0.4.2-beta

FileMind 0.4.2-beta is a small follow-up to 0.4.1-beta. It is free, fully
featured, and runs entirely on your machine — your PDFs are never uploaded
anywhere. If you already have 0.4.1-beta installed, it updates itself: the new
build downloads in the background and installs the next time you quit.

Windows will warn you before it runs this installer. That is expected.

This build is not code-signed yet, so Microsoft Defender SmartScreen shows a
blue full-screen warning the first time you run it:

Windows protected your PC
Microsoft Defender SmartScreen prevented an unrecognized app from starting.
App: FileMind-Setup-0.4.2-beta.exe — Publisher: Unknown publisher

The only button you can see is "Don't run". Click More info first, and a
Run anyway button appears underneath it. "Unknown publisher" means the file
carries no certificate — not that anything is wrong with it. Verify you have the
right file using the checksum below.

---

What changed since 0.4.1-beta

• Ask now works with Claude models that think before answering (Sonnet 5 and
  Opus 5). FileMind was reading past the answer to the model's thinking block,
  finding nothing usable, and retrying three times before giving up with a
  message that pointed at unrelated settings. If you had connected a Claude API
  key and Ask never returned anything, this was why.
• Send Feedback in the sidebar now opens the feedback form inside the app
  instead of launching a browser, and those reports carry the app version and
  platform the way the header button already did.

Everything else — rename, search, Ask, OCR, export, the uncapped trial — is as
in 0.4.0-beta; see that release's notes for the full description.

Install

• Windows 10 or later, 64-bit. 8 GB RAM minimum, 16 GB recommended.
• ~880 MB download, ~2.5 GB installed, plus 2–5 GB if you use a local AI model.
• The setup wizard offers a built-in model you can enable in one click. If
  you would rather use Ollama, LM Studio, or your own OpenAI / Claude / Gemini
  API key, it accepts those instead.

Known issues in this build

• Setup needs a network connection. Every AI provider option requires one and
  the AI step cannot be skipped, so first-run setup will not complete offline.
  Once set up, scanning and searching work fully offline.
• "Ask" is slow on the built-in model — expect several minutes for an answer
  on CPU. The progress counter is real, not a hang. Pointing FileMind at Ollama
  or a cloud API key is much faster.
• If no PDF reader is registered on your machine, "Open in system viewer" does
  nothing rather than telling you why.

Checksum

SHA-256 of FileMind-Setup-0.4.2-beta.exe: c3784dd4c588fa4797c5a4c555223fe181a6a607122be16c4e98e2647bb61faf

FileMind 0.4.1-beta — built-in model priced as free, feedback carries the version

FileMind 0.4.1-beta is a small follow-up to the first public beta. It is free,
fully featured, and runs entirely on your machine — your PDFs are never uploaded
anywhere. If you already have 0.4.0-beta installed, it updates itself: the new
build downloads in the background and installs the next time you quit.

Windows will warn you before it runs this installer. That is expected.

This build is not code-signed yet, so Microsoft Defender SmartScreen shows a
blue full-screen warning the first time you run it:

Windows protected your PC
Microsoft Defender SmartScreen prevented an unrecognized app from starting.
App: FileMind-Setup-0.4.1-beta.exe — Publisher: Unknown publisher

The only button you can see is "Don't run". Click More info first, and a
Run anyway button appears underneath it. "Unknown publisher" means the file
carries no certificate — not that anything is wrong with it. Verify you have the
right file using the checksum below.

---

What changed since 0.4.0-beta

• The built-in model is now correctly treated as free. Calls to the built-in
  model (and to LM Studio or any custom local server) were being recorded as if
  they went to OpenAI: Settings showed a "no known price" warning for them, and
  if you had set a monthly spend limit and reached it, the limit would have
  blocked the local model too. Local inference costs nothing and is never
  blocked now.
• The setup wizard's AI step no longer calls itself "(optional)". It has been
  required since the skip button was removed — Ask needs a model.
• In-app feedback now includes the app version and platform, so a report can be
  matched to the build it came from.

Everything else — rename, search, Ask, OCR, export, the uncapped trial — is as
in 0.4.0-beta; see that release's notes for the full description.

