Drawer Tab

Drawer Tab

The Hone tab in Lumiverse's sidebar drawer is the configuration hub and the stage picker. It's where you set up model profiles, edit presets, toggle auto-refine, and flip between pipeline stages after a multi-stage refine.

You can also access it by right clicking on the Hone widget (hold for mobile).

Hone Control (quick actions)

The big buttons at the top.

Hone Last AI Message

Refines the most recent assistant message in the current chat. The button shows a spinner during generation, and a small badge appears on the drawer tab title (..., then ✓ or ✗, then clears after a few seconds).

Undo Refinement

Appears in place of Hone Last when the last AI message is refined on the current swipe. Two-tap confirm: first click arms, second click within 4 seconds fires. Click outside the drawer to disarm.

Hone All AI Messages

Bulk-refines every assistant message in the active chat. Two-tap confirm. Per-message progress shows on the drawer badge (3/12). A summary modal fires at the end only if any refinements failed. Fully-successful bulk runs are silent.

Stage picker

After a multi-stage or parallel refinement, the drawer shows one button per intermediate output:

• Step 1: Grammar Pass — sequential pipeline stage 1.
• Step 2: Prose & Voice — stage 2.
• Agent 1: Grammar Agent — parallel proposal 1's final output.
• Agent 2: Prose Agent — proposal 2's final output.
• Step 1: Aggregator — aggregator stage 1.

Clicking a stage button swaps the message's live content to that stage's output and shows a diff against the original. Flipping between stages is instant.

The original is preserved across stage flips. No matter which stage you flip to, clicking Undo restores the pre-refinement content.

Stages stream in as they complete.

Automation

Auto-Refine AI

Toggle to automatically Hone every AI message after it finishes generating. Same UI feedback as a manual click: spinner, badge, diff modal, optional ding.

If chat_mutation permission is denied, auto-refine silently no-ops. Failures don't interrupt your chat with a modal, but they show up in debug logs.

Auto-Enhance User (coming soon)

When the feature lands, this will auto-enhance your drafts on send. Until then, manual enhancement via the input-bar Hone button is the only path. See Known Limitations.

Models sub-tab

Full model profile setup. See Model Profiles for the complete reference.

Output sub-tab

The output preset editor. Everything about what happens when you refine an AI message lives here.

Layout:

• Preset bar. Dropdown selector, Rename / Duplicate / Export / Import / New / Delete.
• AI Message POV selector. auto, 1st, 1.5, 2nd, 3rd. See Context and POV#pov.
• Sub-subtab bar:
  • Pipeline. Stage editor. See Pipeline Editor.
  • Prompts. Prompt library. See Prompts and Macros.
  • Shield. Literal-block shielding config — master toggle, include/exclude regex, reset to defaults. See Prompts and Macros#shielding-literal-blocks.
  • Context. Token budgets for {{context}} and {{lore}}.

Built-in presets show a padlock banner and disable all inputs. Click Duplicate to make an editable copy.

Edits to a custom preset save automatically. There's no Save button.

See Presets for everything preset-related.

Input sub-tab

Same layout as Output, but for the input preset. The pipeline that runs when you click the input-bar Hone button.

Differences from Output:

• User Message POV instead of AI Message POV.
• No 1.5 POV option. User messages are always from the user's perspective.
• No Shield sub-subtab. Shielding only applies to output refinement.

Misc sub-tab

Barebones right now, might add stuff to this in the future.

Badge behaviour

The drawer tab title gets a small badge during operations:

• ... — refine in flight.
• 3/12 — bulk refine progress.
• ✓ — successful completion (clears after 3 seconds).
• ✗ — error (clears after 3 seconds).

This is the fastest way to know Hone is working when the drawer is closed.