Padwan CLI

Interactive CLI/TUI for padwan-llm.

Quick start

export OPENAI_API_KEY=...
uvx padwan-cli
# then /help for commands

Install

pip install padwan-cli

Local Development

uv sync --group dev
uv run padwan-cli

One-shot Mode

uvx padwan-cli "Hello" -m gpt-4o-mini

# attach files (images/audio/text, routed by type)
uvx padwan-cli "What's in this recording?" -m gpt-4o-audio-preview -f clip.mp3

Voice Mode

Real-time speech-to-speech chat (needs the PortAudio system lib, e.g. libportaudio2):

uvx --from "padwan-cli[voice]" padwan-talk

Tracing

Pass --trace <backend> (on the one-shot, chat send, and talk commands) to instrument all LLM calls with padwan-llm's OTel GenAI telemetry:

# Langfuse, using the standard LANGFUSE_* env vars
export LANGFUSE_PUBLIC_KEY=... LANGFUSE_SECRET_KEY=... LANGFUSE_BASE_URL=...
uvx --from "padwan-cli[langfuse]" padwan-cli "Hello" --trace langfuse

# any OTLP collector, using the standard OTEL_EXPORTER_OTLP_* env vars
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318 \
  uvx --from "padwan-cli[otel]" padwan-cli "Hello" --trace otlp