Allow All & goal runs
By default Literess proposes and you approve. Two switches change that pace. Allow All skips the per-change approval click, and a goal run lets her keep working on a big job turn after turn until it’s done. A Stop button keeps you in charge the whole way.
On this page
What does Allow All change?
With Allow All off (the default), every change Literess proposes waits on a confirmation chip until you approve it. With it on, her proposals apply the moment she makes them: new block translations, glossary and style guide edits, document moves, workflow runs, workspace settings. She keeps working through the task in the same breath instead of stopping after each proposal. The chip still appears in the chat, already marked “Applied automatically”, so you keep a record of everything she did.
One exception holds no matter what: inviting a teammate always asks first, because an email leaves your workspace.
How do I turn Allow All on and off?
Two places flip the same setting: the switch under Settings → Notifications, and the sparkle button in the header of the Literess panel, lit while the mode is on. You can also just tell her to turn it on. That request arrives on a chip you approve, so the mode can never switch itself on without one deliberate click from you. Turning it off works from anywhere, mid-task included.
How do I stop Literess mid-reply?
While she works, the send button in the composer becomes a square Stop button. Press it and she halts where she is: whatever she already applied stays applied, the words she already wrote stay in the chat, and anything she hadn’t started is dropped rather than left half-done. This works in every conversation, and if a goal run is underway it pauses the run too.
What is a goal run?
A goal run is how Literess handles work that spans many blocks or pages: translate the whole document, chase one recurring mistranslation through all six chapters, check every block against the style guide. Instead of doing one step and waiting for your next message, she sets herself the goal and keeps working turn after turn until it’s reached. There’s no command to learn: describe the job and she takes it on.
While a run is open, a thin strip sits above the composer with the objective, the turns she’s used, the words she’s written, and a small bar for the run’s allowance. Pause on that strip stops the run once the current turn ends. Resume picks it back up.
For bulk translation she normally reaches for the pipeline first: start a translate workflow, watch it run, then put her own effort into reviewing and fixing the result. Her hand-written work is the expensive, careful layer, so she saves it for the places it beats the pipeline. If you want her to write everything herself, say so once and she will.
How does a goal run end?
- She reaches the objective, declares the goal complete, and closes with a summary of what she did.
- She hits something she can’t work around, says so plainly, and tells you what would unblock her.
- The run uses up its built-in allowance of turns and compute. She finishes the turn cleanly and reports how far she got. A fresh goal continues from there.
- You press Stop or Pause. The run waits until you resume it.
What do Literess’s actions cost?
Allow All and goal runs change who presses the button, never the price. The same work bills the same words it always did, from the same balance. On a team document her writes bill the team pool, like any run there.
| Action | Cost |
|---|---|
| Writing a block translation that gets applied | 3 words per source word (the premium rate — her writing runs on the premium model) |
| Running a workflow she set up | The run’s normal estimated cost |
| Re-activating wording a block already has, or switching versions | Free |
| A proposal you reject, or that stays pending | Free |
| Reading, searching, navigating, filling in dialogs | Free |
Does Allow All let Literess spend words without asking me each time?
Yes, for the actions she proposes in chat. A workflow she starts or a translation she writes bills exactly what it would have billed had you pressed the button yourself, and the resolved chip still shows what ran. If you’d rather approve spending case by case, leave Allow All off (the default), or switch it off mid-task. The change takes effect on her next turn.
Can Literess translate my whole document on her own now?
Yes. Describe the job (“translate this whole document”) and she sets a goal run. Usually she starts a translate workflow for the bulk and then reviews the result herself. If you want every block written by her hand, say so and she will. The workflow route bills that run’s normal price; her hand-written blocks bill the premium rate (3 words per source word), since she writes with the premium model.
What keeps a goal run from running forever?
Each run has a built-in allowance of turns and compute. When it’s used up she finishes the turn she’s on, reports her progress, and stops. You can start a fresh goal to continue. Your word balance is a second brake, since her writes stop when it runs out, and Stop and Pause are always one click away.
Does pressing Stop undo what Literess already did?
No. Changes she already applied stay applied, the part of the reply she already wrote stays in the chat, and only the work she hadn’t started yet is dropped. Nothing is rolled back. If you want a block back the way it was, its earlier versions are still in the block’s history.
Why did my goal run pause on its own?
Three causes. You pressed Stop, which pauses the run along with the turn; she finished a turn without finding anything left to act on, so the run parks instead of spinning; or an error interrupted a turn. Press Resume on the goal strip to continue, or tell her what to do next.
With Allow All on, can she also press buttons in dialogs for me?
No. Allow All covers the actions she proposes in chat. On an open dialog (a merge, a generation, a review) she still only fills things in. The button that starts or applies the work stays yours to press.
Still stuck? Ask Literess in the app, or write to [email protected].