Add/xgboost

.github/workflows/test.yml

@@ -32,10 +32,13 @@ jobs:
         uses: actions/setup-python@v4
         with:
           python-version: ${{ matrix.python-version }}
+      - name: Install OpenMP runtime for XGBoost (macOS)
+        if: runner.os == 'macOS'
+        run: brew install libomp
       - name: Installation
         run: |
           python -m pip install --upgrade pip
-          pip install .["streaming"]
+          pip install ".[streaming,xgboost]"
       - name: Test requirements Installation
         run: |
           python otherfiles/requirements-splitter.py

.gitignore

@@ -44,9 +44,13 @@ htmlcov/
 .cache
 nosetests.xml
 coverage.xml
+coverage_*.xml
 *,cover
 .hypothesis/
 
+# Generated by otherfiles/requirements-splitter.py from dev-requirements.txt (CI only)
+test-requirements.txt
+
 # Translations
 *.mo
 *.pot
@@ -82,9 +86,13 @@ celerybeat-schedule
 
 # virtualenv
 .venv/
+.venv*/
 venv/
 ENV/
 
+# Editor / IDE
+.vscode/
+
 # Spyder project settings
 .spyderproject
 

CHANGELOG.md

@@ -7,11 +7,42 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 ## [Unreleased]
 ### Added
 - `.codacy.yaml`
+- `XGBoostChain` class in `pymilo/chains/xgboost_chain.py`
+- `BoosterTransporter` class in `pymilo/transporters/booster_transporter.py`
+- `XGBOOST_SUPPORTED_TABLE`, `XGBOOST_SUPPORTED_CATEGORIES` and `CLI_XGBOOST_NOT_INSTALLED` in `pymilo/pymilo_param.py`
+- `get_xgboost_class` and `get_ml_model_class` functions in `pymilo/utils/util.py`
+- `get_xgboost_version` and `get_framework_versions` functions in `pymilo/pymilo_func.py`
+- `get_ml_model_data` function in `pymilo/pymilo_func.py`
+- `to_ml_model` function in `pymilo/pymilo_func.py`
+- `pymilo_xgb_booster_test` function in `pymilo/utils/test_pymilo.py`
+- `xgboost-requirements.txt`
+- XGBoost models test runner
+- `scenario5` XGBoost test for streaming
+- `PYMILO_CLIENT_DEFAULT_CLIENT_ID` and `PYMILO_CLIENT_DEFAULT_ML_MODEL_ID` in `pymilo/streaming/param.py`
+- `MANIFEST.in`
 ### Changed
 - `get_requires` function in `setup.py`
 - `meta.yaml`
 - `dev-requirements.txt`
 - `requirements.txt`
+- `Export` class in `pymilo/pymilo_obj.py`
+- `Import` class in `pymilo/pymilo_obj.py`
+- `print_supported_ml_models` function in `pymilo/pymilo_func.py`
+- `pymilo/__main__.py`
+- `.github/workflows/test.yml`
+- `scenario3` in `tests/test_ml_streaming/scenarios/scenario3.py`
+- `serialize_spline` function in `PreprocessingTransporter`
+- `deserialize_spline` function in `PreprocessingTransporter`
+- `SUPPORTED_MODELS.md` updated
+- `README.md` updated
+### Fixed
+- PyMilo CLI server launch
+- `generate_dockerfile` `--init` CMD
+- Source distribution packaging
+- `Export.save` directory creation
+### Removed
+- `get_sklearn_data` function in `pymilo/pymilo_func.py`
+- `to_sklearn_model` function in `pymilo/pymilo_func.py`
 ## [1.6] - 2026-04-03
 ### Added
 - `_check_response_error` function in `WebSocketClientCommunicator`

MANIFEST.in

@@ -0,0 +1,5 @@
+# setup.py reads these at build time, so they must ship inside the sdist.
+include requirements.txt
+include streaming-requirements.txt
+include xgboost-requirements.txt
+include CHANGELOG.md

README.md

@@ -192,6 +192,27 @@ The `Import` class facilitates importing of serialized models from JSON files, J
 
 This loaded model is exactly the same as the original trained model.
 
