Help

Around the screen

1. The explorer (left): your projects, their files, your chats, and your maps.
2. The open tab (middle): a file, a doc, a chat, or a map. Tabs sit across the top, like a browser.
3. The chat panel (right): the conversations about the open file.
4. The start page: a new tab shows your recent files and Ask Lyremi for a quick chat.

Good to know

• Chats belong to files. A question about a report lives on that report.
• Everything Lyremi makes (a doc, a brief, a draft) becomes a file you own and can edit.

1. In the explorer, choose New project… and name it.
2. Right-click the project (or tap ⋯) and choose:
   • Upload files…: PDFs, Word, Excel, PowerPoint, CSV, text, and images, up to 50 MB each. You can also drag files onto the explorer. A zip file is unpacked: each file inside it comes in on its own (up to 100), and anything Lyremi can't read is left out.
   • Paste text…: notes, an email, anything you can copy.
   • Add a link…: Lyremi saves the page's readable text, so it's still there if the page changes.
   • Add from Google Drive: once Drive is connected in Settings › Connections.
3. Click a file to open it. PDFs, documents, sheets, slides, and images all open in Lyremi.
4. To move files, drag them onto another project, or right-click › Move to…
5. Chats started from the start page are listed under Chats. To file one with a project, drag it onto the project, or right-click it (or tap ⋯) › Move to project…. It stays the same chat and keeps reading only its own sources. Move back to Chats takes it out again.

Files with no project sit under No project.

Deleting several at once

• Files and docs: select them (Ctrl-click or ⌘-click to pick several, Shift-click for a run), then Delete… in the bar at the bottom of the explorer, or right-click one of them › Delete N items…. You see the list first. Chats that used them stay, marked where the files were.
• Chats: right-click Chats (or a project) › Delete chats…, or right-click a chat › Delete…. Tick the ones to go (Select all helps), then Delete.
• Chats on a file: open the file's chat list and choose Delete… at the top.
• Deleting is permanent. A chat goes with its messages, pictures, and pins; docs it made stay. Chats Lyremi is still working in can be deleted once they finish.

1. Open a file, open the chat panel, and choose New chat.
2. To add files, select them in the explorer and choose + Add sources, or use + in the message box › Add from Lyremi. You can also paste a screenshot (Ctrl+V, or ⌘V on a Mac) or drop files into the message box: they're saved in the project and join the chat.
3. Type your question and press Enter (Shift+Enter for a new line). Or click the mic and speak: your words appear as you talk, and you can edit them before sending.
4. Choose a mode: Fast for everyday questions, Thinking for harder ones, Pro for the deepest reasoning (it uses your allowance faster, and isn't in every plan).
5. Stop ends a reply and keeps what arrived. Closing the tab doesn't stop it: the reply is there when you come back.
6. Under a reply: Copy, Save as doc, Pin, and Regenerate. To change your question, edit your last message.

Good to know

• A chat remembers the conversation. Each question you ask includes every earlier message in that chat, so you can say "shorten that" or "now compare it with the last one." Chats don't share memory: a new chat starts fresh.
• Files are read in full, every time. Every file attached to a chat is read whole with each question, never a summary, so answers come from the entire file. That's true even for a quick follow-up, so a long chat with large files uses more of your allowance. Remove files a chat no longer needs, or start a new chat for a new topic.
• While Lyremi works, the chat says what it's doing (reading your files, searching, writing), and a spinner shows on the file's tab and the chat in the explorer.
• Voice typing uses your browser's own speech recognition (Chrome, Edge, and Safari; not Firefox). The first time, your browser asks to use the microphone.
• Lyremi searches the web when a question needs it, and lists its sources. When your files already answer the question, it answers from them without searching.
• For a picture, choose + › Create image, then describe it.

1. In the explorer, open a project and click Knowledge (the first row in the project), or right-click the project › Knowledge…
2. Under Instructions, write a few lines Lyremi should always keep in mind here: who the work is for, the voice, what to avoid.
3. Under Always follow, add rulebook files from the project: a style guide, a brand voice sheet, a funder's requirements. You can also right-click a file › Add to Knowledge (always follow). Rulebooks are marked Rulebook in the explorer.
4. Choose Save. From your next request, chats in the project show Following … Knowledge, and Lyremi applies it without you attaching anything.

Good to know

• Your request decides what to write: the topic, sections and length. Knowledge decides how, and its rules (like "never invent quotes") still apply unless you say to set one aside, as in "em dashes are fine this time". If a request asks for something a rule forbids, Lyremi says so instead of faking it.
• Knowledge is treated as rules, not sources: Lyremi applies it rather than quoting it.
• Rulebooks are read in full with every request, so long ones use more of your allowance. Keep them to what Lyremi should actually follow.
• Each project has its own Knowledge. Chats outside a project don't use any.

