FND

Fast, free, keyboard-driven document search for macOS. Indexes PDF, DOCX, PPTX, MD and 60 different file types, with strong BM25 ranking, in-file navigation, an "Open with…" launcher, and a lazygit-style TUI.

Linux and Windows are early beta — barely tested. The code paths exist and the automated suite runs against all three OSes in CI, but fnd is developed and used daily on macOS only. Nobody has put real day-to-day use on the Linux build, and the Windows build has had essentially none at all. Expect rough edges, and please open an issue when you hit one. See Platform support for what that means in practice.

Status

On macOS, core features are complete and stable; the project is in a refinement period of finding, fixing, and polishing. On Linux and Windows, treat everything as unproven — see the note above.

Requirements

• macOS (Apple Silicon or Intel) — the supported, tested platform.
• Linux or Windows — early beta, largely untested. See Platform support.
• Nothing else to set up. Each install option below brings Python 3.13 with it.
• A modern terminal is recommended (see Terminal compatibility).

Install

Pick one of the options below. They install the same tool, so you only need one.

Option 1: Homebrew (macOS)

brew install ben-dev-au/tap/fnd

Homebrew is the standard macOS package manager. Install it once, then run the line above. Apple Silicon installs a prebuilt binary; Intel builds from source, which takes a few minutes.

Option 2: uv or pipx (macOS, or Linux / Windows in beta)

uv tool install fndr        # or:  pipx install fndr

To run it once without installing: uvx --from fndr fnd.

---

Either way, launch the app by typing fnd.

On PyPI the package is named fndr (fnd was taken); the command stays fnd.

Releases carry build provenance; see SECURITY.md to verify a download.

Features

• Multi-format indexing: PDF, DOCX, PPTX, Markdown, and plain text.
• Named collections: group sources (per-source roots, include/exclude globs, optional symlink-following) and search them individually or together.
• Strong ranking: BM25 with regime-aware fusion (strong-signal / fusion / cascade) for stable results across corpora of different sizes.
• Expressive query language: phrases, boolean, proximity, fuzzy, field qualifiers, wildcards, date filters, and markdown-frontmatter predicates (see Search how-to).
• lazygit-style TUI: live search as you type, syntax-highlighted preview, and in-file navigation that jumps to the matching PDF page, PPTX slide, or Markdown heading.
• "Open with…" launcher: open a hit in Preview, Skim, Obsidian, VS Code, PDF Expert, or your own configured app, with page/line/heading deep-links where the app supports them (see Open with…).
• Obsidian integration: vault auto-detection, frontmatter filters, and line-precise jumps via the Advanced URI plugin.
• Structured PDF extraction (opt-in): headings, lists, tables, and bold/italic, with a shared content-addressed extraction cache and auto-resume on interrupted reindexes.
• Local and private: no network, no telemetry. The index lives on your machine; state is hardened to 0o700.

Platform support

macOS is the supported platform. It is where fnd is developed, run daily, and where every feature has been exercised by hand against a real corpus.

Linux and Windows are early beta. Platform-specific behaviour is isolated behind four seams — where files live, how to open and reveal them, what the OS calls things, and how cloud-backed files behave — and each has a Linux and Windows implementation. The automated suite runs on macos-14, ubuntu-latest and windows-latest in CI. That establishes the code runs; it does not establish that the app is pleasant, correct, or even usable in daily work on those OSes, because nobody has yet done that. Windows in particular has had almost no hands-on use.

So read the table below as what is implemented, not as what is verified:

Capability	macOS	Linux (beta)	Windows (beta)
Maturity	tested in use	CI only	CI only
Search · indexing · TUI · preview	✓	✓	✓
Open in app / Open with…	✓	✓	✓
Reveal in file manager (R)	Finder	file-manager --select → folder	Explorer /select
PDF page-jump on o	Skim, Preview	Zathura, Okular	SumatraPDF
Structured PDF extra (docling)	✓	✓	✓
Created-date filter	✓ (birth time)	best-effort (statx)	✓ (creation time)
Cloud-only file handling	iCloud Drive	not detectable	OneDrive & co.
Finder tag filtering	✓	—	—

Where an OS has no equivalent for something, fnd degrades rather than erroring. Built-in PDF viewers are auto-detected — the "Open with…" picker lists only the ones actually installed. On Linux/Windows, point [app_defaults].pdf at your preferred viewer, or add any app with a small [apps.<id>] block (see docs/apps.md); the bundled Linux and Windows handlers are written from each app's documented CLI and have not been confirmed against a live install. Frontmatter tags (tags: in YAML) work on every OS; Finder tags are macOS-only. Install with uv/pipx on any OS; Homebrew is macOS-only.

