Luxe 0.1.3: your sessions on your phone, joinable sessions, and nested AGENTS.md

- luxe remote puts the sessions on your computer in reach of your own phone or tablet over
  Tailscale: write, steer, answer prompts, send photos, browse files, diffs and GitHub, with a
  paired device and your own Tailscale login.
- luxe join lets another program on the same machine join a running session over the RPC protocol.
- An AGENTS.md in a subfolder reaches the model the first time it works there.
- Long sessions stay fast: no stall before each model call, lighter memory, cheaper drawing.
- A session is open in one Luxe at a time, web_search shares results across its phrasings, and a
  grep glob starting with ! excludes.

The CHANGELOG has the full list.
This commit is contained in:
Evan Whittaker
2026-10-20 02:40:13 -05:00
parent 99e30d6d0f
commit d2129a886f
423 changed files with 31666 additions and 1669 deletions

View File

@@ -1,10 +1,8 @@
#!/usr/bin/env python3
"""Check shipped prose against `docs/brand/voice.md`.
Taste that cannot be asserted drifts. This asserts the one part of the voice that a machine can
actually judge: the words the rule bans outright. Everything else in that document — evidence with
claims, naming the trade-off, sentence variance — is a human's call, and pretending otherwise would
produce a checker nobody trusts.
This checks the part of the voice a machine can judge: banned words and British spellings. The rest
of that document (plain words, labels over sentences, where wordplay belongs) is a person's call.
Two severities, because they are different promises:
@@ -13,7 +11,7 @@ Two severities, because they are different promises:
Reported, never fatal.
Code is not prose: fenced blocks and inline spans are skipped, so a banned word inside a command is
a command. `docs/brand/voice.md` is skipped entirely — it quotes every one of these on purpose.
a command.
python3 scripts/voice-check.py # everything shipped
python3 scripts/voice-check.py --review # include the judgement calls
@@ -120,8 +118,7 @@ BRITISH = {
"spelt": "spelled",
}
# Text we did not write, plus this file — which contains every banned spelling by definition, the
# same reason `docs/brand/voice.md` is skipped above. Lockfiles are resolver output and can name a
# Text we did not write, plus this file, which contains every banned spelling by definition. Lockfiles are resolver output and can name a
# real package `@img/color`; upstream license grants keep whatever spelling they shipped with. The
# GitHub fixtures are GitHub's own answers, saved verbatim, with strangers' titles in them.
SPELLING_SKIP = (
@@ -149,7 +146,6 @@ SPELLING_SKIP_EXT = (
# Prose that ships. The planning docs and the journal are development-only and write for a different
# reader; holding them to the product's voice would be a category error.
INCLUDE = ["README.md", "docs/guide/*.md", "web/src/**/*.astro", "web/src/**/*.svelte"]
SKIP = {"docs/brand/voice.md"}
def prose(text: str) -> str:
@@ -236,14 +232,7 @@ def main() -> int:
)
args = parser.parse_args()
paths = sorted(
{
p
for pattern in INCLUDE
for p in ROOT.glob(pattern)
if str(p.relative_to(ROOT)) not in SKIP
}
)
paths = sorted({p for pattern in INCLUDE for p in ROOT.glob(pattern)})
banned = [f for p in paths for f in findings(p, BANNED, CASE_SENSITIVE)]
print(f"voice check \u2014 {len(paths)} files")