Install

• Windows 10 or later, 64-bit. 8 GB RAM minimum, 16 GB recommended.
• ~880 MB download, ~2.5 GB installed, plus 2–5 GB if you use a local AI model.
• The setup wizard offers a built-in model you can enable in one click. If
  you would rather use Ollama, LM Studio, or your own OpenAI / Claude / Gemini
  API key, it accepts those instead.

Known issues in this build

• Setup needs a network connection. Every AI provider option requires one and
  the AI step cannot be skipped, so first-run setup will not complete offline.
  Once set up, scanning and searching work fully offline.
• "Ask" is slow on the built-in model — expect several minutes for an answer
  on CPU. The progress counter is real, not a hang. Pointing FileMind at Ollama
  or a cloud API key is much faster.
• If no PDF reader is registered on your machine, "Open in system viewer" does
  nothing rather than telling you why.

Checksum

SHA-256 of FileMind-Setup-0.4.1-beta.exe: 66ee993d9bc559def9f4e3057b5119151d309bb9fb0177fc32f2f097d7574371

FileMind 0.4.0-beta — first public beta

FileMind 0.4.0-beta is the first public build. It is free, fully featured, and
runs entirely on your machine — your PDFs are never uploaded anywhere.

Windows will warn you before it runs this installer. That is expected.

This build is not code-signed yet, so Microsoft Defender SmartScreen shows a
blue full-screen warning the first time you run it:

Windows protected your PC
Microsoft Defender SmartScreen prevented an unrecognized app from starting.
App: FileMind-Setup-0.4.0-beta.exe — Publisher: Unknown publisher

The only button you can see is "Don't run". Click More info first, and a
Run anyway button appears underneath it. "Unknown publisher" means the file
carries no certificate — not that anything is wrong with it. Verify you have the
right file using the checksum below.

---

What you get

• Rename — proposes a clean, citable filename for every paper from its own
  metadata (DOI, arXiv id, title, authors, year), shown in a review queue you
  approve. Nothing is renamed without your say-so, and every applied batch can
  be rolled back — verified to restore each file byte-identical by SHA-256.
• Search — exact (full-text), semantic, and hybrid across your whole library.
• Ask — questions answered from your own papers, with citations back to the
  source page.
• OCR — scanned, image-only PDFs get read and become searchable.
• Export — BibTeX, RIS, CSV, and Zotero.

The embedding, reranking and OCR models are bundled in the installer, so the
first scan works with no network connection and no HuggingFace download.

The trial is uncapped during the beta

The shipping product caps the free trial at 100 PDFs. Beta builds ship with that
limit lifted deliberately — a 100-PDF cap sits below the size where FileMind's
value becomes visible, and we would rather learn what it is worth on your real
library than on a sample of it. Point it at everything.

Install

• Windows 10 or later, 64-bit. 8 GB RAM minimum, 16 GB recommended.
• ~880 MB download, ~2.5 GB installed, plus 2–5 GB if you use a local AI model.
• The setup wizard offers a built-in model you can enable in one click. If
  you would rather use Ollama, LM Studio, or your own OpenAI / Claude / Gemini
  API key, it accepts those instead.

Known issues in this build

• Setup needs a network connection. Every AI provider option requires one and
  the AI step cannot currently be skipped, so first-run setup will not complete
  offline. Once set up, scanning and searching work fully offline.
• "Ask" is slow on the built-in model — expect several minutes for an answer
  on CPU. The progress counter is real, not a hang. Pointing FileMind at Ollama
  or a cloud API key is much faster.
• If no PDF reader is registered on your machine, "Open in system viewer" does
  nothing rather than telling you why.

Verify your download

SHA-256  5f80601792c175b6e57a81f01efc631411a7349e36cbb0dc2e3264465e25380e
Size     921,489,664 bytes

Get-FileHash -Algorithm SHA256 .\FileMind-Setup-0.4.0-beta.exe

Tell us what breaks

This is a beta and we want the bug reports — https://filemind.app/feedback, or
open an issue here. Real libraries on real machines find things a test suite
never will.