If you run fnd on Linux or Windows, reports are genuinely valuable — including "it worked fine", which is the datapoint the table is currently missing.

Quick start

fnd index ~/Documents/papers      # ad-hoc index a folder into the default collection
fnd search "diffusion model"      # search from the terminal
fnd                               # launch the interactive TUI

For ongoing use, define collections (see Collections & sources) and reindex them with fnd collection reindex <name>.

Using the TUI

Run fnd with no arguments for the interactive interface. It has three panes: the query bar at the top, the results tree (hits grouped by file) on the left, and the preview pane on the right showing the matching passage with your search terms highlighted. Just start typing, and results update as you go, and the query language works exactly as it does from the CLI.

Moving around with the keyboard

Key	What it does
↑ / ↓	Move the cursor up/down through results (vim's k / j also work).
⌥↑ / ⌥↓	Skim: hold Option (Alt) and arrow to move through results without loading each preview — browse fast with no per-row mount/lag. The preview loads again on a normal ↑/↓ (the row you land on) or Enter (the exact row you skimmed to).
Enter	Load the highlighted result into the preview (handy right after an Option-skim).
→	Expand the focused file to its matching sections; press again to drill into the first.
←	Collapse the focused node, or back out to its parent (lazygit-style).
Ctrl→ / ⌥→	Expand all: expand the focused node and its whole subtree (results, collections and filters trees).
Ctrl← / ⌥←	Collapse children: fold away every descendant, keeping the node itself open.
Tab	Cycle focus between the query bar, the results tree, and the preview.
/	Jump back to the query bar to refine your search.
↑ / ↓	When the preview pane is focused, scroll the preview.

Option-skim on Apple Terminal: for ⌥↑ / ⌥↓ to reach fnd, enable Settings → Profiles → Keys → Left Option key → Esc+. iTerm2 and most modern terminals work without any change.

Expand/collapse-all — Ctrl or Option? These are bound to both Ctrl and Alt+arrow, because a single physical combo reaches the app under different names per terminal. On macOS, ⌃←/⌃→ are usually captured by Mission Control ("Move a space"), so use ⌥←/⌥→ — your terminal forwards it as whichever of the two the app has bound. On Windows/Linux, Ctrl+arrow works directly.

Opening and acting on a result

Key	What it does
o	Open the hit in its resolved app, jumping to the matching page / slide / line / heading.
O	Open with…: a picker of every app that handles this file type. Use ↑↓ then Enter, or press the letter shown next to an app; Esc cancels.
R	Reveal: show the file in your file manager (Finder on macOS, File Explorer on Windows) with it selected, without opening it — also on the last row of the O picker.
Space	Quick Look the file.
:	Open the Settings & Commands menu: every setting and action in one searchable, full-screen list.
?	Keybindings cheat sheet (press again to dismiss).
Ctrl+F	Toggle auto-fuzzy matching (persists to your config).
h	Toggle search-term highlighting in the preview.
w	Warm the focused file completely, so scrolling anywhere in it is instant. Asks first on a large file; press again on that file to stop.
q / Ctrl+C	Quit. Esc backs out of any overlay or nested screen.

Inside the Settings menu (:) navigate with ↑↓ (or j/k), press Enter to open / edit / toggle the focused row, / to filter rows by label, and Esc or ← to step back.

Terminal compatibility

A modern terminal is required for optimal formatting. For example:

• macOS: iTerm2, Ghostty, Kitty, WezTerm. Terminal.app works but its formatting varies by font (Menlo is a reasonable default); a modern terminal is strongly suggested.
• Linux: Kitty, WezTerm, Alacritty, GNOME Terminal, or Konsole should all be fine — expected to work, not yet confirmed.
• Windows: Windows Terminal (bundled with Windows 11) or WezTerm. The legacy conhost.exe console renders box-drawing and colours poorly. Untested; if the TUI looks wrong, the terminal is the first thing to change.

