config · check · models

See Providers & Auth Modes for the conceptual guide; this page is the command surface.

pdsa config

pdsa config <subcommand> <value>
pdsa config show          # current settings (key masked)

Subcommand	Purpose
key <key>	Set an OpenAI-compatible API key.
key-file <path>	Reference a key file (keeps the key out of config).
model <model>	Model id (default gpt-5.6-terra).
reasoning <level>	none | low | medium | high | xhigh | max.
base-url <URL>	Endpoint base URL.
provider <local|openai-compat [URL]|openai>	Preset an endpoint (may set auth=none).
auth <apikey|oauth|none|codex|claude-cli>	Authentication mode.
claude-cli-path <path>	Pin the claude executable.
claude-cli-timeout <sec>	Cap the claude -p round-trip (default 180s).
allow-insecure-no-auth <true|false>	Opt in to a remote no-auth endpoint.
oauth <endpoint|device-endpoint|client|refresh-token|refresh-token-file> <value>	OAuth settings.
login	Run the OAuth device-code login and persist the token.
lang <en|ko|auto>	Language for help + recorded coaching.
show	Print current settings (key masked).

pdsa check

Verify the configured LLM with a real, minimal round-trip.

pdsa check
# ✔ 성공 (…ms). / ✘ failure with the reason.

Works for every auth mode. On failure it prints the error and a hint to confirm the model exists on the endpoint.

pdsa models

List model ids the endpoint serves (OpenAI-compatible GET /models).

pdsa models [--filter <substring>]
pdsa models --filter gpt-5.6