memex search

What it does

Words, or filters like to:worker-2, tag:namer and since:2d; any frontmatter field works as a filter. Results come grouped, READMEs and headers first, then collections, then logs, so a README is never drowned by log noise. Each hit is a one-liner.

A hit found by a near-miss spelling is marked [~variant]. Zero hits exit 0 with a hint on stderr. rg still finds symbols; memex finds meaning.

Put + before a word to keep only docs that hold it, or ! to leave them out: memex search widget +gizmo, memex search widget '!draft', memex search --in logs '!tag:show'. Both match the word or its stem, never a near miss, and never change the ranking. An id or path also finds the longer ones it begins, so ATM-NAME finds ATM-NAME-01. At an interactive bash or zsh prompt, single-quote a ! term; agents' shells pass it as is.

Flags

--in <names>

Searches only these collections or streams. With no words, prints the roster.

--limit N

Caps each group (or the one group, with --in).

--json

One JSON document, with scores and snippets.

Words, with a typo

$ memex search namer hashs --limit 1
readmes  (showing 1 of 2)
  crates/namer/README.md  Namer: turns atoms into stable class names.
headers  (showing 1 of 31; [~hashes] fuzzy match on hashes)
  crates/namer/src/hash.rs  FNV-1a over canonical atom strings.  [~hashes]
cases  (showing 1 of 6)
  cases:NAME-001  Class names are stable across runs and platforms.