Command reference

Command	What it does
fnd	Launch the interactive TUI.
fnd <query>	Launch the TUI with <query> pre-filled.
fnd -c <collection> <query>	Launch the TUI scoped to a collection (-c all for every collection).
fnd tui [query]	Explicitly launch the TUI (optional seed query).
fnd search "<query>"	Terminal search. Flags: --limit, -c/--collection, --meta, --explain N, plus the filters below.
fnd index <root>	Ad-hoc index a single root into the default collection.
fnd collection list	List configured collections and their sources.
fnd collection add <name>	Add (or extend) a collection in the config TOML.
fnd collection reindex <name>	Index or re-index a configured collection (--rebuild to start fresh).
fnd config show	Print the effective merged config as JSON.
fnd config path	Print the path to the config TOML.
fnd config edit	Open the config TOML in $EDITOR (creates a template if missing).
fnd config validate	Validate the config TOML.
fnd extras list	List optional extras and their installed status.
fnd extras status	Show installed extras with disk usage.
fnd extras install <name>	Install an extra after a disk-impact disclosure prompt.
fnd extras uninstall <name>	Remove an extra (indexed chunks remain).
fnd cache status / info / prune / clear	Manage the PDF extraction cache.
fnd version	Print the fnd version.

Search filters

fnd search takes the same filters as the TUI's Filters pane:

Flag	Meaning
--tag <name>	Only files carrying this tag. Repeatable.
--not-tag <name>	Exclude files carrying this tag. Repeatable.
--tag-match all|any	Combine multiple --tags. Default all.
--created <window>	Created within today/yesterday/week/month/year.
--modified <window>	Modified within the same windows.
--kind <ext>	Restrict to pdf/docx/pptx/md/txt. Repeatable.

fnd search "risotto" --tag recipe --not-tag draft --created month

Tags come from Obsidian-style YAML frontmatter (tags:) and, on macOS, from Finder tags. Nested tags work as a hierarchy — --tag project also matches project/alpha. Tag matching is case-insensitive, and a leading # is optional. Which sources are read is set by defaults.tag_sources in the config TOML; disabling one takes effect immediately, with no re-index.

Created dates come from the filesystem's creation time — macOS birth time, Windows creation time, and statx-capable Linux filesystems (e.g. ext4). Files without one match only the default (unfiltered) window. Only the macOS path has been exercised against a real corpus.

Open with… apps

In the TUI, o opens a hit in its resolved app and O opens the Open with… picker. Built-in handlers ship per OS — Skim, Preview, PDF Expert (macOS), Zathura, Okular (Linux, beta), SumatraPDF (Windows, beta), plus cross-platform Obsidian, VS Code, System Default — and each is offered only where it's installed. The non-macOS handlers are written from each app's documented command line and have not been confirmed against a live install. Where the app and file type allow it, fnd jumps to the matching page, slide, line, or heading. Set a per-file-type default with [app_defaults], or a per-source app, in your config.

You can add your own apps with a small [apps.<id>] block in your config; see the catalogue and schema in docs/apps.md. User templates are passed to apps as argv lists (never a shell) or as percent-encoded URLs handed to open, so file paths can't inject commands.

Collections & sources

A collection is a named group of source folders you search together; each source is a folder plus the include/exclude globs that decide which files in it get indexed. Out of the box fnd searches all your collections; tick individual ones in the sidebar to narrow that, and the selection is remembered for next launch. -c <name> scopes a single launch, and -c all widens it back out again without touching what's remembered.

There are three ways to manage collections, and they're interchangeable, because the UI writes the same config file you can edit by hand.

From the TUI

Press : to open Settings, move to Collections, then:

• Add a collection: choose Add collection and fill the wizard: Name, a Source path (a folder; ~/… is fine), the file types to Include and patterns to Exclude, an optional markdown Frontmatter filter, and a Follow symlinks toggle. Press Ctrl+S to save and index right away (Esc cancels).
• Add a source to an existing collection: open the collection, then Sources → Add source, and set the path, includes/excludes, an optional per-source app, and (for Obsidian) the vault name. Ctrl+S saves and returns; Ctrl+A saves and adds another. Reindex the collection afterward.

From the command line

# Create a collection with one source (repeat --source for more folders)
fnd collection add papers --source ~/Documents/Research

# Narrow it with globs, or a markdown frontmatter filter
fnd collection add notes --source ~/Notes --include "**/*.md" --exclude "drafts/**"
fnd collection add notes --source ~/Vault --filter "NOT ('private' in tags)"

fnd collection list             # show what's configured
fnd collection reindex papers   # build/update the index (--rebuild to start fresh)

From the config file

Run fnd config edit to open the TOML in $EDITOR (the first run writes a commented starter template), then fnd config validate to check it. UI edits preserve your comments and formatting, so hand-editing and the Settings UI mix freely.

