Kai Realtime Voice

Real-time voice streaming via MiniMax WebSocket API. Use for low-latency voice conversations and streaming audio generation.

Security Scan

The skill largely matches its stated purpose (calling MiniMax t2a API with an API key) but contains a few incoherences (mentions WebSocket/Python websockets but implements only a curl REST fallback, and reads optional env vars not declared), so review before installing.

Purpose & Capability

Name/description (real‑time voice via MiniMax) lines up with the actual network calls (requests to https://api.minimax.io/v1/t2a_v2). The single required env var MINIMAX_API_KEY is expected for accessing the MiniMax API.

Instruction Scope

SKILL.md and README text advertise WebSocket-based real-time streaming and require 'Python3 with websockets', but the provided script uses curl and a REST fallback (no WebSocket implementation). The script also reads optional env vars KAI_MINIMAX_WORKSPACE and KAI_ENGLISH_VOICE_ID that are not listed in requires.env. These mismatches are scope/instruction inconsistencies that could confuse users and agents.

Install Mechanism

No install spec; this is instruction-only plus a shell script. No downloads or archive extraction. Low install risk (nothing is written outside the workspace when run, aside from created output file).

Credentials

Only MINIMAX_API_KEY is declared and required, which is appropriate. The script makes optional use of KAI_MINIMAX_WORKSPACE and KAI_ENGLISH_VOICE_ID (defaults provided) — these are not declared in metadata and should be documented. The skill does not request unrelated credentials.

Persistence & Privilege

always:false and no autonomously persistent behavior. The script writes output to a workspace path (default under $HOME/.openclaw) but does not modify other skills or global configuration.

What to consider before installing

This skill appears to do what it claims (call MiniMax t2a endpoint and save audio), but there are small inconsistencies you should understand before installing: (1) SKILL.md mentions WebSocket streaming and Python websockets, but the bundled script currently uses a REST curl fallback — so you will not get true WebSocket low-latency streaming until that is implemented; (2) the script reads optional env vars (KAI_MINIMAX_WORKSPACE, KAI_ENGLISH_VOICE_ID) that are not documented in requires.env — check or set these to avoid files being written where you don't expect; (3) it will send text and your MINIMAX_API_KEY to api.minimax.io, so verify that domain and consider using a scoped/test API key first. If you need real-time WebSocket streaming, request or inspect the implementation update (a Python/websocket client) before relying on low-latency behavior. If you accept the REST fallback and understand where output files are saved, the footprint is small and the install risk is low.

Real-time voice streaming using MiniMax WebSocket API.

Setup

API key already configured in OpenClaw config. Uses same key as kai-minimax-tts.

Usage

Test WebSocket connection:

bash {baseDir}/scripts/kai_realtime.sh --test

Stream text to audio:

bash {baseDir}/scripts/kai_realtime.sh --stream "Hello world"

Requirements

• Python3 with websockets library
• WebSocket support
• MiniMax API key (auto-injected)

Output

Audio streamed to stdout or saved to file.