Getting Started

Get FileMind up and running in under ten minutes. This guide covers installation, the first-run setup wizard, and your first library scan. It describes FileMind 0.4.1-beta.

System Requirements

The search and OCR models are bundled with the installer, so scanning, indexing and search all work with no internet connection. Choosing a language model is a required step in setup — Ask needs one, and rename proposals fall back to it. If you choose the built-in model it is downloaded at that point; the Ollama and LM Studio options use a model already on your machine, and the cloud options need only a key.

Installation

1. Download FileMind-Setup-0.4.1-beta.exe from the download page
2. Run the installer. The beta is not code-signed yet, so Windows shows a blue Windows protected your PC screen. Click More info, then Run anyway. The download page explains the prompt and publishes the SHA-256 checksum so you can verify the file first.
3. FileMind appears in your Start menu and launches when the installer finishes

FileMind currently ships for Windows only. A macOS build is planned — every license already covers both platforms, so there's nothing further to buy when it ships. Beta builds update themselves: when a newer beta is published, FileMind downloads it in the background and installs it on the next restart.

First Launch

On first launch FileMind opens the five-step Setup Wizard: a welcome screen, your library location and folders, a language model, the first scan, and a short tour. Choosing the built-in model is the quickest route — one click, no account, no separate install — but Ollama, LM Studio, and cloud API keys are all supported.

Your First Scan

The wizard runs the first scan for you as soon as you have chosen a model. To scan more folders later:

1. On the Dashboard, click Add Directory and choose a folder — FileMind scans recursively through subfolders
2. Click Step 1: Scan & Extract Metadata
3. Watch progress in Recent Jobs. You can keep using the app — the stat cards update as papers are indexed
4. Click Step 2: Generate Rename Proposals, then review them in the Rename Queue

What Happens During a Scan

For each PDF, FileMind:

1. Extracts text — native extraction for digital PDFs, in reading order, with running headers and footers removed
2. Scores text quality — pages with too little or too garbled text are routed to OCR
3. Runs OCR when needed — PaddleOCR is bundled and runs locally, so scanned papers become searchable too. Files are marked Clean text, Partial OCR, OCR only, or Failed
4. Detects identifiers — looks for a DOI or arXiv ID in the first pages
5. Looks up metadata — arXiv's API for arXiv papers, Crossref for DOIs and titles. Both are free and unmetered, which is why they run by default; OpenAlex is off by default because its anonymous daily budget runs out on a first scan
6. Parses heuristically — extracts title, authors, and year from the document layout when no lookup matches
7. Chunks and embeds the text — for exact, semantic, and hybrid search, and for the related-papers list on every paper

Rename proposals are a separate step so you decide when they are generated. That step composes a filename from the validated metadata and may fall back to the language model only when the deterministic methods fail. See Rename Workflow.

Next Steps

• Review and apply rename proposals
• Search your library or ask questions
• Switch or tune the AI model
• Customize thresholds and settings