# The hub image build (deploy/hub/Dockerfile) only needs the Go module and its
# sources — the embedded webapp comes along under backend/. Everything below is
# either huge, machine-local, or a build output.
.git
dist/
src-tauri/
node_modules/
**/node_modules/
.venv/
.venv-*/
__pycache__/
*.pyc
.playwright-mcp/
.maylyn/
.maylyn-work/
.maylyn-live-test/
artifacts/
artifacts_preview/
boneyard/
models/
*.gguf
.env
.env.*
maylyn
/maylyn-test
*.log
.vscode/
.claude/
docs/
# website/ is NOT excluded — deploy/website/Dockerfile (a second image built
# from this same repo-root context) needs it. Its own node_modules/ is
# already covered by the generic **/node_modules/ rule above, which is the
# only actually-bulky part neither build needs.
# NOT canvasstyle/ or audio/ — hub mode never exercises that code at runtime,
# but Go still needs their source to compile the binary at all (serve_sound.go
# and canvas_image.go import them; the compiler resolves the whole dependency
# graph regardless of which subcommand actually runs). Excluding them broke
# every hub image build with "package maylyn/audio is not in std".
