How to Get the Best Answers Out of Pluely v1

Better inputs get better answers: attach real context, ask for exactly what you want, refine with follow-ups, and match the model to the task. This guide walks through each of those levers.

1. Attach context instead of describing it

Ask mode takes more than typed text — attach a screen capture, an image, or a file to a question instead of describing what's on screen. The real thing beats a secondhand description every time:

1. Instead of typing out an error message or a chart, capture the screen or region directly.
2. Attach the capture (or a file/image) to your question in the overlay.
3. Ask your actual question alongside it — "what does this error mean" next to the captured dialog.
4. Pluely answers with the attached content directly in view, not your paraphrase of it.

For context you reuse constantly, set up a custom system prompt with attached knowledge files — Pluely uses them with every answer, so you stop re-explaining the same background.

2. Be specific about what you want back

A vague question gets a vague answer. Compare "what does this mean" with "explain what's causing this specific error and how to fix it" — the second gives Pluely a clear target (a fix, not just an explanation). Specificity helps most on:

• The format you want (a short answer, a list, a full explanation).
• What you'll do with the answer (send it to someone, act on it yourself, learn from it).
• Any constraint that matters (a specific language, tool, or length).

3. Use follow-ups to refine instead of restarting

First answer not quite it? Don't start over with a longer, more complicated question. Use a follow-up to narrow, correct, or redirect — Pluely already has the original exchange in view, so a short "make that shorter" or "what about for a different case" gets you there faster than rewriting the whole prompt.

4. Pick the right model for the task

The model picker at the top of the overlay groups options into Your CLI, Your Providers, and Pluely Models — see Model selection for the full picture. Matching model to task matters:

• Quick factual questions or short rewrites don't need your most capable (and often slowest) model.
• Nuanced writing, complex reasoning, or anything feeding a real decision deserves a stronger model, even if it's slower.
• A connected local CLI or provider key takes priority automatically — if an answer feels off, check the picker; you may be on a model you didn't intend for that question.

5. Set up Listen mode context correctly before it matters

For live conversations, get system audio capture and the transcription language right before the call starts. A clean transcript is the raw material for every later summary and follow-up — and for automatic responses, which draw on your selected prompt, its attached files, and the live transcript. Messy input caps how good any answer can be, no matter how well you prompt afterward.

Putting it together: a quick checklist

1. Attach the actual screen, image, or file instead of describing it in words.
2. State the kind of answer you want, not just the topic.
3. Refine with a follow-up rather than rewriting a whole new prompt.
4. Check the model picker matches the task's complexity.
5. For live sessions, confirm system audio and language settings before you start.

Tips

• Answer feels shallow? Attach more direct context before assuming the model is at fault.
• Note which model you prefer for which task — quick chat vs. serious writing vs. code — and switch deliberately rather than staying on the default.
• Skim your history and library occasionally; noticing which of your past prompts got the best answers sharpens your own habits.
• Ask a follow-up purely to double-check something — clarifying is cheap and faster than acting on an answer you're unsure about.

Next

• Model selection — how the model groups work and how priority is decided.
• Auto responses — let Pluely answer on its own during live sessions.