feat(f35): local database browser (SQLite)

CHANGELOG.md

@@ -66,6 +66,27 @@ to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
   16,384-column limits are checked before writing (an over-limit export is
   refused up front), and the workbook is committed through the atomic-save
   pipeline, so a failed or cancelled export never touches an existing file.
+- **Local database browser** (F35): open a SQLite database file (palette →
+  "Open database…", or by opening a `.db` / `.sqlite` / `.sqlite3` file) to
+  browse its tables and views — columns with declared type, NOT NULL, defaults
+  and primary keys, plus indexes, foreign keys, WITHOUT ROWID flags, row-count
+  estimates and bounded preview rows. Open any table or view as an indexed,
+  read-only document (paged straight out of the database — a large table is
+  never copied into memory) or import it into a fully editable document
+  (guarded by a memory check, with an explicit "import anyway"). A changed
+  database is detected (rows via `PRAGMA data_version`, schema via a hash) and
+  offers a reload. Export the active document back to a database (palette →
+  "Export to database…"): create a new table, append to a compatible existing
+  one, or replace one after explicit confirmation. Before writing, a preview
+  shows the resolved document-column → SQL-type mapping (from the F31 declared
+  schema, TEXT by default; per-column name/type/primary-key overrides) and a
+  bounded scan of the cells that would fail to convert; the write runs in one
+  transaction that rolls back completely on any failure, with an explicit
+  primary-key conflict policy (abort / skip / replace). Every database read
+  goes through a read-only, authorizer-guarded connection restricted to the
+  files the user explicitly opened. _SQLite only this cycle — DuckDB support is
+  deliberately out of scope (its bundled C++ library cannot be built on the
+  low-memory MinGW development machine)._
 - **Row bookmarks, tags & notes** (F40): mark and annotate records without
   touching the source data. Star or flag a row, apply multiple named tags
   (a per-document tag namespace with usage counts), and attach a row note or

README.md

@@ -69,6 +69,16 @@ and faithful on large, real-world delimited files.**
   1900 leap-year quirk preserved) and leading-zero text stays text. Every
   import creates a new document — there is no in-place `.xlsx` save, so the
   original workbook is never modified.
+- **Open a SQLite database** — browse a `.db` / `.sqlite` file's tables and
+  views (columns with type / NOT NULL / defaults / primary keys, plus
+  indexes, foreign keys, row-count estimates, and bounded preview rows), then
+  open any table or view as a read-only indexed document (paged straight out
+  of the database, never copied into memory) or import it into a fully
+  editable document (guarded by a memory check). A database changed outside
+  CEESVEE is detected — rows and schema — and offers a reload, and every read
+  goes through a read-only, authorizer-guarded connection restricted to the
+  files you explicitly open. _Database support is SQLite only this cycle —
+  DuckDB is deliberately out of scope._
 - Auto-detect the **delimiter** (comma, tab, semicolon, pipe) with a manual /
   custom override — plus an **advanced import** for preambles, comment
   lines, custom quoting/escaping, multi-row headers, and footers.
@@ -140,6 +150,15 @@ and faithful on large, real-world delimited files.**
   back to text). Excel's 1,048,576-row × 16,384-column limits are checked
   before anything is written, and the workbook commits through the same atomic
   save pipeline, so a failed or cancelled export never touches an existing file.
+- **Export to a database** — write the active document into a SQLite table:
+  create a new table, append to a compatible existing one, or replace one
+  after explicit confirmation. A preview shows the resolved document-column →
+  SQL-type mapping (from the declared schema, TEXT by default, with per-column
+  name / type / primary-key overrides) and a bounded scan of the cells that
+  would fail to convert before anything is written; the write runs in one
+  transaction that rolls back completely on any failure, under an explicit
+  primary-key conflict policy (abort / skip / replace) and never altering an
+  existing table's schema.
 
 **Reliability**
 

src-tauri/Cargo.toml

@@ -61,6 +61,12 @@ parquet = { version = "59", default-features = false, features = ["arrow", "snap
 # no extra chrono feature is needed on the writer).
 calamine = { version = "0.36", features = ["dates"] }
 rust_xlsxwriter = "0.96"
+# Local database browser (F35) + SafeQueryEngine (F36): SQLite ONLY this
+# cycle — DuckDB is out of scope (the bundled libduckdb C++ build does not
+# fit the 6 GB dev machine). `vtab` exposes open documents as read-only
+# virtual tables; `hooks` provides the authorizer + progress handler the
+# SafeQueryEngine is built on.
+rusqlite = { version = "0.40", features = ["bundled", "vtab", "hooks"] }
 
 # Single-instance is desktop-only: it forwards a second "Open with" launch
 # (and its file argument) to the already-running window instead of spawning a