diff --git a/.github/workflows/docs.yml b/.github/workflows/docs.yml
new file mode 100644
index 0000000..2e5050c
--- /dev/null
+++ b/.github/workflows/docs.yml
@@ -0,0 +1,68 @@
+name: Deploy Docs
+
+# Builds the mkdocs site and deploys it to GitHub Pages.
+#
+# One-time setup required before this workflow can deploy: in the repo
+# Settings -> Pages -> "Build and deployment" -> Source, switch it from
+# "Deploy from a branch" to "GitHub Actions". Until that switch is made,
+# this workflow will build the site successfully but the deploy job will
+# fail (Pages is still serving from the legacy `documentation` branch).
+on:
+  push:
+    branches: [main]
+    paths:
+      - "docs/**"
+      - "mkdocs.yml"
+      - "src/**"
+      - "requirements.txt"
+      - ".github/workflows/docs.yml"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: "pages"
+  cancel-in-progress: false
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: "pip"
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+          pip install mkdocs "mkdocstrings[python]"
+
+      # mkdocstrings imports the package to pull in its docstrings, so
+      # `python -m mkdocs` (not the bare `mkdocs` entry point) is used to
+      # make sure the repo root (and therefore `src/`) is on sys.path.
+      - name: Build site
+        run: python -m mkdocs build --site-dir _site
+
+      - name: Upload artifact
+        uses: actions/upload-pages-artifact@v3
+        with:
+          path: _site
+
+  deploy:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - name: Deploy to GitHub Pages
+        id: deployment
+        uses: actions/deploy-pages@v4
diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml
new file mode 100644
index 0000000..9d1ca8e
--- /dev/null
+++ b/.github/workflows/publish.yml
@@ -0,0 +1,57 @@
+name: Publish to PyPI
+
+# Triggered when a GitHub Release is published (create a Release from a
+# version tag, e.g. v1.3.0). Uses PyPI Trusted Publishing (OIDC) — no
+# API token needs to be stored as a secret.
+#
+# One-time setup required on PyPI before this workflow can publish:
+#   https://pypi.org/manage/project/teachable_machine/settings/publishing/
+#   Add a trusted publisher with:
+#     Owner:      MeqdadDev
+#     Repository: teachable-machine
+#     Workflow:   publish.yml
+#     Environment: pypi
+on:
+  release:
+    types: [published]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install build tooling
+        run: python -m pip install --upgrade build
+
+      - name: Build sdist and wheel
+        run: python -m build
+
+      - name: Upload build artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+      url: https://pypi.org/project/teachable_machine/
+    permissions:
+      id-token: write # required for PyPI Trusted Publishing
+    steps:
+      - name: Download build artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+      - name: Publish to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
new file mode 100644
index 0000000..ea424a4
--- /dev/null
+++ b/.github/workflows/tests.yml
@@ -0,0 +1,32 @@
+name: Tests
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: "pip"
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements.txt
+
+      - name: Run tests
+        run: pytest -v
diff --git a/CHANGELOG.md b/CHANGELOG.md
new file mode 100644
index 0000000..c687269
--- /dev/null
+++ b/CHANGELOG.md
@@ -0,0 +1,32 @@
+# Changelog
+
+All notable changes to this project are documented in this file.
+
+## [1.3.0]
+
+### Fixed
+
+- **TensorFlow/Keras 3 compatibility**: Teachable Machine's exported `.h5` models embed a legacy `DepthwiseConv2D` layer config (`'groups': 1`) that current Keras (Keras 3, bundled by default since TensorFlow 2.16) rejects during deserialization, causing model loading to fail with `TypeError`/`ValueError: Unrecognized keyword arguments passed to DepthwiseConv2D: {'groups': 1}`. `TeachableMachine` now loads models with a compatibility shim that discards the unused key, so exported models work on up-to-date TensorFlow/Keras installs without pinning an old TensorFlow version. Fixes [#2](https://github.com/MeqdadDev/teachable-machine/issues/2).
+
+### Changed
+
+- `install_requires` now specifies `tensorflow>=2.16` (previously unpinned, which could silently resolve to a broken combination).
+- `python_requires` raised to `>=3.9` to match current TensorFlow's supported Python range; removed unsupported 3.7/3.8 classifiers.
+- `requirements.txt` (dev/test) updated to current dependency versions and now includes `pytest`, `pytest-mock`, and `h5py`.
+
+### Added
+
+- Regression tests covering the legacy `DepthwiseConv2D` config shape, including an end-to-end test that saves a model, tampers its config to reproduce Teachable Machine's legacy export, and verifies it still loads and classifies correctly.
+- CI workflow to run the test suite on pushes/PRs, and a release workflow to publish to PyPI on tagged GitHub releases.
+
+## [1.2.1] - 2024-08-18
+
+- Added `setup.py` for packaging.
+
+## [1.2.0]
+
+- Added package logo, expanded documentation (how-to guide, requirements, code examples), and unit tests.
+
+## [1.1] and [1.0]
+
+- Initial releases of the `TeachableMachine` class for image classification with exported Keras models.
diff --git a/README.md b/README.md
index 293b46b..ff0b36b 100644
--- a/README.md
+++ b/README.md
@@ -23,9 +23,19 @@ Read more about the project (requirements, installation, examples and more) in t
 
