Metadata-Version: 2.4
Name: mystbird
Version: 0.0.1
Summary: Tooling for running AI models locally — starting with an environment probe.
Author: Du Fukang
License-Expression: Apache-2.0
Project-URL: Homepage, https://github.com/Mystbird
Project-URL: Issues, https://github.com/Mystbird/mystbird/issues
Keywords: local,llm,inference,hardware,probe
Classifier: Development Status :: 2 - Pre-Alpha
Classifier: Intended Audience :: Developers
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: System :: Hardware
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Dynamic: license-file

# mystbird

Tooling for running AI models locally.

Deciding whether a machine can run a given local model starts with a boring
question: how much memory, how much disk, and what kind of accelerator is
actually there. This package answers that question, and nothing else yet.

## Install

```bash
pip install mystbird
```

## Use

```bash
mystbird probe
```

```
Mystbird environment probe

  Python       3.13.1 (CPython)
  System       Darwin 25.5.0 / arm64
  CPU          16 logical cores
  Memory       128.0 GiB
  Disk free    1420.6 GiB at /Users/you
  Accelerator  apple-silicon — unified memory, Metal (MPS) expected
```

Add `--json` for machine-readable output, or `--path` to report free space for
a specific filesystem.

From Python:

```python
from mystbird import probe

info = probe()
if info["memory"]["total_gib"] and info["memory"]["total_gib"] < 16:
    print("This machine will need the smaller model tier.")
```

## Design notes

**No third-party dependencies.** A probe that cannot run on a bare interpreter
is useless as a probe, so this package deliberately avoids importing `torch`
or similar to detect accelerators.

**The accelerator field is a hint, not a capability check.** It reports what the
platform and `PATH` suggest — Apple Silicon implies Metal is expected, and a
discoverable `nvidia-smi` implies NVIDIA tooling is installed. Confirming that a
specific backend actually works is the caller's job.

**Unknown beats wrong.** Fields that cannot be determined on a given platform
come back as `None` rather than a guess.

## Status

`0.0.1` is an early release under the `mystbird` name. The wider Mystbird
project — a local expert runtime — is in development; this package will grow
alongside it. Issues and questions: https://github.com/Mystbird

## License

Apache-2.0