1. Make a doc: New doc, Save as doc under a reply, or Make an editable copy of a PDF, Word file, or slides.
2. Write and format with the toolbar, or type / for headings, lists, tables, and images. Lyremi saves as you type.
3. To ask for changes, attach the doc to a chat and ask ("Make the second section shorter").
4. Changes arrive in the doc as suggestions. Accept or reject each one, or Accept all / Reject all. Nothing changes until you do.
5. Version history lists earlier states of the doc; restore any of them.
6. Download gives you a Word file or a PDF. Download › Save to Google Drive puts a copy in a "Lyremi" folder in your Drive as a Google Doc, ready to share. Each save makes a new copy, so nothing in Drive is overwritten.

1. Right-click a project › Show map, or a file › Show on map.
2. Read it left to right: sources, then chats (the dots), then what they made.
3. Click a card to light up its path; double-click to open it.
4. Drag cards to arrange them. Tidy up lays the map out again.
5. Drag a file onto a chat's dot to add it to that chat.
6. Select files, then Start a chat or Composer Agent.
7. Right-click a card (or select it and tap ⋯) to remove it from a chat or from the project. Nothing is deleted.
8. Each Research Agent run is one box, named after its brief. Click its arrow (or double-click it) to list its files; each row works like a card. Click the box itself to select all its files, for example to write from them with Composer Agent, and drag it to move them together.

1. Under a reply, choose Pin to pin all of it.
2. To pin part of a reply, highlight the passage, then choose the pin button that appears.
3. Pins sit at the top of the chat. The Pinboard gathers them from every chat on a file.
4. Click a pin to jump to where it came from. Unpin removes it.

1. In a chat, choose + › Research Agent, type your question, and send.
2. Check the plan: edit, remove, or add the questions it will look into; add sources to prefer or avoid; choose how the brief is written; see the likely cost.
3. Choose Start research. It takes a few minutes and keeps going if you close Lyremi. Stop ends it, and nothing is saved.
4. You get a brief (every claim linked to its source), the pages it relied on saved in your project with a Research Agent label, a summary in the chat, and connections on the map. In the explorer, each run's brief and pages are gathered in one group, named after the brief, so they don't mix with your own files; click it to open or close it. Right-click it (or tap ⋯) to Show on map, Open the brief, or Select its files to start a chat or write from them with Composer Agent.
5. Remove what this research added takes it all back out. Your own files are never touched.

Some sites block automated reading; the card lists those pages so you can open them yourself.

Chat or Composer Agent? Fast, Thinking, and Pro answer in the chat, in seconds. Composer Agent is an agent: it takes a few minutes and gives you a finished doc in your project. Its cards are purple, with a pen, so you always know which one is working.

Two ways to use it

• The mode menu (next to the send button): choose Composer Agent, and under it the mode it writes with. Send your request, and it goes.
• When a request looks like a document ("write a two-page proposal for…", "a 1,500-word blog post on…") and you're in Fast, Thinking, or Pro, Lyremi asks first (on plans that include Composer Agent): Write it with Composer Agent (with its likely cost, in dollars and as a share of your allowance, and about how long it takes) or Answer with Fast. Tick Don't ask again in this project to stop the question there; turn it back on in the mode menu (Offer Composer Agent when a request looks like a document).

What Composer Agent does

1. Plan: reads your request, the chat's files, and the project's Knowledge, and lays out the sections and how long each should be.
2. It starts on its own when the request is clear. It checks with you first when:
   • something essential is missing (it asks one short question; answer, or Skip: use your judgment);
   • the piece is long and you didn't give an outline (you see its outline first);
   • it would use more than 10% of your monthly allowance (you see the plan and cost first).
3. Gather: searches the web for what each section needs and keeps the real pages, so links go straight in. Links it didn't find, or that you didn't give it, are taken out, never guessed.
4. Write: a few sections at a time, so long pieces are actually long.
5. Polish: an editor's read of the whole draft, fixing what breaks your request or the project's rules, and writing thin sections again.
6. Hand over: the doc opens in your project. At the end, For your review recaps your request and gives Composer Agent's own assessment: where it falls short or departs (and why), what it couldn't source, and the facts most worth checking. Delete it when you're done.

The card shows each step, Stop, and the cost: estimated before, actual when done. It keeps going if you close Lyremi.

Keeping its sources: Composer Agent reads web pages while it works but doesn't save them. When it's done, its card offers Save the web sources: the pages the doc links to become files in your project, grouped with the doc in the explorer and on the map, ready for later chats and pieces. Pages a site won't let Lyremi read are listed so you can open them yourself. For sources you want to vet first, or to share across several pieces, run Research Agent first and compose from what it saved.

