fix(ci): disable AppImage bundling in Arch workflow

The previous approach tried to handle AppImage bundling failures, but
the failure occurs during the Tauri build process before the binary
compilation completes.

This fix uses --bundles none to skip all bundling (AppImage, deb, rpm)
and only build the binary. This is perfect for Arch since we create our
own .pkg.tar.zst package from the binary.

This avoids the strip incompatibility issue with modern Arch libraries
that have .relr.dyn sections.

Author: PythonTilk

.github/workflows/linux_packages_arch.yaml

@@ -155,33 +155,17 @@ jobs:
           # Build with Tauri
           # Note: When using pnpm -F desktop, the command runs in apps/desktop context
           # So the config path should be relative to that directory
-          # Note: AppImage build may fail due to strip incompatibility with modern Arch libraries,
-          # but we only need the binary for our Arch package, so we continue on error
+          # Note: We use --bundles none to skip AppImage bundling which fails on modern Arch
+          # due to incompatible strip utility. We only need the binary for our Arch package.
           TAURI_CONF_RELATIVE="./src-tauri/$(basename $TAURI_CONF)"
-          set +e  # Temporarily disable exit on error
           pnpm -F desktop tauri build \
             --target $RUST_TARGET \
             --config "$TAURI_CONF_RELATIVE" \
+            --bundles none \
             --verbose \
             -- --verbose
-          BUILD_EXIT_CODE=$?
-          set -e  # Re-enable exit on error
           
-          if [ $BUILD_EXIT_CODE -ne 0 ]; then
-            echo "⚠️  Tauri build completed with errors (exit code: $BUILD_EXIT_CODE)"
-            echo "This is expected - AppImage bundling fails due to strip incompatibility,"
-            echo "but the binary itself was built successfully and that's what we need."
-            
-            # Verify the binary was actually built
-            BINARY_CHECK=$(find /workspace/apps/desktop/src-tauri/target/${RUST_TARGET}/release -maxdepth 1 -type f -executable -name "hyprnote" | head -n 1)
-            if [ -z "$BINARY_CHECK" ]; then
-              echo "❌ Error: Binary was not built. This is a real failure."
-              exit 1
-            fi
-            echo "✅ Binary verified at: $BINARY_CHECK"
-          else
-            echo "✅ Tauri build completed successfully"
-          fi
+          echo "✅ Tauri build completed successfully"
           
           # Create PKGBUILD directory
           mkdir -p /pkgbuild

CLAUDE.md

@@ -0,0 +1,70 @@
+# AGENTS.md
+
+This file gives automated agents concise guidance for working in this repository.
+
+## Overview
+This is a polyglot monorepo:
+- Rust crates under `crates/` and `plugins/` plus some standalone dirs.
+- Tauri / desktop and web apps under `apps/`.
+- Shared TypeScript packages under `packages/`.
+- Scripts (Python, shell, Swift) under `scripts/`.
+
+## Build
+- Rust: `cargo build` for whole workspace; `cargo build -p <crate>` for a single crate.
+- Use feature flags to avoid heavy ML deps: e.g. `cargo test -p tauri-plugin-listener --no-default-features --features export-types` for specta export without local LLM/STT.
+- Prefer optional dependencies + a dedicated feature (e.g. `connector`) instead of unconditional linking.
+- Node/TS: `pnpm install`, then `pnpm run build` (Turbo orchestrates). For a single package: `pnpm --filter <name> run build`.
+
+## Testing
+- Rust: `cargo test` (workspace). Single crate: `cargo test -p <crate>`. With feature gating: `cargo test -p <crate> --no-default-features --features <feat>`.
+- Run a single Rust test: `cargo test -p <crate> <test_name>`.
+- TypeScript/Vitest: `pnpm --filter <pkg> test`. Single test name: `pnpm --filter <pkg> vitest run path/to/file.test.ts -t "test name"`.
+- Run snapshot tests: `cargo install cargo-insta` then `cargo test`.
+
+## Formatting & Lint
+- Rust formatting via `dprint fmt` (uses exec rustfmt). Run before committing.
+- Keep Rust imports grouped: std, external crates, workspace crates, local modules.
+- TypeScript formatting also via `dprint fmt`. Do not introduce other formatters unless necessary.
+- Avoid trailing whitespace; keep line length reasonable (<120 typical).
+
+## i18n / Translations
+- Uses Lingui for internationalization. After modifying UI text, run: `task i18n` (or `pnpm -F desktop lingui:extract`).
+- **CRITICAL**: Always run `pnpm -F desktop lingui:compile` after extract and before building to sync translation keys.
+- Build order for desktop: `poetry run python scripts/pre_build.py` → `pnpm -F desktop lingui:compile` → `pnpm -F ui build` → `pnpm -F desktop tauri build`.
+- Translation catalogs: `apps/desktop/src/locales/{en,ko}/messages.{po,ts}`. The `.ts` files are compiled from `.po`.
+
+## Conventions (Rust)
+- Modules & functions: `snake_case`; types & enums: `CamelCase`.
+- Errors: use `thiserror`; prefer a central `Error` enum and `Result<T, crate::Error>`.
+- Instrument async/public functions with `#[tracing::instrument(skip(...))]` when adding tracing.
+- Feature gating: wrap variants/APIs with `#[cfg(feature = "feat")]`; supply fallbacks when disabled.
+
+## Adding Features
+- Introduce new optional deps with a matching feature name; add to `default` only if broadly needed.
+- For specta/TS type export steps, minimize dependency surface (avoid heavy ML crates) by disabling default features.
+- Extend builders (e.g. event/command registration) conditionally behind feature flags.
+
+## TypeScript/Apps
+- Prefer explicit types for public APIs. Use consistent naming: `camelCase` for variables/functions, `PascalCase` for types/components.
+- Centralized config and shared utilities live in `packages/utils` and `packages/ui`.
+- Desktop app: `pnpm -F desktop tauri:dev` for development. Uses Vite + React + Tauri v2.
+
+## Scripts
+- Python uses `poetry.lock` / `pyproject.toml`; prefer `poetry run python <script>` if dependencies matter.
+- Shell scripts in `scripts/` should be idempotent; do not add interactive prompts.
+
+## Performance / Heavy Deps
+- Whisper / Llama and STT/LLM crates are large; gate them and avoid including them in lightweight export/test commands.
+
+## Pull Requests
+- Keep diffs minimal and focused; explain "why" in commit/PR body.
+- Ensure `dprint fmt` and targeted tests pass before requesting review.
+
+## Code Review
+- Use `coderabbit --prompt-only` for AI-powered code analysis; let it run fully to completion.
+- **Always run `coderabbit --prompt-only` at the end of every task** to ensure code quality.
+
+## Do Not
+- Do not commit large model binaries.
+- Do not add new formatters or global toolchains without discussion.
+- Do not enable heavy features for simple type export or CI smoke tests.