← Databases

Redis

The driver is built on the redis crate (RESP3 when the server speaks it, with rustls for TLS). It opens Redis 6 and later, and Redis Stack for JSON keys. The editor takes Redis commands, one per line. The driver is part of the app; nothing is downloaded.

Connect

The default port is 6379. The path (or the Database field) is the database number, 0 when empty; anything else is an error. The user and password are the ACL's: a password alone signs in as the default user.

TLS is on for rediss:// URLs and for SSL modes from require up (redis:// means no TLS and prefer does not turn it on, because Redis takes TLS on a port of its own). require does not check the certificate; verify-ca and verify-full do. A CA file and a client certificate go in the TLS fields. The session opens one connection per database it uses, because SELECT on a shared connection would move every other user of it.

Privileges

The ACL rules of the login decide what the navigator and the commands can do. The navigator reads INFO and, for each key type, runs SCAN … TYPE; browsing a type runs BROWSE, never KEYS. There are no administration tabs.

What works

• Commands in the editor, quoted as redis-cli quotes them; # and -- lines are comments. Replies are rows, with nested replies as JSON in the document tree (F7).
• SELECT n moves the session to another database. MULTI queues commands until EXEC, which sends them as one transaction and shows each reply.
• The navigator lists the databases that hold keys, each with a table per key type: string, hash, list, set, zset, stream, and json with RedisJSON. Opening one runs BROWSE db0.hash, optionally with MATCH and LIMIT, reading big keys a chunk at a time.
• The grid edits rows: values, TTLs, key names, hash fields, members and scores; new rows add keys and items, deleted rows remove them. The preview shows the commands, which run as one MULTI/EXEC; a command the server refuses does not undo the others. An Undo script puts the values back. A key, field or member that already exists stops the save before anything runs.
• FLUSHDB and FLUSHALL ask first, like an UPDATE without WHERE; on production connections commands that change the server (CONFIG SET, SHUTDOWN) ask as DDL does. Read-only connections run only commands that read.
• Cancel stops a blocking command (BLPOP, XREAD BLOCK).
• Terminal mode runs redis-cli if it is on the PATH, with the password in its environment.
• All the commands and key types are described in FEATURES.md.

What does not

• Subscriptions (SUBSCRIBE and the like) and MONITOR are refused: they would take over the connection.
• No manual commit: MULTI/EXEC is the only transaction.
• No explain, no object DDL, no SQL tools (table editor, ER diagram, schema comparison, query builder, SQL import), no snapshots before changes.

Agents

run_query and sample_rows (BROWSE … LIMIT n) work; any command that is not a read is asked about first. The commands that show the server's credentials (ACL GETUSER, ACL LIST, and a CONFIG GET whose pattern matches a password setting) are refused to read-only agents. Refused on Redis: explain, get_er_diagram, propose_edits and query_across_databases. See Agents and the CLI.

Known limits

• Older than 6.0: it connects with a warning that some features may not work.
• A table's key count shows only when the database has up to 10,000 keys.
• Values that are not UTF-8 show quoted with escapes.

Tested on

The conformance suite runs on Redis Stack 7.4 (with RedisJSON): metadata, BROWSE on every type, commands and their errors, saving and undoing grid edits, cancel of blocking commands, ACL users and database selection, and a lost connection that reopens. Redis 6 is the lowest version the driver accepts but is not run by the suite.

See also: Connecting, Troubleshooting.