Explorer UI¶

DataPress ships an optional browser-based explorer that lets you browse registered datasets, inspect schemas, run structured API queries, and open a live DuckDB-WASM terminal — all without leaving the browser.

It is served at a configurable path (default /explore) directly by the DataPress process. No separate web server is needed.

Build¶

The explorer is opt-in at compile time:

cargo build --release -p datapress-duckdb --features docs,swagger,explorer

When the binary is built without the explorer feature but [explorer] enabled = true is set in the TOML, the server logs a warning at startup and skips the mount. The feature flag has no effect on the HTTP API.

Configuration¶

[explorer]
enabled = true        # default; set false to hide the UI at runtime
path    = "/explore"  # mount point

Key	Default	Notes
`enabled`	`true`	Master switch. Set `false` to suppress the UI even when the feature is compiled in.
`path`	`"/explore"`	Mount point. Must start with `/` and not end with `/`. Cannot collide with `/api`, `/health*`, `/version`, or other reserved mounts.

What it shows¶

Open http://localhost:8080/explore (or your configured path) in a browser.

Discovery tab¶

Lists every registered dataset with:

row count, column count, and backing file size
source kind (parquet / delta / lazy) and location
schema — column names and inferred types
equality index configuration (when enabled)

Datasets are sortable by name, row count, or column count.

API Query tab¶

An in-browser query builder with two modes:

Structured JSON — builds a POST /api/v1/datasets/<name>/query request (the same structured QueryRequest body used by the HTTP API). Supports column projection, predicates, grouping, aggregation, sorting, pagination, and result export as CSV, JSON, or Parquet (Parquet export uses a bundled DuckDB-WASM instance entirely in-browser — no server round-trip).

Raw SQL — sends a POST /api/v1/sql request. This tab is only active when [sql] enabled = true; otherwise it shows a warning banner. A "To JSON query" button translates the SQL to a structured QueryRequest and switches back to JSON mode — useful when the SQL endpoint is disabled but you want to draft a query interactively.

Both modes support an Arrow IPC response toggle for faster large results, custom request headers, and a timing readout showing time-to-first- byte, total transfer time, body size, and row count.

DuckDB terminal tab¶

An embedded DuckDB-WASM terminal that queries the Parquet export of each dataset directly in the browser. The terminal connects to the server's Parquet download endpoint for each dataset; no SQL is executed server-side and no credentials are transmitted to the WASM sandbox. The DuckDB-WASM bundle is self-hosted and served by DataPress itself (no CDN).

Security note¶

The explorer uses the same session / cookie context as the browser. If [auth] is enabled, the explorer's API requests inherit the browser's Authorization header from the page that opened the explorer — or the session cookie for same-origin requests. There is no separate explorer credential. Disable the explorer (enabled = false) or restrict network access if you do not want the UI reachable from untrusted networks.

From Python¶

from datap_rs.datapress import DataPress, DataPressConfig, DatasetConfig

config = DataPressConfig(
    backend="duckdb",
    port=8080,
    explorer_enabled=True,
    explorer_path="/explore",
)