When something is off, check the status first
Most issues come down to four things: ① engine status (the dot next to the logo), ② indexing progress, ③ OCR model status, and ④ the LLM connection. Find your symptom below and follow the checks in order.
Checklist
Symptom → check → fix
| Symptom | What to check | Fix |
|---|---|---|
| Very few results | Are folders registered in Settings → Folders? Is a scan or indexing still in progress? | Add the folders that hold your documents; run a file scan from Settings → Engine if needed. |
| Precise search empty | Is the file type enabled in Settings → Analysis? Is the file under the size limit? Has document indexing finished? | Enable the type or raise the size limit, then rescan. If indexing is still running, retry shortly. |
| Luna is disabled | In Settings → LLM, check the on/off switch and the provider, key, and model. | Configure a provider and run the connection test. If it fails, verify the key, URL, and network (proxy/firewall). |
| New files missing | Is the file inside a watched folder? Is it on a network path where change watching is limited? | Run a manual file scan from Settings → Engine. Split overly broad network roots into subfolders. |
| PDF content not found | For a scanned-image PDF, check OCR on/off, recognition languages, model load status, and indexing progress in Settings → Engine. | Enable OCR, select the document language, and save. If you added a model, refresh → select the language → save, then run the file scan again. |
| Custom OCR language missing | Is the ocr-models\<lang> folder lowercase, with both inference.onnx and inference.yml present? |
Correct the names and location, then refresh OCR models. Once listed, check the language and save before it will be used. |
| Indexing too slow | Is the performance preset set to Eco? Are unnecessary huge folders being watched? | Turn on Boost (lightning button), raise the preset, or let Idle Boost work overnight. |
| Window won't open | Is the tray icon present? Does the shortcut clash with another app? | Open from the tray icon; change the shortcut in Settings → General if it clashes. If no icon, relaunch Findex. |
| Recurring engine errors | Check the status message and engine test in Settings → Engine, and look for a port conflict. | Try engine restart → change the port → reset search data; if it persists, contact us with the logs below. |
Image-only PDFs
What to check when OCR does not run
- Confirm that the PDF is image-onlyIf text on the page cannot be selected, the document likely needs OCR.
- Enable OCR, choose languages, and saveIn Settings → Engine, select the document language. Korean uses the ko recognition model; Chinese and Japanese use the default official PP-OCRv5 recognition model.
- Check the run mode and queueIdle-only waits while the PC is in use. Switch to Balanced for continuous processing, and allow extra time for large scanned documents while watching indexing progress.
- Run the file scan againAfter saving, use File scan in the Engine tab so the document is picked up again.
Custom model folder: %LOCALAPPDATA%\Findex\ocr-models\<lang>\ must contain both inference.onnx and inference.yml. Put the shared det.onnx directly under ocr-models. See OCR settings for the full structure and language coverage.
FAQ
Frequently asked questions
Q. Are my documents uploaded anywhere?
No — indexing, search, and OCR happen on your PC. However, when you use Luna with a cloud LLM connected, questions and conversation, relevant document content, filenames and metadata, and tool results may be sent to that provider. When an MCP tool runs, selected files and tool input/output may also be sent. If you don't want that, use a local model (Ollama, LM Studio) or keep the LLM off.
Q. Can I use Findex without an LLM?
Yes. Quick and precise search work fully without one. An LLM is needed for Luna's answers and the document or external-tool workflows Luna runs.
Q. How much disk space does the index use?
It depends on your library — typically hundreds of MB up to a few GB. To reduce it, limit the analyzed file types or lower the max file size in Settings → Analysis.
Q. Does indexing restart from scratch after a relaunch?
No — progress is saved and resumes on the next launch.
Q. Which languages do the bundled OCR models read?
The default model recognizes Simplified and Traditional Chinese, English, and Japanese. The ko model recognizes Korean, English, and digits. Selecting only the languages you need keeps the loaded model set minimal.
Q. How do I start over completely?
Run Reset everything in Settings → Engine. All data is erased and the setup wizard starts again. This cannot be undone.
Support
Finding logs and contacting support
For anything unresolved, sending the logs speeds things up a lot. You can find them here:
-
Paste into the Explorer address bar:
%LOCALAPPDATA%\Findex\logsOpen Windows Explorer, paste the path, and press Enter. - Grab the recent filesPick the app (Tauri) and engine logs closest to when the problem occurred.
- Email sami_info@samitech.krInclude the symptom, when it happened, the logs, and ideally your Windows and Findex versions.
Keep reading: Getting Started · Search · Luna · Settings