Configuration

The config lives in your platform's app-data directory (run fnd config path to see the exact location): ~/Library/Application Support/fnd/ (macOS), ~/.local/share/fnd/ (Linux), or %LOCALAPPDATA%\fnd\ (Windows). fnd also reads ~/.config/fnd/config.toml on any OS if you keep it there. fnd config show prints the effective merged config; fnd config validate checks it before you rely on it.

Each collection is one or more [[collections.<name>.sources]] tables. A minimal, annotated config:

[defaults]
collection    = "all"      # scope a fresh profile starts with: "all" or a name
result_limit  = 200        # max results per query
fuzzy_enabled = true       # auto-fuzzy in the cascade fallback (toggle with Ctrl+F)

# A collection named "papers" with two source folders.
[[collections.papers.sources]]
path     = "~/Documents/Research"
includes = ["**/*.pdf", "**/*.md"]        # omit to index all supported types
excludes = ["**/.git/**", "archive/**"]
follow_symlinks = false

[[collections.papers.sources]]
path               = "~/Notes"
includes           = ["**/*.md"]
frontmatter_filter = "Status == 'published' AND NOT ('private' in tags)"  # md only

# Default app per file type for the `o` shortcut. Built-in ids:
# system, obsidian, vscode (all OSes); skim, preview, pdf_expert (macOS);
# zathura, okular (Linux); sumatra (Windows).
[app_defaults]
pdf = "skim"
md  = "obsidian"

# Define your own app (ready-made blocks live in docs/apps.md).
[apps.marked]
display_name = "Marked 2"
handles      = ["md"]
argv         = ["open", "-a", "Marked 2", "{path}"]

The [defaults] table also controls preview behaviour and auto-resume; run fnd config edit to see every option documented inline. After changing collections or sources, run fnd collection reindex <name> (or Reindex from the Settings UI) to apply it.

Indexing

Structured PDF extraction (opt-in)

PDFs render as flat extracted text by default. The opt-in pdf-structure extra adds headings, lists, tables, bold/italic, and recovered image-rendered tables. It is installed via uv (see the uv docs for the one-line installer on your OS).

In the TUI: Settings → Indexing → Status / Install… shows current state, disk impact (~900 MB), and a tight disclosure before any download. Install runs in a modal with progress; Esc sends it to the background, c cancels (SIGTERM).

From the CLI:

fnd extras install pdf-structure   # ~900 MB total, with disclosure prompt
fnd extras list                    # show available + installed
fnd extras status                  # disk usage per installed extra
fnd extras uninstall pdf-structure # revert; indexed chunks remain in index

After installing, reindex from Settings → Collections → ‹name› → Reindex (or fnd collection reindex <name>). New PDFs added later are extracted structurally automatically.

Two packages: pymupdf4llm (which pulls pymupdf-layout, Polyform Noncommercial 1.0) and docling-slim[standard] (Apache-2.0). fnd redistributes neither — the install fetches them onto your machine, so Polyform's non-commercial restriction binds your use of pymupdf-layout directly. Check it before installing this extra in a commercial setting. ML weights (~400 MB) download on first use. Uninstall removes the packages; indexed structured chunks remain in the index until the next reindex.

Cost on first reindex

~30 s per PDF on M1 Max (pymupdf4llm; longer for pages routed through the docling fallback). A 200-book corpus is roughly a 2-hour one-time cost. Subsequent reindexes only re-process changed files.

Cache

Extracted chunks are content-addressed in fnd's cache directory — ~/Library/Caches/fnd/pdf-structure/ on macOS, the platform cache dir elsewhere (fnd cache info prints it). Shared across collections: the same file in two collections is extracted once.

In the TUI: Settings → Indexing → Cache size shows entries + disk; Cache maintenance… drills to Prune stale (recoverable) and Clear (destructive, confirms with ⚠ Cannot be undone).

From the CLI: fnd cache status / info / prune / clear.

Auto-resume on launch

A Ctrl+C, sleep, terminal close, or fnd quit during reindex leaves the cache and a state file in fnd's data directory at reindex/<collection>.state.toml.