With an outline first: + › Composer Agent, outline first (or Composer Agent on the map) always shows the outline and waits. Rename, reorder, or add sections; for each gap, choose Leave out, I'll write it, or Research this; choose whether it may search the web, and the mode; then Write it.

Afterwards, ask in the chat for changes; they arrive in the doc as suggestions. Remove this draft takes it back out.

When to use it: you're starting something with several parts and aren't sure where to begin. Composer Agent writes a finished doc for you; Roadmap Agent gives you the structure, and you write.

1. In a chat, choose + › Roadmap Agent (on a phone, tap Roadmap Agent above the message box). On a new tab, Start a content project with a roadmap does the same. Describe what you're making, who it's for, and any deadline. If your first message sounds like the start of a project, Lyremi offers a roadmap; Just answer skips it.
2. A project first: a roadmap's docs live in a project. If the chat isn't in one, name a new project or choose one you have.
3. Your files: if the project has files, Lyremi asks whether to build from them. Choose the ones that help. A doc you've already started can become one of the parts instead of a new empty doc.
4. Check the plan: you see the parts, in order, before anything is made. Rename a part, change what it's for, move it up or down, remove it, or add one. Nothing is made until you choose Make the docs (the button says how many, for example Make 6 docs). Cancel stops it, and nothing is made.
5. What you get: one doc per part in the project, each opening with a short brief: what the part is for, questions to answer, and which of your files help, with room for your draft. The map shows the parts in order, each marked Empty or Started.
6. Remove on the card takes the roadmap back out. Docs you've written in are kept; your own files are never touched.

How many: every plan includes some roadmaps a day (Settings › Plan and billing shows how many). After those, each one is a small extra from your balance, and the price shows above the message box before you send. One that doesn't finish doesn't count.

1. Phone: the bar at the bottom has Chats (every chat, newest first; the chips at the top show one project's), Files, Messages, and +.
2. + starts a new chat (in the project you're looking at, or choose another), takes a photo, uploads a file, pastes text, adds a link, or makes a project.
3. In a chat, Research Agent, Composer Agent, and Roadmap Agent sit right above the message box: tap one, then say what you want. The message box stays above the keyboard.
4. Docs open to read. Select text to comment on it in a shared project, and Ask about this doc starts a chat about it. Writing and editing docs, and the map, are for a larger screen.
5. Add Lyremi to your home screen and it opens full screen, like an app. iPhone: in Safari, tap Share, then Add to Home Screen. Android: in Chrome, tap the menu, then Add to home screen or Install app.
6. Offline: once you've opened Lyremi, it opens without a connection too, with your chats, files, and docs. Changes wait and save when you're back; Lyremi needs a connection to answer.
7. Tablet: Lyremi works as it does on a computer. Where a computer uses right-click, tap ⋯.

1. Open the account menu (top right) › Settings. It opens over your work, with a menu on the left: General, Plan and billing, Usage, Balance, Workspace, Connections, and Your data. Type in Search to jump to any setting.
2. Usage shows how much of this month's AI allowance you've used, and when it starts again. Your month runs from the day you joined: join on the 14th, and it starts again on the 14th of each month. Research Agent and Composer Agent show their likely cost before you start, and Composer Agent shows what it actually spent when it's done.
3. Storage shows how much you've stored, of your cap.
4. At 80%, a note appears above the message box. At 100%, new replies and agent runs pause until your allowance starts again (the date is in Settings › Usage); your files, docs, and chats still work.
5. Plan and billing names your plan, with what it includes. Every plan has Fast chat, docs, and the map. Thinking, Research Agent, Composer Agent, Pro, and creating images are included, available as extras paid from your balance, or not included, depending on the plan. Extras show their price in the menus and on the Research Agent and Composer Agent cards before anything runs. Anything your plan doesn't include still shows in the menus, marked Not included in your plan.
6. Balance shows what you have: a monthly premium balance on plans that come with one (spent first, and it starts again each month), a one-time welcome balance to try an extra, and anything added (which doesn't expire). The history lists every charge, with what it was for and the chat it was in, and refunds: an extra that fails, or that you stop before anything comes back, is refunded. Extras don't use up your everyday allowance.
7. Free replies a day: some plans include a number of everyday replies a day, refilling at midnight. When the day's or the month's included replies are used, a note above the message box offers Keep going with my balance, and replies then come from your balance until your month starts again. What each one used shows in your balance history. You can turn it off in Settings › Balance.
8. Web searches: some plans include a number of web searches a month, shown in Settings › Usage. At 80% a note appears above the message box; once they're used up, web search pauses until your month starts again, and Lyremi says so above the message box. Replies still work from your files and what Lyremi knows, Composer Agent writes from your files, and Research Agent waits until searches start again.
9. Appearance (in General): System, Light, or Dark. Docs always keep a light page.
10. Your data: Download my data gives you everything you've made. You can delete your account at any time.