Introducing the Prisment Local AI Assistant: Your Folder, Your Model, Your Machine

Meet Prisment's local AI assistant: ask your folder questions, make changes you can undo, and keep every document on your computer.

Introducing the Prisment Local AI Assistant: Your Folder, Your Model, Your Machine

Prisment's desktop assistant answers questions about your documents and makes changes you can review and undo, using an AI model that runs on your own computer — your documents stay on your machine. It is out of beta, with Qwen2.5-Coder 14B as the model we recommend, and important answers still need checking. Think of it as a helper that reads the folder you choose: ask "What's in this folder?" or "Summarize these documents", and Prisment finds the relevant text and hands it to your local model.

1. What can you do with it?

Ask your folder questions, have documents changed by asking, and see how your files connect — all with a model on your own machine. You ask in plain language; there are no commands and no @file syntax.

You want to What happens
Ask about your documents Prisment retrieves relevant passages, and the answer ends with the lines it came from
Summarize or review Each summary point names its section; a review starts with problems the app found itself
Change a document It edits, adds sections, renames with links repaired, or deletes — with a before-and-after view and an Undo button
Read a web page you mention Every new page asks you first: once, for this chat, or for this workspace

The assistant panel beside the Networking document: the question "What happens to a call that keeps failing?", the model's name and the line Context — 5 files, then an answer that cites docs/networking.md and says a call still failing after three attempts is parked on the dead-letter queue

It needs the Prisment desktop app for macOS, Windows or Linux, and Ollama or LM Studio with a chat model. Prisment never downloads a model for you.

2. What is it built to do?

Four goals: make a small, free model useful, keep your documents private by construction, keep you in control of every change, and be honest about where each answer came from. Where behaviour falls short, how we test it says so.

Four goals: small models made useful by fitting the right page into a 3B model's window; private by construction, with the model on your computer and a connection that cannot reach another; you stay in control, with every change a diff and one press from undone; honest about sources, with every answer naming the files it read

3. How does it find an answer?

Prisment finds the relevant parts of your documents, gives them to your local model, and shows which sources were supplied. The model can ask the app to look further before replying, and a source in the list means the model received it, not that every claim is correct.

How a question is answered, all on your computer: your question; 1 search over words, names in code, typos and any language; 2 the graph of files your best matches link to; 3 fit the whole file, passages or an outline to the model's window; 4 your model in Ollama or LM Studio, which can look further by listing, searching, reading and following links; 5 the answer with the files it read, or a change to review

Step 2 is why connections between files help: a troubleshooting note may point to the file that explains a feature even when the two share no words, so the answer can sit one link away. The Knowledge panel draws those connections as a map and shows what nothing links to. The steps in full are in the guide: How a question actually gets answered.

The Workspace Graph beside the Workspace Knowledge sidebar: seven documents as dots, docs/networking.md selected and joined by strong lines to observability.md, platform-overview.md and retry-storm.md, with a bar below reading referenced by 3, references 1

4. What does the app check in every answer?

A small model is good at wording and weak at bookkeeping, so the app does the bookkeeping: it checks each answer against the exact text it gave the model and says what it found under the answer.

The app checks What you see
Which lines each sentence came from Sources: docs/networking.md (lines 17–18) under the answer
A number or name no supplied line contains Not found in the supplied sources — and the sentence is never given a source
Whether a summary covered what you asked Each point ends with its section; a section nobody covered is named
Whether a review points at real problems Repeated or empty sections, unfinished TODOs, ignored exceptions in code — found by the app, with their lines
A to-do list Counted in TODO.md: 4 of 7 tasks done; still open: …

The model's own review suggestions are labelled as such: the app checks that their quotes are really in the file, not that their reasoning is right.

The assistant after the request "In docs/networking.md, change 'three times' to 'five times'": a to-do list ticked Done, the reply, and an Updated 1 file card for networking.md showing one line added and one removed with an Undo button, beside the document now saying five times

5. What should you expect from each model?

A small model is quick and usually finds the fact but may guess when your folder has no answer; a larger one is steadier on harder questions and noticeably slower, and the app checks both the same way. This is a description, not a score: on our own folders we are the judge, so we do not grade ourselves.

