Using the eLLM chat

What this product is for

The chat helps you get reliable answers without hunting through files yourself. You ask a question in plain language; eLLM searches the documents you are allowed to see, finds the most relevant passages, and writes an answer that cites those passages. You can also use it for general questions that don't need any documents at all.

Main features

• Ask questions in natural language and get clear, written answers.
• See numbered citations that link back to the exact source document and page.
• Turn document grounding on or off with the Documents toggle.
• Choose which AI model answers, grouped into local and cloud models.
• Apply a skill to prime the assistant with a saved template or tone.
• Attach a document or image to a single question, or paste an image from your clipboard.
• Continue earlier conversations from a side panel, and start fresh with New Chat.
• Answers are nicely formatted and stream in live as they are written, including highlighted code blocks where relevant.
• Re-ask the same question with a different model, without retyping it.
• Compare two models side by side on the same question.
• Export any answer as a PDF, Word, or Excel file.
• Optional memory that quietly remembers lasting facts about you to personalise later answers.

How to use it

Asking a question

• Type your question in the input box at the bottom of the chat page.
• Press send.
• eLLM searches your accessible documents, then writes an answer above the input box.

Your answer may include numbered citation badges such as [Source 1]. Select a badge to open the citation panel and see the passage, document, and page the answer drew from.

Grounding answers in documents (the Documents toggle)

Next to the input you will find a Documents toggle. It is on by default and your choice is remembered in your browser.

• On: answers are grounded in your indexed documents and include citations. This is best for questions about company information.
• Off: ELLM answers as a plain assistant with no document sources. This is best for general questions, drafting, or brainstorming.

Choosing a model

If your administrator has registered more than one AI model, a model picker appears. Models are grouped into Local models (running on your own servers) and Cloud models (external providers). Pick the one you want, or leave the default. If only one model is available, the picker is hidden and that model is used automatically.

Applying a skill

A skill is a saved set of instructions that shapes how the assistant responds, for example a fixed format or a particular tone. If skills are available, a skill picker appears in the bar next to the model selector and Documents toggle. Choose a skill and it is applied to your questions until you change it. You can manage your own personal skills from the same bar. See the Skills article for details.

Attaching a document or image

• Use the attach control to add a document to a single question. It is added to the context for that message only and is not permanently saved to the library.
• You can attach an image, or simply paste an image from your clipboard into the input box. A small thumbnail appears, and the image is sent to the model so it can describe or reason about it (the chosen model must support images).

Continuing and managing conversations

• eLLM remembers the recent back-and-forth so follow-up questions make sense in context.
• When you reload the page, your current conversation continues in place.
• Past conversations are listed in a side panel. Select one to reopen it.
• Each past conversation can be deleted individually, and there is a Delete all button to clear them.
• Use New Chat to start a fresh conversation with a clean context.

Re-asking with a different model

Each answer has a Re-ask control. It lets you send the same question to a different model without retyping it, keeping your current Documents and skill choices. A new, tagged answer is added below, so you can see how another model responds.

Comparing two models side by side

The Compare button opens a side-by-side view: one question, two models you choose, both answers shown in parallel. Underneath, one of the models writes a short summary of how the two answers differ. This is useful when you want to weigh a fast local model against a more capable cloud one before trusting an answer.

Exporting an answer as a document

Each answer has an Export control to turn it into a downloadable PDF, Word, or Excel file, or choose Auto to let eLLM pick the best format. eLLM lays the content out properly with headings, lists, and tables. See the Exporting answers as documents article for details.

How memory personalises your answers

If memory is enabled, eLLM quietly remembers a few lasting facts about you, such as your team or how you like answers written, and uses them to tailor later replies. You never have to set this up, and you can review, remove, or switch it off in your settings. See the Personal memory article.

Common tasks

• Find a policy answer: keep the Documents toggle on, ask your question, and open the citations to verify the source.
• Draft an email or brainstorm: turn the Documents toggle off and ask freely.
• Compare answers from different models: use Re-ask to resend to another model, or Compare to see two side by side.
• Turn an answer into a document: use Export to download it as PDF, Word, or Excel.
• Reuse a format: apply a skill before asking.
• Start over: click New Chat to clear the current context.

Things to know

• Answers are only as good as the documents and models available. If a question isn't covered by any document you can access, eLLM may have little to ground its answer on.
• You only ever see citations from documents your group is allowed to access. ELLM never reveals content you are not permitted to see.
• Attached documents and pasted images apply to the single message you send them with, not to your whole conversation.
• The model picker only appears when more than one model is available to you. Some models may be restricted to certain groups.
• Answers render formatted text and highlighted code. Any raw HTML the model produces is shown as plain text rather than run, for safety.

Troubleshooting

• My answer has no citations: either the Documents toggle is off, or no document closely matched your question. Try rephrasing, or check the toggle is on.
• I don't see a model picker: only one model is available to your account, so it is used automatically.
• An image didn't work: the selected model may not support images. Choose a model that does, or describe the content in text.
• A follow-up question seems to ignore earlier context: if you started a New Chat, the previous context was cleared. Continue in the same conversation to keep context.
• I want a general answer but keep getting document-based ones: turn the Documents toggle off for plain chat.

Frequently asked questions

How do I see where an answer came from?

Select a numbered citation badge in the answer to open the source passage, document, and page.

What's the difference between local and cloud models?

Local models run on your own servers; cloud models are external services. See the Local models and Cloud models articles for guidance.

Are my past conversations saved?

Yes. They appear in the side panel and can be reopened or deleted. Use Delete all to remove them.

Can I attach a file just for one question?

Yes. Attaching a document or pasting an image affects only the message you send it with.

Does turning Documents off delete anything?

No. It only changes whether the current answer uses your documents. Your library is untouched.

Can I send the same question to another model without retyping it?

Yes. Use the Re-ask control on the answer, or Compare to run two models side by side.

Can I save an answer as a file?

Yes. Use Export to download it as a PDF, Word, or Excel document.

Does eLLM remember things about me?

Only if memory is enabled. It quietly keeps a few lasting facts to personalise your answers, and you can review or clear them in settings.

Summary

The chat page is the heart of eLLM. Type a question, decide whether to ground it in your documents, optionally pick a model or skill, and read a clear answer with citations you can check. Conversations are remembered and easy to revisit, and you can attach a document or image when a single question needs extra context.