 **Image Classification**: use exported keras model from Teachable Machine platform.
 
+## Compatibility with recent TensorFlow/Keras releases
+
+Teachable Machine's exported `.h5` models embed a legacy `DepthwiseConv2D` layer config that current Keras (Keras 3, bundled by default since TensorFlow 2.16) rejects with an error such as:
+
+```
+TypeError: Unrecognized keyword arguments passed to DepthwiseConv2D: {'groups': 1}
+```
+
+Since `v1.3.0`, this package patches the model loader so exported models load correctly on up-to-date TensorFlow/Keras installs, with no need to pin an old TensorFlow version. See [issue #2](https://github.com/MeqdadDev/teachable-machine/issues/2) for background.
+
 ## Requirements
 
-``` Python >= 3.7 ```
+``` Python >= 3.9 ```
 
 ## How to install package
 
diff --git a/docs/changelog.md b/docs/changelog.md
new file mode 100644
index 0000000..c687269
--- /dev/null
+++ b/docs/changelog.md
@@ -0,0 +1,32 @@
+# Changelog
+
+All notable changes to this project are documented in this file.
+
+## [1.3.0]
+
+### Fixed
+
+- **TensorFlow/Keras 3 compatibility**: Teachable Machine's exported `.h5` models embed a legacy `DepthwiseConv2D` layer config (`'groups': 1`) that current Keras (Keras 3, bundled by default since TensorFlow 2.16) rejects during deserialization, causing model loading to fail with `TypeError`/`ValueError: Unrecognized keyword arguments passed to DepthwiseConv2D: {'groups': 1}`. `TeachableMachine` now loads models with a compatibility shim that discards the unused key, so exported models work on up-to-date TensorFlow/Keras installs without pinning an old TensorFlow version. Fixes [#2](https://github.com/MeqdadDev/teachable-machine/issues/2).
+
+### Changed
+
+- `install_requires` now specifies `tensorflow>=2.16` (previously unpinned, which could silently resolve to a broken combination).
+- `python_requires` raised to `>=3.9` to match current TensorFlow's supported Python range; removed unsupported 3.7/3.8 classifiers.
+- `requirements.txt` (dev/test) updated to current dependency versions and now includes `pytest`, `pytest-mock`, and `h5py`.
+
+### Added
+
+- Regression tests covering the legacy `DepthwiseConv2D` config shape, including an end-to-end test that saves a model, tampers its config to reproduce Teachable Machine's legacy export, and verifies it still loads and classifies correctly.
+- CI workflow to run the test suite on pushes/PRs, and a release workflow to publish to PyPI on tagged GitHub releases.
+
+## [1.2.1] - 2024-08-18
+
+- Added `setup.py` for packaging.
+
+## [1.2.0]
+
+- Added package logo, expanded documentation (how-to guide, requirements, code examples), and unit tests.
+
+## [1.1] and [1.0]
+
+- Initial releases of the `TeachableMachine` class for image classification with exported Keras models.
diff --git a/docs/index.md b/docs/index.md
index a0cbe0d..7376c8a 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -29,6 +29,16 @@ Source Code is published on [GitHub](https://github.com/MeqdadDev/teachable-mach
 