When you ask… Llama 3.2 3B — small, runs on a laptop Qwen2.5-Coder 14B — larger, needs more memory
For a fact in one document Usually right, with its lines Usually right, with its lines
About something spread over several documents Often joins them; can mix up which figure belongs where Joins them more reliably, but can still conclude two different values match
Something your folder does not say May guess — the app flags a figure it cannot find Usually says plainly that the folder does not say
How long it takes Seconds for most questions Noticeably slower; reviews can take a while

6. Where do you still need to check its work?

The assistant is a helper, not an authority on your documents, and it can sound confident when its answer is incomplete or wrong. A small model may guess instead of saying the folder has no answer; search can miss a passage in different words or another language, so ask in the document's own language; a request with several changes may be only partly done, so check the task list; and the model's own count can disagree with the one the app prints beneath it. Start with a focused question, check its sources, and use Undo freely.

7. What stays private?

Your documents and the model stay on your computer, and a web page is fetched only when you allow it — a request that carries the page's address and nothing from your files. The connection to the model is guarded in the app, not by a setting, and refuses any destination that is not your own machine. The Ask / Edit switch separates questions from changes. These promises concern your documents; the website uses analytics, which the privacy policy explains.

8. Try it

Install the desktop app, install Ollama with a model, open a folder and ask — about fifteen minutes, most of it the model download.

  1. Install the Prisment desktop app from the download page; the assistant is in the desktop app only.
  2. Install Ollama and pull a model: ollama pull qwen2.5-coder:14b (about 9 GB, recommended), or ollama pull llama3.2:3b (about 2 GB) with less memory, and the limits above.
  3. Open a folder in Prisment. Indexing reads it without writing index files into it; edits are a separate, requested action.
  4. Detect the model: in Settings → Local AI, press Detect.
  5. Ask a question in the assistant panel. The line above the answer says what the model read, and Context lists every file that went to it.

The guide covers picking a model for your memory and what a small model can and cannot be trusted to change. If the assistant gets something wrong in your folder, that is the report we most want.

Frequently asked questions

What is the Prisment local AI assistant?

An assistant in the Prisment desktop app that answers questions about the folder you opened and changes its documents when you ask. It uses a model running on your own computer through Ollama or LM Studio: we recommend the free Qwen2.5-Coder 14B, and the smaller Llama 3.2 3B runs on an ordinary laptop with limits we describe. Prisment finds the passages that answer, fits them into the amount the model can read, and shows which files it read.

Can a local AI model answer questions about my documents without uploading them?

Yes. In the Prisment desktop app the model runs on your computer, and Prisment talks to it through a connection that cannot reach another machine. Nothing is uploaded and no account is needed. The one exception is a web page you choose to let it read, and that request carries the page address and nothing from your files.

How accurate is a small local model with Prisment?

It can help find and explain information, but it can still miss details or misunderstand a document. The app checks each answer against the text it supplied: it names the lines an answer came from, flags a figure or name it could not find, anchors each summary bullet to its section, starts a review from problems it found itself, and counts a to-do list. The model can still word things wrongly, so check important answers against the named lines. A small model such as Llama 3.2 3B answers quickly but may guess where your folder has no answer; Qwen2.5-Coder 14B, the model we recommend, is steadier and noticeably slower. Ask in the language a document is written in. We do not grade the assistant on our own test folders, and we do not claim a leading benchmark rank.

Can the assistant change my files, and can I undo it?

Yes, if you ask. It can edit Markdown, text, YAML and SVG files, write new ones, add sections, rename files with their links repaired, and delete documents. Every change has a before-and-after view and an Undo button, a deletion always asks first, and source code is read but never written.

Which local models work with Prisment?

Any chat model you run in Ollama or LM Studio. We recommend qwen2.5-coder:14b, about 9 GB: it is the model the assistant was checked on before it left beta. llama3.2:3b, about 2 GB, runs on an ordinary laptop, with limits. The model you choose and how much text it can read affect both quality and memory use; no model is guaranteed to handle every request. Prisment never downloads a model for you.

Does the assistant work in the browser?

No. The assistant runs in the Prisment desktop app only, because only the desktop app can guard the connection to your model. The workspace index, its graph and context packs work in the browser too, with no model at all.