Tabularis Goes NoSQL: Redis Ships, MongoDB and ClickHouse Need You
When I shipped the plugin system in v0.9.0, the promise was simple: you should not need to wait for us to support your database. Build a driver, list it in the registry, and Tabularis users can use it today. That promise is being kept — and the NoSQL world is where the community has moved fastest.
Redis: Two Independent Drivers, Both Excellent
Redis is the first NoSQL database to land in the official plugin registry, and it arrived not once but twice — two independent implementations from two different contributors, each with a distinct design philosophy.
Redis (Go)
The first Redis driver comes from @gzamboni and is written in Go. It models Redis as a set of virtual tables: Strings, Hashes, Lists, Sets, and ZSets each get their own tab in the Tabularis data grid. Write operations — SET, HSET, LPUSH, and the rest — are available directly from the UI without dropping to a terminal. If your team uses Redis as a primary data store and you want a visual interface for daily operations, this is the driver to reach for.
Tabularis driver for Redis databases. Supports virtual table views for Strings, Hashes, Lists, Sets, and ZSets with native UI write operations.
Redis (Rust)
The second implementation comes from @nicholas-papachriston and takes a different approach. Written in Rust, it focuses on Hash and RedisJSON row storage — treating Redis keys as structured records rather than raw key-value pairs. If you store JSON documents in Redis and want to query and browse them like rows in a table, this driver fits more naturally.
Redis key-value store with Hash and RedisJSON row storage
Two drivers for the same database is not redundancy — it is a sign that the plugin system is working as intended. Different use cases, different mental models, different implementations. Users can pick the one that matches how they use Redis.
The Next Frontier: DuckDB, MongoDB, and ClickHouse
Redis proved the model. Now the ecosystem needs to grow into the rest of the NoSQL and analytical space. Three databases are in active development, and all three could use experienced contributors.
DuckDB
DuckDB is the reference plugin — the first one we ever shipped — but being first does not mean it is finished. The current driver covers the basics: open a local .db file, run queries, browse results. What it does not yet handle well: in-memory databases, Parquet and Arrow file sources passed directly as connection targets, deeper type fidelity for DuckDB-native types like HUGEINT, MAP, and UNION, and ATTACH/DETACH multi-database workflows. If you know DuckDB internals and want to make the Tabularis integration best-in-class, the codebase is a Rust project with a clean JSON-RPC interface.
MongoDB
MongoDB is the most-requested NoSQL database on the issue tracker. A driver is in development but has not yet reached a releasable state. The core challenge is the impedance mismatch: MongoDB documents are schema-less and deeply nested, while Tabularis renders data in a tabular grid. Solving this well — flattening documents intelligently, handling arrays, surfacing the right projection controls in the UI — requires someone who has built against the MongoDB wire protocol before.
If you have experience with the MongoDB Rust driver, the Node.js driver, or the Go official driver, and you want to bring full MongoDB support to Tabularis, open a discussion on GitHub or reach out directly. The JSON-RPC protocol is documented in the Plugin Guide and a skeleton implementation is included to get you started quickly.
ClickHouse
The ClickHouse plugin landed in the registry at v0.1.0 — it connects, authenticates, and runs queries. That is a foundation, not a finished product. ClickHouse has a rich type system (tuples, maps, nested arrays, nullable variants), a native HTTP and binary protocol, and features like distributed tables, materialized views, and TTL policies that deserve first-class representation in the UI. The driver today exposes none of that.
If you work with ClickHouse at scale and know what a power user actually needs from a GUI client, the v0.1.0 driver is a starting point. Help shape what v1.0 looks like.
ClickHouse columnar OLAP database driver
How to Contribute
The plugin protocol is language-agnostic. MongoDB contributors have used Go, Rust, Python, and Node.js in existing plugins — pick the runtime you are most productive in. What matters is the JSON-RPC 2.0 interface over stdin/stdout.
Start here:
- Read the Plugin Guide — covers the full protocol, manifest format, and type system.
- Clone an existing plugin as a reference. The DuckDB plugin (Rust) and the CSV plugin (Python) are both intentionally simple and well-commented.
- Open an issue or discussion on the main repo before starting, so we can coordinate and avoid duplicate work.
The Redis contributions showed what happens when someone with domain expertise builds for their own use case: two production-quality drivers shipped in a matter of weeks. DuckDB, MongoDB, and ClickHouse deserve the same treatment. If you have the expertise, we have the platform.