 **Image Classification**
 
+## Compatibility with recent TensorFlow/Keras releases
+
+Teachable Machine's exported `.h5` models embed a legacy `DepthwiseConv2D` layer config that current Keras (Keras 3, bundled by default since TensorFlow 2.16) rejects with an error such as:
+
+```
+TypeError: Unrecognized keyword arguments passed to DepthwiseConv2D: {'groups': 1}
+```
+
+Since `v1.3.0`, this package patches the model loader so exported models load correctly on up-to-date TensorFlow/Keras installs, with no need to pin an old TensorFlow version. See [issue #2](https://github.com/MeqdadDev/teachable-machine/issues/2) for background.
+
 ## Links
 
 - [PyPI](https://pypi.org/project/teachable-machine/)
diff --git a/docs/requirements.md b/docs/requirements.md
index 15a58af..8ec0e90 100644
--- a/docs/requirements.md
+++ b/docs/requirements.md
@@ -1,12 +1,22 @@
 ## Requirements
 
 ### Python Version
-``` Python >= 3.7 ```
+``` Python >= 3.9 ```
 
 ### Dependencies
 
 ```bash
 numpy
 Pillow
-tensorflow
+tensorflow>=2.16
 ```
+
+## Compatibility with recent TensorFlow/Keras releases
+
+Teachable Machine's exported `.h5` models embed a legacy `DepthwiseConv2D` layer config that current Keras (Keras 3, bundled by default since TensorFlow 2.16) rejects with an error such as:
+
+```
+TypeError: Unrecognized keyword arguments passed to DepthwiseConv2D: {'groups': 1}
+```
+
+Since `v1.3.0`, this package patches the model loader so exported models load correctly on up-to-date TensorFlow/Keras installs, with no need to pin an old TensorFlow version. See [issue #2](https://github.com/MeqdadDev/teachable-machine/issues/2) for background.
diff --git a/mkdocs.yml b/mkdocs.yml
index 16b9a09..fd63dfd 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -12,4 +12,5 @@ nav:
   - Requirements: requirements.md
   - Code Examples: codeExamples.md
   - Explanation: explanation.md
+  - Changelog: changelog.md
   - Contributing: contribution.md
diff --git a/requirements.txt b/requirements.txt
index c28aedc..3ad4e18 100644
Binary files a/requirements.txt and b/requirements.txt differ
diff --git a/setup.py b/setup.py
index c1e8d40..509d80a 100644
--- a/setup.py
+++ b/setup.py
@@ -5,15 +5,13 @@
 
 setup(
     name="teachable_machine",
-    version="1.2.1",
+    version="1.3.0",
     description="A Python package designed to simplify the integration of exported models from Google's Teachable Machine platform into various environments. \
     This tool was specifically crafted to work seamlessly with Teachable Machine, making it easier to implement and use your trained models.",
     py_modules=["teachable_machine"],
     package_dir={"": "src"},
     classifiers=[
         "Programming Language :: Python :: 3",
-        "Programming Language :: Python :: 3.7",
-        "Programming Language :: Python :: 3.8",
         "Programming Language :: Python :: 3.9",
         "Programming Language :: Python :: 3.10",
         "Programming Language :: Python :: 3.11",
@@ -30,9 +28,9 @@
     install_requires=[
         "numpy",
         "Pillow",
-        "tensorflow",
+        "tensorflow>=2.16",
     ],
-    python_requires=">=3.7",
+    python_requires=">=3.9",
     url="https://github.com/MeqdadDev/teachable-machine",
     author="Meqdad Dev",
     author_email="meqdad.darweesh@gmail.com",
diff --git a/src/teachable_machine.py b/src/teachable_machine.py
index efa43bf..68734bf 100644
--- a/src/teachable_machine.py
+++ b/src/teachable_machine.py
@@ -1,8 +1,32 @@
 from tensorflow.keras.models import load_model
+from tensorflow.keras.layers import DepthwiseConv2D
 from PIL import Image, ImageOps, ImageDraw, ImageFont
 import numpy as np
 
