transformers-haystack 1.0.0__tar.gz → 1.1.0__tar.gz

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/CHANGELOG.md

@@ -1,5 +1,23 @@
 # Changelog
 
+## [integrations/transformers-v1.0.1] - 2026-09-23
+
+### 🐛 Bug Fixes
+
+- Serialize missing init params in `to_dict` (transformers, amazon_bedrock) (#3873)
+
+
+## [integrations/transformers-v1.0.0] - 2026-09-10
+
+### 🐛 Bug Fixes
+
+- Fix (transformers): fix linting errors in transformers integration (#3867)
+
+### 🚜 Refactor
+
+- [**breaking**] Transformers - lifecycle refactor (#3939)
+
+
 ## [integrations/transformers-v0.3.0] - 2026-08-24
 
 ### 🐛 Bug Fixes

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: transformers-haystack
-Version: 1.0.0
+Version: 1.1.0
 Summary: Haystack integration for transformers
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/transformers#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/src/haystack_integrations/components/generators/transformers/chat/chat_generator.py

@@ -24,7 +24,6 @@ from haystack.tools import (
     flatten_tools_or_toolsets,
     serialize_tools_or_toolset,
 )
-from haystack.tools.utils import warm_up_tools
 from haystack.utils import ComponentDevice, Secret, deserialize_callable, serialize_callable
 from haystack.utils.hf import convert_message_to_hf_format, deserialize_hf_model_kwargs, serialize_hf_model_kwargs
 from huggingface_hub import model_info
@@ -238,7 +237,7 @@ class TransformersChatGenerator:
 
     def warm_up(self) -> None:
         """
-        Initializes the component and warms up tools if provided.
+        Initializes the Transformers pipeline and the executor owned by the component.
         """
         if self.pipeline is None:
             pipeline_kwargs = _with_hf_token(self.huggingface_pipeline_kwargs, self.token)
@@ -251,10 +250,7 @@ class TransformersChatGenerator:
                 raise ValueError(msg)
             pipeline_kwargs["task"] = task
 
-            hf_pipeline = pipeline(**pipeline_kwargs)
-            if self.tools:
-                warm_up_tools(self.tools)
-            self.pipeline = hf_pipeline
+            self.pipeline = pipeline(**pipeline_kwargs)
 
         if self._owns_executor and self.executor is None:
             self.executor = ThreadPoolExecutor(

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/src/haystack_integrations/components/readers/transformers/extractive_reader.py

@@ -150,6 +150,7 @@ class TransformersExtractiveReader:
             answers_per_seq=self.answers_per_seq,
             no_answer=self.no_answer,
             calibration_factor=self.calibration_factor,
+            overlap_threshold=self.overlap_threshold,
             model_kwargs=self.model_kwargs,
         )
 

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/tests/test_chat_generator.py

@@ -365,133 +365,6 @@ class TestComponentLifecycle:
             device="cpu",
         )
 