+### XGBoost
+PyMilo transparently exports and imports XGBoost models, just like scikit-learn ones.
+
+⚠️ In order to use XGBoost support, make sure you've installed the `xgboost` mode of PyMilo:
+
+```console
+pip install pymilo[xgboost]
+```
+
+Both the scikit-learn-style wrappers (`XGBRegressor`, `XGBClassifier`, `XGBRanker`, `XGBRFRegressor`, `XGBRFClassifier`) and the native `xgboost.Booster` are supported.
+
+```python
+from xgboost import XGBRegressor
+from pymilo import Export, Import
+
+model = XGBRegressor(n_estimators=20).fit(X_train, y_train)
+Export(model).save("xgb.json")
+
+restored = Import("xgb.json").to_model()
+```
+
 ### ML streaming
 You can easily serve your ML model from a remote server using `ML streaming` feature of PyMilo.
 
@@ -314,24 +335,7 @@ result = pymilo_client.predict(x_test)
 
 
 ## Supported ML models
-| scikit-learn | PyTorch | 
-| ---------------- | ---------------- | 
-| Linear Models ✅ | - | 
-| Neural Networks ✅ | -  | 
-| Trees ✅ | -  | 
-| Clustering ✅ | -  | 
-| Naïve Bayes ✅ | -  | 
-| Support Vector Machines (SVMs) ✅ | -  | 
-| Nearest Neighbors ✅ | -  |  
-| Ensemble Models ✅ | - | 
-| Pipeline Model ✅ | - |
-| Preprocessing Models ✅ | - |
-| Cross Decomposition Models ✅ | - |
-| Feature Extractor Models ✅ | - |
-| Composite Models ✅ | - |
-
-
-Details are available in [Supported Models](https://github.com/openscilab/pymilo/blob/main/SUPPORTED_MODELS.md).
+See [Supported Models](https://github.com/openscilab/pymilo/blob/main/SUPPORTED_MODELS.md).
 
 ## Issues & bug reports
 

SUPPORTED_MODELS.md

@@ -1,6 +1,6 @@
 # Supported Models
 
-**Last Update: 2025-12-30**
+**Last Update: 2026-09-07**
 
 
 <h2 id="scikit-learn">Scikit-Learn</h2> 
@@ -798,3 +798,47 @@
 		<td>>=1.5</td>
 	</tr>
 </table>
+
+<h2 id="xgboost">XGBoost</h2>
+
+Available when PyMilo is installed with the optional `xgboost` extra (`pip install pymilo[xgboost]`).
+
+📚 <a href="https://xgboost.readthedocs.io/" target="_blank"><b>XGBoost Documentation</b></a>
+
+<table>
+	<tr align="center">
+		<th>ID</th>
+		<th>Model Name</th>
+		<th>PyMilo Version</th>
+	</tr>
+	<tr align="center">
+		<td>1</td>
+		<td><b>XGBRegressor</b></td>
+		<td>>=1.7</td>
+	</tr>
+	<tr align="center">
+		<td>2</td>
+		<td><b>XGBClassifier</b></td>
+		<td>>=1.7</td>
+	</tr>
+	<tr align="center">
+		<td>3</td>
+		<td><b>XGBRanker</b></td>
+		<td>>=1.7</td>
+	</tr>
+	<tr align="center">
+		<td>4</td>
+		<td><b>XGBRFRegressor</b></td>
+		<td>>=1.7</td>
+	</tr>
+	<tr align="center">
+		<td>5</td>
+		<td><b>XGBRFClassifier</b></td>
+		<td>>=1.7</td>
+	</tr>
+	<tr align="center">
+		<td>6</td>
+		<td><b>Booster (native)</b></td>
+		<td>>=1.7</td>
+	</tr>
+</table>

dev-requirements.txt

@@ -7,6 +7,7 @@ requests==2.33.0; python_full_version >= "3.10"
 requests>=2.0.0; python_full_version < "3.10"
 websockets==15.0.1
 pydantic==2.13.2
+xgboost==3.4.1
 setuptools>=40.8.0
 vulture>=1.0
 bandit>=1.5.1