 
+class _CompatDepthwiseConv2D(DepthwiseConv2D):
+    """
+    Drop-in replacement for Keras' DepthwiseConv2D that tolerates the
+    stray 'groups' argument found in models exported by Google's
+    Teachable Machine platform.
+
+    Teachable Machine's Keras/.h5 export embeds 'groups': 1 in every
+    saved DepthwiseConv2D layer config. That key was never meaningful
+    for a depthwise convolution, but older Keras ignored it. Keras 3
+    (bundled by default since TensorFlow 2.16) validates layer configs
+    strictly and raises:
+        TypeError/ValueError: Unrecognized keyword arguments passed to
+        DepthwiseConv2D: {'groups': 1}
+    This subclass discards the unused key on construction so exported
+    models keep loading on current TensorFlow/Keras installs, without
+    requiring users to pin an old TensorFlow version.
+    """
+
+    def __init__(self, *args, **kwargs):
+        kwargs.pop("groups", None)
+        super().__init__(*args, **kwargs)
+
+
 class TeachableMachine(object):
     """
     Create a TeachableMachine object to run pre-trained AI models.
@@ -30,7 +54,11 @@ def __init__(
 
     def _load_model(self, model_path: str):
         try:
-            self._model = load_model(model_path, compile=False)
+            self._model = load_model(
+                model_path,
+                compile=False,
+                custom_objects={"DepthwiseConv2D": _CompatDepthwiseConv2D},
+            )
         except IOError as e:
             print("LoadingModelError: Error while loading Teachable Machine model")
             raise IOError("Error loading model") from e
@@ -159,18 +187,30 @@ def show_prediction_on_image(
         draw = ImageDraw.Draw(image)
 
         font_size = int(image.height * 0.04)  # 4% of the image height
-        font = ImageFont.truetype("DejaVuSans-Bold.ttf", font_size)
-
-        text_width, text_height = draw.textsize(text, font=font)
+        try:
+            # Not bundled with Pillow on every platform (notably Windows),
+            # where a bare name can't be resolved and raises OSError.
+            font = ImageFont.truetype("DejaVuSans-Bold.ttf", font_size)
+        except OSError:
+            font = ImageFont.load_default(size=font_size)
+
+        # ImageDraw.textsize was removed in Pillow 10; use textbbox instead.
+        # A first measurement at the origin gives the text's extent, used only
+        # to decide where to place it.
+        _, top, _, bottom = draw.textbbox((0, 0), text, font=font)
+        text_height = bottom - top
         position = (10, image.height - text_height - 10)
 
+        # textbbox's (left, top, right, bottom) are offsets from the point
+        # passed in, and that offset (particularly `top`, from font ascent)
+        # is usually not zero -- reusing the origin-measured box as if it
+        # started exactly at `position` shifts it up and clips the text's
+        # descenders. Re-measure at the actual draw position instead, and
+        # pad it a little so the box comfortably covers the glyphs.
+        padding = 4
+        box = draw.textbbox(position, text, font=font)
         draw.rectangle(
-            [
-                position[0],
-                position[1],
-                position[0] + text_width,
-                position[1] + text_height,
-            ],
+            [box[0] - padding, box[1] - padding, box[2] + padding, box[3] + padding],
             fill=(0, 0, 0, 128),
         )
 
diff --git a/tests/test_teachable_machine.py b/tests/test_teachable_machine.py
index 8bbd585..0b1258d 100644
--- a/tests/test_teachable_machine.py
+++ b/tests/test_teachable_machine.py
@@ -1,11 +1,14 @@
+import json
+
 import pytest
-from src.teachable_machine import TeachableMachine
+from src.teachable_machine import TeachableMachine, _CompatDepthwiseConv2D
 from PIL import Image
 import numpy as np
 
 
 @pytest.fixture
 def teachable_machine(mocker):
+    """Return a TeachableMachine instance with load_model/open mocked out."""
     # Create a TeachableMachine instance
     mocker.patch("src.teachable_machine.load_model", return_value="mock_model")
 
@@ -99,6 +102,7 @@ def test_open_image(mocker):
 
 
 def test_preprocess_image_output_shape(teachable_machine):
+    """_preprocess_image should resize any input image to the model's fixed input shape."""
     # Create a sample image
     sample_image = Image.new("RGB", (300, 200))
 