Reopen the TUI and indexing auto-resumes silently in the background. Already-cached files return in milliseconds, so resume effectively starts where you left off.

Toggle off from Settings → Indexing → Auto-resume on launch, or set defaults.indexer_auto_resume = false in your config.

Search how-to

fnd's query bar accepts plain words, phrases, boolean expressions, fuzzy and proximity matches, wildcards and regex, field qualifiers, date filters, and markdown frontmatter filters. They compose freely.

The basics

You type	What it does
entropy	One term. Searches body, title, headings, and filename. Stemmed (entropy = entropies).
cross entropy loss	Several terms, ranked. Docs matching more terms rank higher; all-term docs reach the top.
cross AND entropy	Require both terms.
"cross entropy loss"	Exact phrase, in order. Also matches cross-entropy loss.
cross OR entropy	Either term.
entropy NOT regression	Has entropy, excludes regression.
+rust -python	+ require, - exclude (shorthand for AND / NOT).
(loss OR cost) AND function	Group with parentheses, to any depth.

Phrases

Quoting is the biggest precision win — quote any common phrase:

You type	Matches
man in the middle	The four words anywhere in a chunk. Noisy.
"man in the middle"	The four words together, in order. Also matches man-in-the-middle.

Proximity

Find terms near each other, in any order — {N} and NEAR/N are equivalent:

You type	Means
{5} cross entropy	The two terms within 5 tokens of each other.
cross NEAR/5 entropy	Same.
{20} man in the middle attack	All five words within ~one line of text.
{60} buffer overflow exploit	Within ~a few lines.
{500} race condition mitigations	Within ~one page.

• Scale: 5 ≈ very near, 20 ≈ a line, 60 ≈ a few lines, 500 ≈ a page.
• Order doesn't matter — quote ("cross entropy") when it does.
• {N} covers the words right after it, up to the first operator, (, or filter, so {10} buffer overflow kind:pdf slops only buffer overflow.
• Can't cross a chunk boundary; if terms are far apart, drop the {N}.
• NEAR/N takes exactly two words.

Fuzzy matching for typos and variants

Suffix ~1 or ~2 to allow that many edits per term — an adjacent transposition (ir ↔ ri) counts as one edit:

You type	Matches
mitochondira~1	mitochondria, mitochondrial, etc.
kubernates~2	kubernetes and near spellings.

Works on a single term or alongside others (powerhouse mitochondira~1). Use sparingly on short terms: cat~2 matches almost everything.

Field qualifiers

A field qualifier is a hard filter: it narrows the result set (it does not just boost), so you can combine it with search terms to constrain them.

You type	What it does
title:transformer	Only documents whose title contains transformer.
heading_path:"chapter 4"	Only sections under that heading path.
author:dijkstra	Only documents with that author metadata.
kind:pdf	Only a file type (pdf, docx, pptx, md, txt).
path_tokens:thesis	Only paths containing thesis.
title:(rust OR golang)	Group alternatives within one field.
has:author	Only documents that have a non-empty author field.

Combine with terms to constrain them — kind:pdf "diffusion model" finds the phrase in PDFs only.

Collections

fnd organises sources into named collections. The shorthand c: scopes a search to one or more:

You type	What it does
c:wine attack	Search the wine collection only.
c:notes,papers transformer	Search two collections.

Without c: the active collection (settings menu) is used.

Page, slide, and date filters

Numeric ranges use [low TO high]. Shorthand for one-sided comparisons:

You type	What it does
page:5	Exact page 5.
page:>20	Page 21 onward.
page:[10 TO 20]	Pages 10 to 20 inclusive.
slide:<5	First four slides.
mtime:today	Modified today.
mtime:week / mtime:month / mtime:year	Within the last 7 / 30 / 365 days.
mtime:>2024-01-01	Modified on or after 2024-01-01.
mtime:[2024-01-01 TO 2024-06-30]	Modified in that ISO range.

Wildcards and regex

You type	Matches
crypto*	Words starting with crypto.
gr?y	? = exactly one character: gray, grey.
/cryp.*/	A regular expression over indexed words.

* only works at the end of a word. Leading or infix wildcards (*tion, de*ce) match almost nothing — search strips word endings before matching. Use a trailing crypto* or a /regex/ instead.

You rarely need *: search already matches word variants (entropy finds entropies). Wildcards, fuzzy, regex, and phrases all work inside AND / OR / NOT / ().

