I’m building Findrail, an open-source local search app, and I’d like feedback on a workflow for people whose reference files extend beyond a single vault.
The idea is to leave Markdown notes in their existing vault folder and search them alongside code or text PDFs in other selected folders. Findrail is a standalone Go app with a local browser UI. It does not require an Obsidian plugin or change the source documents.
The workflow:
- Start with
findrail demo to try three fictional documents with a separate temporary index.
- Stop the demo and run
findrail start "/path/to/vault".
- While it runs, register another folder with
findrail index "/path/to/reference-pdfs".
- Search a word you remember. Preview the indexed passage and copy the original location; a PDF match includes its page number.
The second command and the additional index command must use the same data directory if you override the default. On Windows, use findrail.exe and Windows folder paths.
Folder changes refresh the index while the app runs. Documents are not uploaded, and there are no runtime analytics. The index itself stores plaintext copies of extracted text and paths.
This is an early alpha. Search is literal keyword search; multiple query words must occur in the same document. Scanned PDFs need OCR, which is not included. Preview shows indexed text, and Copy location returns a file URI. It doesn’t open Obsidian at a block or launch a PDF viewer.
Project, downloads and demo: GitHub - Lephiziel/findrail: Local-first search for notes, code and text PDFs. Built in Go, with matching passages, PDF page numbers and live folder refresh. No document uploads. · GitHub
I’m looking for a few people with a Markdown vault and reference PDFs outside it to try two real searches. Does this workflow help with something you currently find awkward? I’d also like to hear where existing tools already work better for you.
Update: Findrail 0.1.0-alpha.4 is available.
Since the original post, I’ve been working on the steps after finding a passage: getting back to the file and keeping useful excerpts for later.
- Desktop launcher: launch Findrail from your OS menu and reopen the same running instance. It still uses a local browser interface.
- Open original / containing folder: an opt-in action for eligible local files, enabled with
findrail start --enable-originals or findrail-desktop --enable-originals. Preview shows the indexed text; opening the original uses the current file. PDF viewers may open at their default page.
- Saved, named collections: select exact passages, add notes, arrange them and press Save. In normal
start or desktop sessions, collections can be loaded after restarting with the same data directory. Export them as Markdown citations for your notes.
For an Obsidian workflow, you can add your vault folder and a separate reference-PDF folder through Sources → Add folder, search across them and bring selected citations back into a note. Findrail leaves your source files untouched and needs no Obsidian plugin or AI account. It also supports DOCX body text.
To try fictional documents first, extract your platform archive and run ./findrail demo (Windows: .\findrail.exe demo). Search idempotency and preview the PDF match on page 2. Demo collections remain temporary; persistent collections are available in start / desktop sessions.
This is still an unsigned alpha with lexical search, not semantic search. Scanned PDFs need OCR, which is not included. If upgrading an existing index, stop Findrail and back up the whole data directory first; alpha.4 upgrades it to schema 7.
Downloads and installation instructions are in the GitHub repository linked in the first post.
If you keep reference PDFs outside your vault, I’d like to hear about one real search: did you find the passage you wanted, and was it useful to save or export it? Your OS and any setup step that got in the way would help too.