diff --git a/.github/workflows/python-package.yml b/.github/workflows/test-with-tox.yml
similarity index 78%
rename from .github/workflows/python-package.yml
rename to .github/workflows/test-with-tox.yml
index 63a1767..3f05ccf 100644
--- a/.github/workflows/python-package.yml
+++ b/.github/workflows/test-with-tox.yml
@@ -10,14 +10,11 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["2.7", "3.10"]
-
+        include:
+        - tox-envs: "sqla06-py27,sqla07-py27,sqla08-py27,sqla09-py27,sqlalchemy-py27"
+        - tox-envs: "sqlalchemy-py3"
     steps:
     - uses: actions/checkout@v3
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v3
-      with:
-        python-version: ${{ matrix.python-version }}
     - name: Install dependencies
       run: |
         sudo apt-get -y install graphviz
@@ -30,4 +27,4 @@ jobs:
         flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
     - name: Test with tox
       run: |
-        tox
+        tox -v -e ${{ matrix.tox-envs }} -- -v --color=yes
diff --git a/tests/test_schema_graph.py b/tests/test_schema_graph.py
index a2f30e3..50b19fe 100644
--- a/tests/test_schema_graph.py
+++ b/tests/test_schema_graph.py
@@ -32,7 +32,7 @@ def test_no_args():
 def test_empty_db(metadata):
     graph = sasd.create_schema_graph(metadata=metadata)
     assert isinstance(graph, pydot.Graph)
-    assert graph.create_plain() == 'graph 1 0 0\nstop\n'
+    assert graph.create_plain() == b'graph 1 0 0\nstop\n'
 
 
 def test_empty_table(metadata):
@@ -40,8 +40,8 @@ def test_empty_table(metadata):
         'foo', metadata,
         Column('id', types.Integer, primary_key=True))
     result = plain_result(metadata=metadata)
-    assert result.keys() == ['1']
-    assert result['1']['nodes'].keys() == ['foo']
+    assert list(result.keys()) == ['1']
+    assert list(result['1']['nodes'].keys()) == ['foo']
     assert '- id : INTEGER' in result['1']['nodes']['foo']
 
 
@@ -51,8 +51,8 @@ def test_empty_table_with_key_suffix(metadata):
         Column('id', types.Integer, primary_key=True))
     result = plain_result(metadata=metadata, show_column_keys=True)
     print(result)
-    assert result.keys() == ['1']
-    assert result['1']['nodes'].keys() == ['foo']
+    assert list(result.keys()) == ['1']
+    assert list(result['1']['nodes'].keys()) == ['foo']
     assert '- id(PK) : INTEGER' in result['1']['nodes']['foo']
 
 
@@ -64,7 +64,7 @@ def test_foreign_key(metadata):
         'bar', metadata,
         Column('foo_id', types.Integer, ForeignKey(foo.c.id)))
     result = plain_result(metadata=metadata)
-    assert result.keys() == ['1']
+    assert list(result.keys()) == ['1']
     assert sorted(result['1']['nodes'].keys()) == ['bar', 'foo']
     assert '- id : INTEGER' in result['1']['nodes']['foo']
     assert '- foo_id : INTEGER' in result['1']['nodes']['bar']
@@ -80,7 +80,7 @@ def test_foreign_key_with_key_suffix(metadata):
         'bar', metadata,
         Column('foo_id', types.Integer, ForeignKey(foo.c.id)))
     result = plain_result(metadata=metadata, show_column_keys=True)
-    assert result.keys() == ['1']
+    assert list(result.keys()) == ['1']
     assert sorted(result['1']['nodes'].keys()) == ['bar', 'foo']
     assert '- id(PK) : INTEGER' in result['1']['nodes']['foo']
     assert '- foo_id(FK) : INTEGER' in result['1']['nodes']['bar']
@@ -96,8 +96,8 @@ def test_table_filtering(metadata):
         'bar', metadata,
         Column('foo_id', types.Integer, ForeignKey(foo.c.id)))
     result = plain_result(tables=[bar])
-    assert result.keys() == ['1']
-    assert result['1']['nodes'].keys() == ['bar']
+    assert list(result.keys()) == ['1']
+    assert list(result['1']['nodes'].keys()) == ['bar']
     assert '- foo_id : INTEGER' in result['1']['nodes']['bar']
 
 def test_table_rendering_without_schema(metadata):
diff --git a/tests/test_uml_graph.py b/tests/test_uml_graph.py
index b546624..ff73257 100644
--- a/tests/test_uml_graph.py
+++ b/tests/test_uml_graph.py
@@ -33,8 +33,8 @@ class Foo(Base):
         __tablename__ = 'foo'
         id = Column(types.Integer, primary_key=True)
     result = plain_result(mappers(Foo))
-    assert result.keys() == ['1']
-    assert result['1']['nodes'].keys() == ['Foo']
+    assert list(result.keys()) == ['1']
+    assert list(result['1']['nodes'].keys()) == ['Foo']
     assert '+id : Integer' in result['1']['nodes']['Foo']
     out, err = capsys.readouterr()
     assert out == u''
diff --git a/tests/utils.py b/tests/utils.py
index 2f92d82..e716207 100644
--- a/tests/utils.py
+++ b/tests/utils.py
@@ -6,7 +6,8 @@
 
 def parse_graph(graph):
     result = {}
-    sio = StringIO(graph.create_plain())
+    graph_bytes = graph.create_plain()
+    sio = StringIO(graph_bytes.decode('utf-8'))
     graph = None
     for line in sio:
         line = line.strip()
diff --git a/tox.ini b/tox.ini
index 19db8af..bf29f6e 100644
--- a/tox.ini
+++ b/tox.ini
@@ -2,38 +2,30 @@
 
 [testenv]
 commands = py.test --cov-report html --cov-report term --cov sqlalchemy_schemadisplay {posargs}
-
-[basedeps]
 deps =
     pytest
     pytest-cov
 
 [testenv:sqla06-py27]
-basepython = python2.7
 deps =
-    {[basedeps]deps}
-    sqlalchemy>0.6dev,<0.7dev
+    {[testenv]deps}
+    sqlalchemy==0.6.*
 
 [testenv:sqla07-py27]
-basepython = python2.7
 deps =
-    {[basedeps]deps}
-    sqlalchemy>0.7dev,<0.8dev
+    {[testenv]deps}
+    sqlalchemy==0.7.*
 
 [testenv:sqla08-py27]
-basepython = python2.7
 deps =
-    {[basedeps]deps}
-    sqlalchemy>0.8dev,<0.9dev
+    {[testenv]deps}
+    sqlalchemy==0.8.*
 
 [testenv:sqla09-py27]
-basepython = python2.7
 deps =
-    {[basedeps]deps}
-    sqlalchemy>0.9dev,<1.0dev
+    {[testenv]deps}
+    sqlalchemy==0.9.*
 
 [testenv:sqlalchemy-py27]
-basepython = python2.7
-deps =
-    {[basedeps]deps}
-    sqlalchemy>0.9dev
+
+[testenv:sqlalchemy-py3]