their supported thinking levels can vary; use the options shown for the selected route. Compare models using the same bounded task and an inspected result. A longer model list is less useful than examples showing the trade-off between quality, time and cost. In Ideatura
  1. Open the model selector.
  2. Choose an available model.
  3. Check its supported thinking levels.
  4. Run a small task and inspect the output.
Check the result The model is available and its result meets the task’s needs.

Thinking level

Overview Thinking level controls the reasoning setting offered by the selected model. Choose from the levels shown in the picker; available levels can differ between models. For a straightforward rename, begin with a modest setting. For a family with interacting constraints, compare a higher setting on a small test. More thinking is not a guarantee of a correct model. In Ideatura
  1. Select a model.
  2. Open its available thinking settings.
  3. Choose a level suited to the task.
  4. Compare inspected result, duration and recorded cost.
Check the result The choice improves the task outcome at an acceptable cost.

Attach files

Overview Use the plus control to supply task references. Before sending, check the attached file names, versions and the information each reference establishes. A drawing can establish geometry or dimensions; a photo can establish appearance; a specification can establish a requirement. Explain those roles when attaching several sources. In Ideatura
  1. Attach the reference files.
  2. Open or inspect their previews.
  3. Name scale, units and source roles in the request.
  4. Remove accidental attachments before sending.
Check the result The assistant receives the intended references and their interpretation is clear.

Skills and Guidelines

Overview The composer includes a skill control and can show a Guidelines chip. Use a skill for a repeatable method; use the current request for the specific input, scope and result. Inspect which instructions are active before sending. Removing a chip from a draft is a different operation from deleting a shared skill or editing workspace Guidelines. In Ideatura
  1. Read the relevant skill or Guidelines.
  2. Select the needed skill in the composer.
  3. Check the visible chips.
  4. Send a task that explicitly uses the method.
Check the result The task has the intended instructions without unintended extra material.

Pin context

Overview The prompt and viewer expose Pin controls. Use the pin state to keep important material associated with the conversation where the current interface supports it. Attach, Pin and Clone are separate actions. Confirm what is pinned, what is added to the next message, and what has become an editable copy. In Ideatura
  1. Open a useful reference.
  2. Inspect Attach, Pin and Clone actions.
  3. Pin the intended material.
  4. Check the visible pinned state before the next request.
Check the result You can identify the material being carried into the task.

Send and stop

Overview Send starts the request using the currently selected context and controls. During work, read the progress and tool/output messages before deciding whether to refine or stop. History’s Stop all is broader than stopping one task. Stopping work does not by itself reverse an application change already made. In Ideatura
  1. Check the prompt, target and attachments.
  2. Send the bounded task.
  3. Read progress and the final report.
  4. Inspect the actual result and issue a precise correction.
Check the result You know whether work completed, stopped or needs attention.

Usage indicators

Overview Hover over a usage indicator and read its tooltip before interpreting the number. For account allowances and Workspace Storage, open Settings. Context-window usage, token usage and credits measure different things. See Context window and Tokens and credits to compare them. In Ideatura
  1. Inspect the indicator tooltip or details.
  2. Compare before and after a bounded task.
  3. Read Settings → General for the account allowance.
Check the result You can identify the unit, time period and scope of each figure.