* fix(db): forward DatabaseOpts to tauri-plugin-turso
NativeDatabaseService.open ignored its opts parameter, dropping any
experimental feature flags (e.g. 'index_method' needed for FTS / vector
indexes) and any encryption config before they could reach the Tauri
plugin. Translate DatabaseOpts to the plugin's LoadOptions shape and
forward as the single argument Database.load accepts. Skip translation
when no relevant opts are set so the existing path-string call shape is
preserved for callers without experimental/encryption needs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(db): cover Turso vector primitives + add benchmark harness
Verifies the Turso functions Reedy retrieval depends on, with a
brute-force per-book kNN test that runs against every DatabaseService
backend (node, native, WASM):
SELECT vector_distance_cos(embedding, vector32(?)) AS d
FROM book_chunks
WHERE book_hash = ?
ORDER BY d ASC LIMIT k
This is the path Turso's own founder recommended in
tursodatabase/turso#3778 ("First, focus on efficient SIMD-accelerated
brute-force search") and what shipped at commit 1aba105df4f. Native
vector index modules don't exist in this engine: `libsql_vector_idx`,
`vector_top_k`, and `USING vector/hnsw/diskann/ivfflat` all parse-error
against @tursodatabase/database@0.6.0-pre.28 (libsql_vector_idx is a
libSQL/sqld fork feature; DiskANN was closed not-planned upstream in
#832). The test asserts cross-book isolation and nearest-first ordering
using only `WHERE book_hash = ?` and `ORDER BY` — no DDL, no identifier
interpolation, no index plumbing.
Also adds bench/ harness for manual perf checks:
pnpm bench [name] run benchmarks (refuses in CI)
pnpm bench --list list available benchmarks
pnpm bench --no-record skip results.jsonl append
pnpm bench --force override the CI guard
Uses Node 24's --experimental-strip-types so no tsx devDep is needed.
Appends one JSON line per run to bench/results.jsonl (gitignored, local
history; share by pasting tabular stdout into PRs/issues). Explicitly
NOT in CI — shared-tenant variance makes synthetic-benchmark regression
detection unreliable; production telemetry (reedy_metrics, plan §M1.9)
is the right tool for that.
First benchmark: vector-retrieval. Measured on M1 Pro:
400 chunks × 384 dim → 0.35 ms / query
400 chunks × 768 dim → 0.45 ms / query
2000 chunks × 768 dim → 2.23 ms / query
10000 chunks × 768 dim → 14.00 ms / query
400 chunks × 1536 dim → 0.70 ms / query
Per-chunk cost ~1.1 µs at 768 dim = ~1.4 ns/dim. NEON-class on
Apple Silicon, ~50× faster than scalar — confirms SIMD acceleration is
active in 0.6.0-pre.28. Per-query latency stays sub-ms at Reedy MVP
corpus sizes; the ceiling is ~10K chunks per book before phone-class
hardware notices.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(reader): custom dictionaries (StarDict + MDict)
Adds a pluggable dictionary provider system. Built-in Wiktionary +
Wikipedia (extracted from the legacy single-popup model into a tabbed
shell) plus user-importable StarDict (.ifo/.idx/.dict.dz/.syn) and MDict
(.mdx/.mdd) bundles.
Settings → Language → Dictionaries: import / enable / drag-reorder /
delete (delete-mode toggle mirrors CustomFonts). Drag uses @dnd-kit with
pointer/touch/keyboard sensors. Reader popup: tabbed UI, per-tab lookup
history, scroll-aware back button, last-active tab persists. Tabs grow
to natural width up to a cap, truncate with ellipsis when crowded; phantom
bold layer prevents layout shift on focus.
StarDict reader is self-contained (replaces unused foliate-js/dict.js),
with lazy random-access binary search on .idx + .syn (~420 KB Int32Array
of byte offsets vs ~10 MB of parsed JS objects), lazy DictZip chunk
decompression via fflate streaming Inflate (cmudict/eng-nld both
chunked), and an optional .idx.offsets sidecar generated at import to
skip the init scan. Cmudict 105K-entry init drops from ~10 MB heap and
2 MB IO to ~1.7 MB heap and ~500 KB IO.
MDict uses the readest/js-mdict fork (added as a submodule, consumed via
tsconfig paths so deps stay out of readest's pnpm-lock) which adds a
browser-friendly BlobScanner reading via blob.slice(...).arrayBuffer()
— slices are lazy when the Blob is Readest's NativeFile / RemoteFile.
encrypt=2 (key-info-only) MDX is fully supported via ripemd128-based
mdxDecrypt; encrypt=1 (record-block, needs user passcode) surfaces as
unsupported.
Wikipedia annotation tool removed (Wikipedia is now a tab inside the
unified popup); legacy WiktionaryPopup / WikipediaPopup deleted. Stale
annotationQuickAction === 'wikipedia' coerced to 'dictionary' on settings
load. iOS-friendly external links: skip target="_blank" on Tauri to
avoid the WebView's "open externally" path triggering the shell scope
error; the popup's container click handler routes through openUrl.
i18n: 939 strings translated across 31 locales (30 base keys + CLDR
plural forms for ar/he/sl/pl/ru/uk/ro/it/pt/fr/es).
Test fixtures bundled: cmudict (StarDict, 105K entries), eng-nld
(StarDict, smaller), and a Longman Phrasal Verbs MDX (encrypt=2).
3396 unit tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(stardict): resolve fflate from js-mdict source for vitest + Next
js-mdict is consumed as TypeScript source via tsconfig paths from
packages/js-mdict/src/. Its sources `import 'fflate'` directly, but
fflate is only installed under apps/readest-app/node_modules — so
vite's import-analysis (and Next/Turbopack's resolver) can't find
fflate when it walks up from the redirected js-mdict source location.
CI's fresh checkout exposes this; locally a leftover
packages/js-mdict/node_modules/fflate from the old workspace setup
masked it.
Pin fflate resolution to apps/readest-app/node_modules/fflate in:
- vitest.config.mts (Vite alias)
- next.config.mjs (webpack alias + Turbopack resolveAlias — Turbopack
rejects absolute paths so use a project-relative form)
- tsconfig.json (paths entry so tsgo / Biome see it)
Verified by deleting packages/js-mdict/node_modules locally and
re-running pnpm test (3396 pass), pnpm lint (clean), and both
pnpm build-web and a tauri-platform Next build (clean).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix: add missing TXT worker files for large import path
* fix(ios): decode import paths and stabilize large TXT conversion
* style(txt): run prettier on TXT conversion files
* fix(txt): restore chunk iterator API and stream cancel behavior
* fix(txt): avoid unused private segment iterator
* chore: clean up log for unit tests
---------
Co-authored-by: Huang Xin <chrox.huang@gmail.com>
Move Radix/shadcn-based UI primitives into a dedicated
`components/primitives` directory to better reflect
their low-level, composable nature.
Imports remain ergonomic via an alias to `components/ui`.
App-level components should continue to use PascalCase filenames.
Huang Xin