Writing Generators

Writing Generators starts the assistant with a bounded job and visible source. Instead of asking for “something better,” choose the document or selection, add constraints, and review the exact prompt before it is sent.

Four focused tasks

• Abstract asks for a draft grounded in the paper as written, without inventing claims.
• Summarize condenses the open document or selected passage.
• Paraphrase rewrites a selection while preserving its meaning.
• Thesis outline proposes chapters around material already present in the draft.

These are starting points, not one-click authorship. The assistant can miss nuance, weaken a qualification, or repeat a claim the source does not support.

Generate with context

1. Open a project, then open Writing Generators or run /generators.
2. Choose a task.
3. Select Document or Selection where the task allows it.
4. Add a concrete instruction, such as Keep it under 150 words and preserve all numerical results.
5. Read the prompt preview.
6. Choose Generate. Oleafly opens the assistant and sends the reviewed prompt.
7. Treat the response as a proposal and compare every claim with the source.

When to skip it

Write directly when the passage carries a personal voice, a delicate interpretation, an ethical judgment, or a claim whose evidence you have not checked. A generator cannot decide what your results mean or which contribution is defensible.

The selected source and prompt go to the assistant you configured. Nothing is sent until you choose Generate. A local model keeps the request on your computer; a hosted provider receives the reviewed prompt and source context. If no assistant is configured, Oleafly opens setup before sending anything.

Keep the request testable

Name the audience, length, facts that must remain, and changes you do not want. Then ask the assistant to flag missing evidence instead of filling gaps with plausible text.

For smaller changes inside the editor, use inline AI edits.