See the engine.
Follow the response.
Real Qwen3.5-9B-8bit execution, from a repaired Python function to a structured streamed response. 真实本地运行，可查看截图、回放和来源。
A practical code repair.
The local browser workspace repaired a Python clamp function and produced the same token sequence as direct engine execution.
```python
def clamp(x, lo, hi):
return max(lo, min(hi, x))
```Single recorded interactive request, 62 input tokens and 26 generated tokens including EOS; measured on M4 Pro / 24 GiB. Request timing describes this interaction.
Source JSON ↗
Watch a real text stream.
A structured extraction request from the recorded baseline API control. Playback follows captured content-event timestamps at your selected speed. All inference ran locally when recorded.
Press play to replay the captured response.
Recorded replay uses the real baseline content events and elapsed timestamps, including the recorded prefill wait. Playback speed changes presentation only.
Source request, SSE events and token controls ↗Your next inference runs here.
Apple Silicon · Python 3.12 / 3.13 · uv. Start the bundled local UI and API with the locked 9B model.
git clone --branch main --single-branch https://github.com/ZEVIN17/deepinfer-mac.git
cd deepinfer-mac
uv sync --frozen --extra mlx
uv run deepinfer-mac doctor
uv run deepinfer-mac models --download qwen3.5-9b-8bit --target-only
uv run deepinfer-mac serveDoctor reports Metal, Python, available memory and disk.
Visit http://127.0.0.1:8765 and select qwen3.5-9b-8bit. The default path is baseline generation.
Use CLI, streamed API or the bundled browser UI. Explore the explicit adaptive entry in the architecture guide.