Markdown frontmatter filter

Filter markdown notes by their YAML frontmatter with a […] predicate. The same expression is also a source's frontmatter_filter in the config. String values use single quotes; double quotes mark a field name with spaces ("Due Date"):

You type	What it does
mitm [Course == 'Security Foundations']	Notes where the Course field equals that value.
[Notes_Type == 'Lecture' OR Notes_Type == 'Tutorial']	Either value (there are no list literals — use OR).
entropy [Course == 'ML' AND Year >= 2024]	Compound predicate.
['urgent' in tags]	urgent is an element of the tags list.
[NOT ('private' in tags)]	Exclude a tag, keeping notes that have no tags:.
[Course ~~ 'Design *']	Glob a string value (not the body-search ~N fuzzy).
["Due Date" < 2026-01-01]	A field name with a space, double-quoted.

Operators: == != < <= > >= ~~ (glob, string fields), in / not in (list membership), AND, OR, NOT, parentheses. Values are single-quoted strings, numbers, ISO dates, or true/false/null. Only markdown is filtered; other kinds pass through.

A missing field fails every comparison, including negative ones. On a note with no tags:, ['x' not in tags] is false, so the note is dropped; [NOT ('x' in tags)] is true, so it is kept. To exclude a tag without also losing untagged notes, negate the whole test: NOT (… in …). The same rule makes [Status != 'draft'] skip notes that have no Status at all.

in needs a real YAML list. If some notes write tags: private as a bare string, add AND NOT (tags ~~ '*private*') to catch them too.

Composing: worked examples

"buffer overflow"                                  # exact phrase
{10} buffer overflow exploit kind:pdf              # three terms within 10 tokens, PDFs only
c:notes mitm [Course == 'Security Foundations']    # term + collection scope + frontmatter filter
title:"chapter 4" heading_path:proof               # constrain to one chapter's proofs
kind:pptx slide:>10 attention                      # later-half slides mentioning attention
mtime:month crypto*                                # recently-modified docs mentioning crypto-anything
crypto* AND wallet                                 # a wildcard required inside a boolean
(loss OR cost) AND function~1                      # grouping with a fuzzy term
"defence in depth" OR diverse                      # an exact phrase OR a loose term

A few common pitfalls

• Quoting a single word does nothing useful. "entropy" is the same as entropy. Quotes only help for multi-word phrases.
• OR and AND are case-sensitive. Lowercase or / and are treated as ordinary terms. Always uppercase boolean operators.
• Standalone stopwords are dropped. the man searches just man — common words (the, in, of, …) are removed from unquoted queries. To match a phrase that includes them, quote it: "man in the middle".
• Proximity is per-chunk. A phrase or {N} query can't span a chunk boundary. If the terms are paragraphs apart, drop to a loose multi-term query.
• * only works at the end of a word. Leading/infix wildcards (*tion, de*ce) match almost nothing — use crypto* or /regex/.

Contributing

Bug reports and focused PRs are welcome; see CONTRIBUTING.md for dev setup and the "Open with…" app-catalogue workflow.

Security

fnd is local-only (no network, no telemetry). For the threat model and private vulnerability reporting, see SECURITY.md.

Support

fnd is free and always will be. If it's earned a spot in your workflow and you feel like buying a broke student dev a coffee, the button's there. Much gratitude if you do, but I hope you find the tool useful either way.

License

GNU AGPL-3.0-or-later © 2026 Ben Davidson

Use it, read it, modify it, run it — privately, for any purpose, with no obligations at all. If you distribute it, or run a modified version as a network service, that version has to ship its source under this same licence. Put plainly: your use is either private, or it is open source.

fnd links PyMuPDF (AGPL-3.0-or-commercial), so any distributed combination already carried AGPL terms; this licence states what was true of fnd all along.

Releases up to and including fndr 0.0.5 were published under MIT and remain MIT.

Acknowledgments

Some design choices in fnd's search layer are adapted from sibling open-source projects:

• tobi/qmd (MIT): the strong-signal bypass (skip parallel sub-queries when the literal probe is already unambiguous), the score normalization s / (1 + s) that makes its thresholds (0.85 score, 0.15 gap) corpus-stable, and the intent: line in the multi-line query DSL.
• The Reciprocal Rank Fusion constant k = 60 and rank-position bonuses follow Cormack/Clarke/Buettcher (2009).