qork

A simple, beautiful CLI for asking LLMs questions from your terminal. Fast defaults, clean output, and optional conversational threading.

Highlights

• Backend: OpenAI Responses API
• Optional streaming output (--stream)
• Optional global thread mode (--thread) that stores a single previous_response_id
• Plaintext or pretty Rich output

Install

pip install qork

Prerequisites

• Environment: OPENAI_API_KEY must be set
• Models: you can pass -m/--model at call time; otherwise defaults apply (see below)

Quick start

• Default (Responses API, non-streaming):

qork "Say hello in one short sentence."

• Responses API with explicit model:

qork -m gpt-5-mini "Give me a five-word poem."

• Presets:

qork --profile nano "Explain this in one sentence."
qork --profile high "Think carefully and give the best answer."

• Streaming:

qork --stream "List 3 colors."

• Plaintext output (easier to copy/paste):

qork -pt "Plain output please."

• Debug info (tokens/cost where available):

qork -d "How many seconds in a day?"

Backends and defaults

• Responses API (only backend)
  • Non‑streaming by default; use --stream to enable streaming
  • Default model if not specified: QORK_MODEL or gpt-5-mini

Thread mode (simple and explicit)

qork can optionally reuse a single global previous_response_id to continue one thread across invocations.

• Enable with -t/--thread
• State file lives at: ~/.qork/history/session.id
• This is global and will clobber across shells (intentionally simple)
• To reset: delete ~/.qork/history/session.id

CLI flags

• -m, --model Set model name
• --profile Preset: nano|mini|large|high (high sets reasoning.effort=high)
• --stream/--no-stream Stream output tokens
• -t, --thread Continue a single global thread (stores previous_response_id)
• -pt, --plaintext Plain stdout (no rich panels/markdown)
• -d, --debug Show token usage/cost when available

Python API

Call from notebooks and scripts using the same behavior as the CLI.

from qork.ask import ask

# Responses API (non-streaming)
text = ask("One short sentence.", stream=False, plaintext=True, return_text=True)

# Responses API (streaming)
text = ask("Print three facts.", stream=True, plaintext=True, return_text=True)

Parameters you’ll likely use:

• prompt: str (required)
• model: Optional[str]
• stream: bool
• plaintext: bool (stdout formatting)
• debug: bool (token/cost info)
• previous_response_id: Optional[str] (continue a thread)
• return_text: bool (return text value in addition to printing)

Examples

• Continue a single global thread:

qork -t "Start a thread in one sentence."
qork -t "Continue in one sentence."

Tests (end‑to‑end)

These tests hit live APIs (no mocks). Set your key first.

export OPENAI_API_KEY=sk-...
pytest -q tests/test_e2e_cli.py::test_cli_responses_session_persistence
pytest -q tests/test_e2e_cli.py::test_cli_plaintext_non_stream
pytest -q tests/test_e2e_cli.py::test_cli_plaintext_stream
pytest -q tests/test_e2e_python_api.py

You can select a model for tests with QORK_E2E_MODEL or rely on defaults.

Troubleshooting

• “API key not set”: ensure OPENAI_API_KEY is exported in your shell
• No thread carry‑over: ensure you used -t/--thread; check ~/.qork/history/session.id (delete it to reset)
• Streaming: use --stream if you prefer incremental output

---

Designed for fast, accurate answers from the terminal with minimal ceremony.