@@ -110,6 +114,7 @@ def test_preprocess_image_output_shape(teachable_machine):
 
 
 def test_preprocess_image_normalization(teachable_machine):
+    """_preprocess_image should scale pixel values into the [-1, 1] range."""
     # Create a sample image with known values
     sample_array = np.full((300, 200, 3), 127, dtype=np.uint8)
     sample_image = Image.fromarray(sample_array)
@@ -123,6 +128,7 @@ def test_preprocess_image_normalization(teachable_machine):
 
 
 def test_preprocess_image_different_sizes(teachable_machine):
+    """_preprocess_image should normalize inputs of any size to the same output shape."""
     # Test with different image sizes
     sizes = [(100, 100), (500, 300), (224, 224)]
 
@@ -133,6 +139,7 @@ def test_preprocess_image_different_sizes(teachable_machine):
 
 
 def test_preprocess_image_content(teachable_machine):
+    """_preprocess_image should preserve the relative spatial layout of image content."""
     # Create a sample image with a specific pattern
     sample_array = np.zeros((300, 200, 3), dtype=np.uint8)
     sample_array[:100, :100] = 255  # White square in top-left corner
@@ -143,3 +150,71 @@ def test_preprocess_image_content(teachable_machine):
 
     # Check if the white square is still in the top-left corner (approximately)
     assert np.mean(processed[0, :50, :50]) > np.mean(processed[0, 50:, 50:])
+
+
+def test_compat_depthwise_conv2d_ignores_stray_groups_kwarg():
+    """
+    `_CompatDepthwiseConv2D` must silently drop an unrecognized 'groups'
+    kwarg, which Teachable Machine's exported .h5 configs always include.
+    Plain `tf.keras.layers.DepthwiseConv2D` raises on this under Keras 3
+    (TensorFlow >= 2.16); this is the direct unit-level check of the fix.
+    """
+    layer = _CompatDepthwiseConv2D(kernel_size=3, groups=1)
+    assert layer.kernel_size == (3, 3)
+
+
+def test_load_model_with_legacy_groups_config(tmp_path):
+    """
+    Regression test for https://github.com/MeqdadDev/teachable-machine/issues/2.
+
+    Teachable Machine's Keras/.h5 export embeds a stray 'groups': 1 key
+    in every saved DepthwiseConv2D layer config. Current Keras (bundled
+    by default since TensorFlow 2.16) validates configs strictly and
+    raises on that unrecognized key. This test builds a tiny model,
+    tampers its saved .h5 config to reproduce that legacy shape, and
+    verifies TeachableMachine still loads and classifies with it.
+    """
+    h5py = pytest.importorskip("h5py")
+    tf = pytest.importorskip("tensorflow")
+
+    model = tf.keras.Sequential(
+        [
+            tf.keras.layers.Input(shape=(224, 224, 3)),
+            tf.keras.layers.DepthwiseConv2D(
+                kernel_size=3, name="expanded_conv_depthwise"
+            ),
+            tf.keras.layers.Flatten(),
+            tf.keras.layers.Dense(2, activation="softmax"),
+        ]
+    )
+
+    model_path = tmp_path / "keras_model.h5"
+    model.save(model_path)
+
+    # Reproduce Teachable Machine's legacy DepthwiseConv2D config by
+    # injecting the unused 'groups' key into the saved model config.
+    with h5py.File(model_path, "r+") as f:
+        model_config = json.loads(f.attrs["model_config"])
+        layers = model_config["config"]["layers"]
+        patched_layers = [
+            layer for layer in layers if layer["class_name"] == "DepthwiseConv2D"
+        ]
+        assert patched_layers, "test setup: no DepthwiseConv2D layer to tamper"
+        for layer in patched_layers:
+            layer["config"]["groups"] = 1
+        f.attrs["model_config"] = json.dumps(model_config)
+
+    labels_path = tmp_path / "labels.txt"
+    labels_path.write_text("0 Class A\n1 Class B\n")
+
+    # Without the compatibility patch, this raises TypeError/ValueError
+    # on Keras 3: "Unrecognized keyword arguments... {'groups': 1}".
+    tm = TeachableMachine(
+        model_path=str(model_path), labels_file_path=str(labels_path)
+    )
+
+    sample_image = Image.new("RGB", (224, 224), color=(128, 128, 128))
+    result = tm._get_image_classification(sample_image)
+
+    assert result["class_name"] in {"0 Class A", "1 Class B"}
+    assert result["predictions"].shape == (2,)