DBDelve: a native Rust database client with its own DBeaver benchmark
DBDelve is a native Rust database client for many engines and ODBC. Its DBeaver comparison covers one Mac and a deliberately oversized Postgres table.
Project facts
GitHub Ecosystem- License
- GPL-3.0
- Language
- Rust
- Stars
- 1,073
- Data checked
- 2026-11-12
Snapshot figures reflect the check date and may change over time.
DBDelve is a Rust database client for developers who want one native desktop app for relational and NoSQL connections. It is built with GPUI, without Electron or a JVM, and its README links a comparison with DBeaver. GitHub showed 1,073 stars on November 12, 2026. The benchmark is useful as a look at one workload on one machine; it should be read with those conditions in view.
Core features
- A broad engine list: The README names Postgres, MySQL, MariaDB, SQLite, Turso, DuckDB, Snowflake, SQL Server, MongoDB, Redis, Valkey, KeyDB, Dragonfly, and ClickHouse. Other databases can connect through the bundled ODBC bridge, though each still needs its vendor ODBC driver.
- SQL and result review: The editor includes schema-aware completion and query plans, with result rows shown in a grid. Editing a cell generates SQL. CSV, TSV, and JSON imports let users map columns and review statements before applying the data.
- Local files through DuckDB: CSV, Excel, Parquet, and JSON files can be opened as DuckDB data. The project says they are held in memory, so file size and available RAM matter.
- Connection organization: Connections can be grouped into projects and assigned access modes. The README describes importing saved connections from DBeaver and, on macOS, TablePlus. Diagnostic logs stay local, but verbose logging can contain hostnames, database names, and information emitted by a driver.
- Desktop packages: Builds are listed for Apple Silicon Macs, Linux x86_64 and aarch64, and Windows x86_64. Linux users can choose an AUR package, tarball, or AppImage; Windows has an installer and ZIP.
Typical use cases
- Backend developers who move among PostgreSQL, Redis, SQLite, and other systems and want one desktop connection list.
- Analysts who need to query a local CSV, Excel, or Parquet file with SQL without first importing it into a separate database.
- Teams using Snowflake, SQL Server, or an ODBC-only source. Check the driver and ODBC guide before making it part of a daily workflow.
- Developers evaluating DBeaver alternatives. Read the benchmark conditions, then repeat the queries that matter on your own connections. DbGate is another multi-database client in our directory, with a different feature range and application shape.
Quick start
On macOS, the README provides this install script:
curl -fsSL https://raw.githubusercontent.com/ShayanAbbas1/dbdelve/main/install.sh | sh
It targets Apple Silicon and macOS 12 or later. The project also has a Homebrew Cask, though the README notes that the app is not notarized. Linux users can choose AUR, tarball, or AppImage packages; Windows users can download an installer or ZIP from Releases. The installer is unsigned, so Windows SmartScreen warns on first run. After opening the app, create a connection with its address and credentials, then use the SQL editor. For an ODBC connection, install the matching vendor driver first; ODBC grid editing is conditional: it works only with the Generic SQL or PostgreSQL dialect, on one named table whose primary key the driver reports; Import Rows is off for every ODBC connection.
Benchmarks
The project’s BENCHMARKS.md reports DBDelve 0.3.6 against DBeaver 25.3.1 on October 11, 2026. The test machine was a 32 GB M1 Pro MacBook Pro. Its database was an emulated amd64 Postgres 15 instance with a one-million-row, 107-column table occupying 3 GB on disk.
| Measure | DBDelve 0.3.6 | DBeaver 25.3.1 |
|---|---|---|
| Cold launch to connected window | 1.13 s | 5.20 s |
| Idle memory | 133 MB | 606 MB |
| Query 1,000 rows | 0.21 s | 1.96 s |
| Query 1,000,000 rows | 55.6 s | 207.5 s |
| Process memory at 1,000,000 rows | 4.9 GB | 17 GB |
Each query size had five alternating runs, with medians reported. A result counted as finished after three seconds without a screen change and CPU below 10%. Both apps had row limits disabled, and DBeaver’s heap was raised to 16 GB; the repository says its default 1 GB heap runs out of memory at about 43,000 rows for this table. The million-row test intentionally reads the entire table into memory, while DBeaver normally fetches 200 rows by default. The harness is not in the repository, and the comparison covers one machine, database, and table. Result panes differ in size, and DBeaver’s launch CPU includes JVM warm-up. These results describe that setup, not a general speed ratio for everyday database work.
Summary
DBDelve is interesting for its native Rust desktop app, broad engine list, and ability to query local CSV, Excel, Parquet, or JSON files through DuckDB. The benchmark adds evidence, but the million-row result is a stress case with row limits disabled and DBeaver given a 16 GB heap. Re-run your own workload before choosing a client. DBDelve uses GPL-3.0; code contributed before the switch stays under MIT, and the main language is Rust. Its macOS app is not notarized, ODBC needs a separate vendor driver, and ODBC grid editing is limited to Generic SQL or PostgreSQL on a named table with a key; consult the project’s docs for driver and TLS restrictions. This entry relies on the README, BENCHMARKS.md and docs; the tool was not installed or run here.