-    @patch("haystack_integrations.components.generators.transformers.chat.chat_generator.pipeline")
-    def test_warm_up_with_tools(self, pipeline_mock):
-        """Test that warm_up() calls warm_up on tools and is idempotent."""
-
-        # Create a mock tool that tracks if warm_up() was called
-        class MockTool(Tool):
-            warm_up_call_count = 0  # Class variable to track calls
-
-            def __init__(self):
-                super().__init__(
-                    name="mock_tool",
-                    description="A mock tool for testing",
-                    parameters={"x": {"type": "string"}},
-                    function=lambda x: x,
-                )
-
-            def warm_up(self):
-                MockTool.warm_up_call_count += 1
-
-        # Reset the class variable before test
-        MockTool.warm_up_call_count = 0
-        mock_tool = MockTool()
-
-        # Create TransformersChatGenerator with the mock tool
-        generator = TransformersChatGenerator(
-            model="mistralai/Mistral-7B-Instruct-v0.2",
-            task="text-generation",
-            device=ComponentDevice.from_str("cpu"),
-            tools=[mock_tool],
-        )
-
-        # Verify initial state - warm_up not called yet
-        assert MockTool.warm_up_call_count == 0
-        assert generator.pipeline is None
-
-        # Call warm_up() on the generator
-        generator.warm_up()
-
-        # Assert that the tool's warm_up() was called
-        assert MockTool.warm_up_call_count == 1
-        assert generator.pipeline is not None
-
-        # Verify pipeline was initialized
-        pipeline_mock.assert_called_once()
-
-        # Call warm_up() again and verify it's idempotent (only warms up once)
-        generator.warm_up()
-
-        # The tool's warm_up should still only have been called once
-        assert MockTool.warm_up_call_count == 1
-        assert generator.pipeline is not None
-        # Pipeline should still only have been called once
-        pipeline_mock.assert_called_once()
-
-    @patch("haystack_integrations.components.generators.transformers.chat.chat_generator.pipeline")
-    def test_warm_up_with_no_tools(self, pipeline_mock):
-        """Test that warm_up() works when no tools are provided."""
-
-        generator = TransformersChatGenerator(
-            model="mistralai/Mistral-7B-Instruct-v0.2", task="text-generation", device=ComponentDevice.from_str("cpu")
-        )
-
-        # Verify initial state
-        assert generator.pipeline is None
-        assert generator.tools is None
-
-        # Call warm_up() - should not raise an error
-        generator.warm_up()
-
-        # Verify the component is warmed up
-        assert generator.pipeline is not None
-        pipeline_mock.assert_called_once()
-
-        # Call warm_up() again - should be idempotent
-        generator.warm_up()
-        assert generator.pipeline is not None
-        # Pipeline should still only have been called once
-        pipeline_mock.assert_called_once()
-
-    @patch("haystack_integrations.components.generators.transformers.chat.chat_generator.pipeline")
-    def test_warm_up_with_multiple_tools(self, pipeline_mock):
-        """Test that warm_up() works with multiple tools."""
-
-        # Track warm_up calls
-        warm_up_calls = []
-
-        class MockTool(Tool):
-            def __init__(self, tool_name):
-                super().__init__(
-                    name=tool_name,
-                    description=f"Mock tool {tool_name}",
-                    parameters={"type": "object", "properties": {"x": {"type": "string"}}, "required": ["x"]},
-                    function=lambda x: f"{tool_name} result: {x}",
-                )
-
-            def warm_up(self):
-                warm_up_calls.append(self.name)
-
-        mock_tool1 = MockTool("tool1")
-        mock_tool2 = MockTool("tool2")
-
-        # Use a LIST of tools, not a Toolset
-        generator = TransformersChatGenerator(
-            model="mistralai/Mistral-7B-Instruct-v0.2",
-            task="text-generation",
-            device=ComponentDevice.from_str("cpu"),
-            tools=[mock_tool1, mock_tool2],
-        )
-
-        # Call warm_up()
-        generator.warm_up()
-
-        # Assert that both tools' warm_up() were called
-        assert "tool1" in warm_up_calls
-        assert "tool2" in warm_up_calls
-        assert generator.pipeline is not None
-        pipeline_mock.assert_called_once()
-
-        # Track count
-        call_count = len(warm_up_calls)
-
-        # Verify idempotency
-        generator.warm_up()
-        assert len(warm_up_calls) == call_count
-        # Pipeline should still only have been called once
-        pipeline_mock.assert_called_once()
-
 
 class TestRun:
     def test_run(self, mock_pipeline_with_tokenizer, chat_messages):

{transformers_haystack-1.0.0 → transformers_haystack-1.1.0}/tests/test_extractive_reader.py

@@ -150,6 +150,7 @@ def test_to_dict(initialized_token: Secret):
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {
                 "torch_dtype": "torch.float16",
                 "device_map": ComponentDevice.resolve_device(None).to_hf(),
@@ -158,6 +159,15 @@ def test_to_dict(initialized_token: Secret):
     }
 
 
+def test_overlap_threshold_survives_a_serialization_round_trip():
+    """overlap_threshold decides which overlapping answers are deduplicated away in run()."""
+    reader = TransformersExtractiveReader("my-model", token=None, overlap_threshold=0.5)
+
+    restored = TransformersExtractiveReader.from_dict(reader.to_dict())
+
+    assert restored.overlap_threshold == 0.5
+
+
 def test_to_dict_no_token():
     component = TransformersExtractiveReader("my-model", token=None, model_kwargs={"torch_dtype": torch.float16})
     data = component.to_dict()
@@ -178,6 +188,7 @@ def test_to_dict_no_token():
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {
                 "torch_dtype": "torch.float16",
                 "device_map": ComponentDevice.resolve_device(None).to_hf(),
@@ -206,6 +217,7 @@ def test_to_dict_empty_model_kwargs(initialized_token: Secret):
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {"device_map": ComponentDevice.resolve_device(None).to_hf()},
         },
     }
@@ -239,6 +251,7 @@ def test_to_dict_device_map(device_map, expected):
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {"device_map": expected},
         },
     }
@@ -261,6 +274,7 @@ def test_from_dict():
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {"torch_dtype": "torch.float16"},
         },
     }
@@ -325,6 +339,7 @@ def test_from_dict_no_token():
             "answers_per_seq": None,
             "no_answer": True,
             "calibration_factor": 0.1,
+            "overlap_threshold": 0.01,
             "model_kwargs": {"torch_dtype": "torch.float16"},
         },
     }