What it can do
- Summarise a long PDF or Word document in a few lines, or section by section.
- Answer questions about a document: dates, names, amounts, what a clause says.
- Compare two versions of a document and list what changed.
- Pull out figures from an Excel sheet or the tables in a report.
- Read pictures and scans: the text in a screenshot, a photo of a receipt or a scanned letter, read on your PC by Windows’ own text recognition.
- Write something new from them, in agent mode: a summary as a Word document, a short PDF report or a PowerPoint outline. It asks before it saves anything.
1. Set up Laghu and a model
Download Laghu and install it; it is free, with no account. Then download a model from the library. Any model can read an attached document; Qwen2.5 7B or Llama 3.1 8B gives good summaries on most laptops. For agent mode, choose a model badged Agent, such as Qwen2.5 Coder 7B. How to run an LLM locally explains the first steps.
2. Attach a document and ask
Click the paperclip in the message box, or drag the file onto the window. Laghu reads its text on your PC, which takes a moment for a long PDF, and the file appears above the message box. Then ask, as you would a colleague: “Summarise this in five bullet points.” You can attach several files at once, for example two versions of a budget to compare.
You can attach files before choosing a model. Click the model name in the message box to see which models on your PC read everything you attached; a picture is seen only by a model that can see, and any other model gets the text read from it. Laghu says so under the files, with the model to switch to.
If a document is longer than the model can take in at once, Laghu uses the start of it and marks the file trimmed. For a long report, ask about one part at a time, or use agent mode.
3. A whole folder: agent mode
To work through many documents, or to have Laghu write new files, switch on Agent in the message box (it is in beta) and choose the folder that holds your documents. Laghu works only inside that folder. It can read the files there freely, and it asks before it writes, changes or deletes anything. Name the files and say what you want: “Read the PDFs here and make a one-page summary of each, as a Word document.” Every step it takes is listed under the answer, and recorded in an activity log you can check later.
Good things to ask
- “What are the payment terms in this contract?” (with contract.docx attached)
- “Compare these two budgets and list the biggest changes.” (with both spreadsheets attached)
- “Translate the introduction of this paper into Hindi.” (with paper.pdf attached)
- “Read the PDFs in this folder and make a one-page summary of each, as a Word document.” (agent mode)
- “Turn notes.docx into a ten-slide PowerPoint outline.” (agent mode)
Good to know
- Scans and photos: scanned PDFs and pictures are read with Windows’ own text recognition, in the languages Windows has it for (English nearly always). Clear, straight, well-lit pages read best; handwriting often doesn’t. Of a long scanned PDF, the first 30 pages are read.
- Very long documents: a document longer than the model can take in at once is trimmed, and the file says so. For the best answers, ask about one part at a time.
- Check what matters: AI models can be wrong, so check figures and quotes against the document before you rely on them.
- Speed: on a modest laptop, reading and summarising a long document takes a few minutes; more memory makes it faster.
Questions
Are my documents uploaded anywhere?
No. The documents and the AI model are both on your PC, and nothing is sent to us or to anyone else. Our privacy policy says exactly what goes online, and when.
Does it work without the internet?
Yes. Once Laghu and a model are downloaded, reading and summarising documents works offline.
Which file types can it read?
PDF, Word (.docx), PowerPoint (.pptx) and Excel (.xlsx), plain text files such as .txt, .md and .csv, and pictures: PNG, JPEG, HEIC, TIFF, WebP and more.
Can it read text in a photo or a screenshot?
Yes. Attach the picture and Laghu reads the text in it on your PC. A model that can see, such as Qwen2.5-VL or SmolVLM2, gets the picture as well, so you can also ask about what is in it.
Do I need agent mode?
No. To ask about a document, attach it to a message. Agent mode is for working through a folder of documents, or for writing new files.
Can it save its summary as a document?
Yes, in agent mode. It can write PDF, Word and PowerPoint files into your folder, and it asks before it saves each one.
Is it free?
Yes. Laghu is free, with no account and no subscription.