QwenPaw-TUI
A terminal chat UI for QwenPaw.
QwenPaw-TUI is a small, fast Textual front-end
for an existing QwenPaw installation. It spawns qwenpaw acp and drives it
over ACP (Agent Client Protocol) — streaming replies and thinking, rendering
tool calls as inline panels, handling permission prompts, and forwarding slash
commands (/clear, /compact, …) straight to the agent. Its command-line
tool is paw.
It only speaks ACP and never imports QwenPaw, so it stays light and ships independently. Everything else — provider keys, model selection, memory, tools — lives in QwenPaw; paw just drives the agent you've already configured.
Install
Install QwenPaw first (see the QwenPaw docs), then install the UI:
pip install qwenpaw-tui # finds `qwenpaw` on your PATH
paw does not bundle or install QwenPaw, so there's nothing to conflict
with your existing setup — it just drives whatever qwenpaw is on your PATH
(or an explicit --agent-cmd).
Usage
paw # interactive chat with your QwenPaw
paw --agent writer # pick a specific agent
paw --agent-cmd "qwenpaw acp" # drive an explicit ACP command
For non-interactive / one-shot use, run QwenPaw directly: qwenpaw chat.
Inside the chat: enter sends, shift+enter inserts a newline, esc
interrupts the current turn, and ctrl+c quits.
Type / to open an overlay suggestion list. It includes paw's own commands,
QwenPaw commands advertised over ACP, and argument completions such as
/theme cyberpunk. Use up/down to pick, tab or enter to insert a
highlighted suggestion, and esc to dismiss. Once an exact full command is
typed, enter submits it.
Useful local commands:
/themeopens the theme gallery;/theme <prompt>generates a persistent vibe from your words./inspecttoggles between friendly chat and deeper tool/thought inspection.
Model and provider commands (e.g. /model) are QwenPaw's — paw forwards them
to the agent. Configure providers and models with QwenPaw's own tools
(qwenpaw models config-key, qwenpaw models set-llm).
Pasting a file path or data URL attaches it to the prompt. Pasting long text stores it as a temporary attachment and inserts a reference, keeping the input usable.
How it finds QwenPaw
paw resolves the agent to drive in this order:
--agent-cmd "<command>"— used verbatim.- PATH — runs
qwenpaw acp.
If QwenPaw isn't on your PATH, install it first or pass --agent-cmd.
How it works
paw is an ACP client. It spawns the QwenPaw agent as a subprocess and
exchanges JSON-RPC over stdio. Because QwenPaw already ships a full ACP
agent (qwenpaw acp), paw reuses the backend for tools, memory, advertised
slash commands, permissions, and model switching.
The agent's stderr is drained to a log file under paw's state dir
(PAW_STATE_DIR, or an OS default) so chatty tools (e.g. a headless browser)
can't deadlock the stdio stream.
Develop
pip install -e ".[dev]"
pytest -m "not integration" # unit + transport + UI + CLI tests
Run the live development-checkout integration tests with provider keys:
DASHSCOPE_API_KEY=... \
OPENAI_API_KEY=... \
ANTHROPIC_API_KEY=... \
GEMINI_API_KEY=... \
DEEPSEEK_API_KEY=... \
pytest tests/test_qwenpaw_dev_integration.py -m integration -vv
The integration suite defaults to
/Users/erkang.zhu/code/ContinueLearningBench/submodules/QwenPaw, or set
QWENPAW_DEV_CHECKOUT=/path/to/QwenPaw. It starts QwenPaw through
uv run --project <checkout> python -m qwenpaw acp, verifies the imported
module comes from that checkout, and sets temporary QWENPAW_WORKING_DIR,
QWENPAW_SECRET_DIR, and PAW_STATE_DIR so it does not write to
~/.qwenpaw or ~/.copaw. It covers multi-turn conversation, a live
read_file tool call through ACP, and two-turn provider smoke tests for
DashScope, OpenAI, Anthropic, Gemini, and DeepSeek. Provider cases skip
cleanly if the corresponding API key is absent.
Default live-test models can be overridden with:
PAW_E2E_DASHSCOPE_MODELPAW_E2E_OPENAI_MODELPAW_E2E_ANTHROPIC_MODELPAW_E2E_GEMINI_MODELPAW_E2E_DEEPSEEK_MODEL
Against a local QwenPaw checkout
To test paw against an in-development QwenPaw (e.g. a sibling ../QwenPaw
editable install) without touching your normal QwenPaw setup, point paw at
that checkout's interpreter with --agent-cmd, and isolate its data with
QWENPAW_WORKING_DIR:
QWENPAW_WORKING_DIR="$PWD/.devdata" \
paw --agent-cmd "/path/to/QwenPaw/.venv/bin/python -m qwenpaw acp"
paw forwards its environment to the spawned agent, so any vars you set
(QWENPAW_WORKING_DIR, provider keys, etc.) reach QwenPaw. The agent uses
.devdata/.devdata.secret for its config, sessions, and secrets, leaving
~/.qwenpaw untouched.
First time, seed a provider key and pick a model into .devdata using
QwenPaw's own config (run the dev interpreter with the same working dir):
DEV="/path/to/QwenPaw/.venv/bin/python"
# store an API key for a provider (id: `dashscope` or `openai`) — paste the
# key at the prompt, e.g. $DASHSCOPE_API_KEY / $OPENAI_API_KEY
QWENPAW_WORKING_DIR="$PWD/.devdata" "$DEV" -m qwenpaw models config-key dashscope
# then choose the active model
QWENPAW_WORKING_DIR="$PWD/.devdata" "$DEV" -m qwenpaw models set-llm
The key is encrypted into .devdata.secret, so it stays out of your normal
install (and out of git).
License
MIT
Metadata
Release files for qwenpaw-tui 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| qwenpaw_tui-0.1.0.tar.gz | 61.6 kB | With provenance | Details for qwenpaw_tui-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| qwenpaw_tui-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for qwenpaw_tui-0.1.0-py3-none-any.whl |
Total release size: 113.4 kB
Release files / qwenpaw_tui-0.1.0.tar.gz
| Download URL | qwenpaw_tui-0.1.0.tar.gz |
|---|---|
| Size | 61.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
63eee4a8f198b38e42fbcc5930458de87978b75e45f718fd9b324dc58b47ea80
|
|
BLAKE2b-256 checksum How to use checksums |
a00e5478051d3c5855dd8fe846b54ba4d51a29b8fd87718ea419e9f7cd8320e1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 8, 2026.
Transparency logRelease files / qwenpaw_tui-0.1.0-py3-none-any.whl
| Download URL | qwenpaw_tui-0.1.0-py3-none-any.whl |
|---|---|
| Size | 51.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
50b07cc95d6e3a0ec879b28483eab1a81970ec38cdfbd8abdc3a27b6149f7e16
|
|
BLAKE2b-256 checksum How to use checksums |
ef2ca164ba76b653bb88eb29515c35c64fc82618b07d9db7418bcc45f01b8db4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 8, 2026.
Transparency logSecurity for qwenpaw-tui 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for qwenpaw_tui-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jun 8, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.