Skip to content

cli.md

Latest commit

 

History

History
83 lines (63 loc) · 2.53 KB

cli.md

File metadata and controls

83 lines (63 loc) · 2.53 KB
Raw

CLI reference

kutsu has four subcommands. Global flag --log-format text|json (KUTSU_LOG_FORMAT) applies to all of them.

All subcommands read the same layered config (see configuration.md); GEMINI_API_KEY is always required.

kutsu init

Write a documented default config.

kutsu init                              # -> ./kutsu.toml (refuses to clobber)
kutsu init --force                      # overwrite
kutsu init --path /etc/kutsu/kutsu.toml

The content is kutsu.example.toml, embedded at build time so the CLI output and the in-repo example never drift.

kutsu mcp

Run the MCP server. See mcp.md.

kutsu mcp
kutsu mcp --transport streamable-http --bind 0.0.0.0:8090 --auth-token "$TOKEN"
Flag Env Default
--transport KUTSU_MCP_TRANSPORT stdio
--bind KUTSU_MCP_BIND 127.0.0.1:8090
--auth-token KUTSU_MCP_TOKEN (none)

kutsu call

Place one outbound call from the CLI and print the transcript.

kutsu call +15551234567
kutsu call +15551234567 --scenario ./scenario.json

--scenario is an optional JSON file with the per-call layer; without it a minimal default (empty goal schema, no context) is used:

{
  "goal_schema": { "type": "object", "required": ["disposition"],
    "properties": { "disposition": { "type": "string",
      "description": "Call outcome" } } },
  "context": { "name": "Alex Carter" },
  "prompt_override": null
}

The persona comes from config, not the scenario file — see prompts.md.

kutsu live

Developer harness: run one conversation against Gemini Live from a scenario plus an audio file (no SIP). Useful for iterating on prompts and audio handling.

kutsu live --scenario ./scenario.json --audio-in ./in.wav \
  --audio-out ./out.wav --transcript ./t.jsonl --goal-out ./goal.json
Flag Meaning
--scenario Scenario JSON (goal_schema, context?, prompt_override?).
--audio-in Mono PCM16 WAV or raw .pcm at 16 kHz.
--audio-out Output WAV (model speech, 24 kHz).
--transcript Transcript JSONL output.
--goal-out Filled goal JSON output.
--model Override model: half | native.
--voice Override voice.
--tail Seconds to keep the session open after input ends (default 8).
--no-net-check Skip the network preflight (offline debugging).
--greet-after-silence-ms Override the greet delay (0 = reactive).