The Workbench is an interactive command-line interface built directly into the dashboard. Open it at http://localhost:8080 and select workbench in the sidebar to start running commands against your Cache-Pot instance from the browser. Every command executes in-process through the same pipeline as TCP client commands, so it appears in the Profiler and is recorded in the AOF.
Running commands
Type a command in the input box at the bottom of the page and press Enter or click run. The scrollback area above shows your command echoed back, the server’s response, and the time taken in microseconds.
Responses are rendered with full RESP type information:
| Response type | Display |
|---|
| Simple string | Plain text in green |
| Bulk string | Quoted text, e.g. "hello" |
| Integer | Labelled, e.g. (integer) 42 |
| Array | Numbered list, nested as needed |
| Error | Prefixed with (error) in red |
| Nil | (nil) or (empty array) |
Ctrl+L to clear the scrollback without losing your history.
Example session:
SET greeting "hello world"
GET greeting
HSET user:1 name Alice age 30
HGETALL user:1
Syntax hints
As you type, a hint strip beneath the input box shows the full command signature. Typing the start of a command name shows matching command names; completing a full command name shows the syntax and a short description.
For example, typing VSEARCH shows:
VSEARCH key f1 f2 ... fn [TOPK k] [WITHSCORES] — search for nearest vectors
Command history
The Workbench keeps a rolling history of up to 200 commands, saved in your browser’s localStorage so it persists across page reloads and browser sessions.
- Press ↑ to move back through previous commands.
- Press ↓ to move forward, returning to your current draft on the final press.
Use the Workbench to prototype commands before integrating them into your application code. The exact command syntax you confirm here will work identically over a TCP connection.
Running AI commands
The Workbench is the most convenient way to experiment with Cache-Pot’s AI and vector extensions. All AI commands follow standard RESP syntax and respond in the scrollback like any other command:
# Store vectors (3-dimensional)
VSET myembeds v1 0.1 0.9 0.2 META "first document"
VSET myembeds v2 0.8 0.1 0.7 META "second document"
# Find the nearest neighbours
VSEARCH myembeds 0.15 0.85 0.18 TOPK 2 WITHSCORES
# Semantic cache (requires embedding config)
SCACHE.SET "What is Go?" "Go is a statically typed language." 3600
SCACHE.GET "Tell me about the Go language"
# Agent memory
REMEMBER session-1 topic "vector databases"
RECALL session-1
VSET appear in the Key Browser as vector type, where you can inspect the vector count and dimension. Use the Workbench for all vector reads and writes.
Blocked commands
A small set of commands are not available in the Workbench because they require persistent connections or streaming behaviour that the browser CLI cannot support. Use the dedicated dashboard pages instead:
| Blocked command | Use instead |
|---|
MONITOR | Profiler page — real-time command stream with pause and filter |
SUBSCRIBE, UNSUBSCRIBE, PSUBSCRIBE, PUNSUBSCRIBE | Pub/Sub page — subscribe to channels and publish messages |
QUIT, RESET | Clients page — view and terminate connections |
Attempting to run a blocked command in the Workbench returns an error message directing you to the appropriate dashboard page. No connection state is affected.
