R
RadMah AIDOCS
Sign In
Interfaces · CLI

rady — the terminal-first RadMah AI client

A friendly, chatty, colored CLI that drives the full RadMah AI platform — every job, every chat session, every engine, every evidence bundle — without leaving your shell. Same binary across Free, Sovereign, and Enterprise tiers; only the --base-url changes.

One command to log in

Paste an API key once, server-verified, persisted in your OS credential store (macOS Keychain / Windows Credential Manager / Linux Secret Service). Zero plaintext on disk, zero env var leaks.

rady auth login — banner, endpoint info, API key paste prompt, verifying spinner, logged in confirmation, next-step hint
Shell
rady auth login
rady auth whoami
rady auth logout

One-shot data generation from a natural prompt

Describe what you want in English; rady fabricates a sealed contract seal, submits a mock job, waits for completion, downloads the artefact. Credits metered server-side — client-side budget ceiling via --timeout.

rady mock — prompt echo, seal created, job submitted, waiting spinner, bytes written, evidence-bundle hint
rady mock \
    --prompt "100 SaaS accounts with MRR 50-5000 and churn_risk low/medium/high" \
    --rows 100 --seed 42 \
    --output ./mock.csv

Chat with the AI assistant

Interactive REPL with prompt_toolkit history (↑/↓), slash commands (/jobs, /approve, /quit), and rich Markdown rendering of assistant output.

rady chat — session open, user asks for 100 SaaS customer records, assistant proposes a schema preview, /approve slash-command confirms the plan
rady chat --mode auto
# Inside the REPL:
#   /jobs     → list recent jobs
#   /approve  → approve the current plan
#   /quit     → exit (session stays open server-side)

Job lifecycle in the shell

List, inspect, wait for, cancel, fetch artefacts, and pull sealed evidence bundles — everything the dashboard shows, scriptable for CI.

rady jobs list — a color-coded table of recent jobs with id, kind, engine, status (running/succeeded), rows, and created timestamp
rady jobs list --limit 10
rady jobs get <id>
rady jobs wait <id>
rady jobs cancel <id> --yes
rady jobs artifacts <id>
rady jobs download <job-id> <artifact-id> -o out.csv
rady jobs evidence <id>

Why a separate CLI

One binary, every tier

Free, Sovereign, and Enterprise all use the same rady binary. The only difference is --base-url — Enterprise customers point at their own Docker stack (http://radmah.internal:8000).

Credentials never leak

API keys stored in the OS-native credential store via keyring. Never written to plaintext, never exported as an env var, never shown in error messages. API-key fingerprints are deterministic hash-6 — safe to log.

Engines stay server-side

The CLI is a thin wrapper around the SDK, which is a thin wrapper around REST. No engine code ever reaches the user's machine through this package — matches the Commerce.md §9 “SDK + VPLC Docker model” architecture.

Chatty by design

Progress spinners on every long call, colored outcome markers ( ! ), pretty tables for list output, and italic hints pointing at the next useful command after each action.

Access

SDK + CLI access is provided under agreement. Contact sales@radmah.ai to get set up.