otherfiles/meta.yaml

@@ -26,6 +26,7 @@ requirements:
         - fastapi>=0.68.0
         - pydantic>=1.5.0
         - websockets>=9.0
+        - xgboost>=1.7.0
 
 about:
     home: https://github.com/openscilab/pymilo

pymilo/__main__.py

@@ -9,14 +9,17 @@
     CLI_MORE_INFO,
     CLI_UNKNOWN_MODEL,
     CLI_ML_STREAMING_NOT_INSTALLED,
+    CLI_XGBOOST_NOT_INSTALLED,
+    XGBOOST_SUPPORTED_TABLE,
 )
 from .pymilo_func import print_supported_ml_models, pymilo_help
 from .pymilo_obj import Import
-from .utils.util import get_sklearn_class
+from .utils.util import get_ml_model_class
 
 ml_streaming_support = True
 try:
     from .streaming import PymiloServer, Compression, CommunicationProtocol
+    from .streaming.param import PYMILO_CLIENT_DEFAULT_CLIENT_ID, PYMILO_CLIENT_DEFAULT_ML_MODEL_ID
 except BaseException:
     ml_streaming_support = False
 
@@ -95,9 +98,12 @@ def main():
         _model = _model.to_model()
     elif args.init:
         model_name = args.init
-        model_class = get_sklearn_class(model_name)
+        model_class = get_ml_model_class(model_name)
         if model_class is None:
-            print(f"{CLI_UNKNOWN_MODEL}\n{print_supported_ml_models()}")
+            if model_name in XGBOOST_SUPPORTED_TABLE:
+                print(f"{CLI_XGBOOST_NOT_INSTALLED}\n{CLI_MORE_INFO}")
+            else:
+                print(f"{CLI_UNKNOWN_MODEL}\n{print_supported_ml_models()}")
             return
         run_ps = True
         _model = model_class()
@@ -109,12 +115,16 @@ def main():
         pymilo_help()
         parser.print_help()
     else:
-        PymiloServer(
-            model=_model,
+        pymilo_server = PymiloServer(
             port=_port,
             compressor=_compressor,
             communication_protocol=_communication_protocol,
-        ).communicator.run()
+        )
+        if _model is not None:
+            pymilo_server.init_client(PYMILO_CLIENT_DEFAULT_CLIENT_ID)
+            pymilo_server.init_ml_model(PYMILO_CLIENT_DEFAULT_CLIENT_ID, PYMILO_CLIENT_DEFAULT_ML_MODEL_ID)
+            pymilo_server.set_ml_model(PYMILO_CLIENT_DEFAULT_CLIENT_ID, PYMILO_CLIENT_DEFAULT_ML_MODEL_ID, _model)
+        pymilo_server.communicator.run()
 
 
 if __name__ == '__main__':

pymilo/chains/util.py

@@ -11,6 +11,11 @@
 from .cross_decomposition_chain import cross_decomposition_chain
 from ..utils.util import get_sklearn_type, check_str_in_iterable
 
+try:
+    from .xgboost_chain import xgboost_chain as _xgboost_chain
+except BaseException:
+    _xgboost_chain = None
+
 
 MODEL_TYPE_TRANSPORTER = {
     "LINEAR_MODEL": linear_chain.transport,
@@ -23,6 +28,9 @@
     "CROSS_DECOMPOSITION": cross_decomposition_chain.transport,
 }
 
+if _xgboost_chain is not None:
+    MODEL_TYPE_TRANSPORTER["XGBOOST"] = _xgboost_chain.transport
+
 
 def get_concrete_transporter(model):
     """
@@ -37,6 +45,9 @@ def get_concrete_transporter(model):
         if upper_model in MODEL_TYPE_TRANSPORTER.keys():
             return upper_model, MODEL_TYPE_TRANSPORTER[upper_model]
 
+    if _xgboost_chain is not None and _xgboost_chain.is_supported(model):
+        return "XGBOOST", _xgboost_chain.transport
+
     if linear_chain.is_supported(model):
         return "LINEAR_MODEL", linear_chain.transport
     elif neural_network_chain.is_supported(model):