transformers-haystack 0.2.0__tar.gz → 0.3.0__tar.gz

transformers_haystack-0.3.0/CHANGELOG.md

@@ -0,0 +1,28 @@
+# Changelog
+
+## [integrations/transformers-v0.2.0] - 2026-07-06
+
+### 📚 Documentation
+
+- Replace old haystack core imports with haystack_integrations paths (#3545)
+
+### 🧪 Testing
+
+- Improve del_hf_env_vars fixture (#3428)
+- Trust test modules under Haystack 3.0's deserialization allowlist (#3537)
+- Make Tool/Agent serialization assertions version-agnostic for Haystack 2.x/3.x (#3533)
+- Force Transformers and Sentence Transformers integration tests to run on CPU (#3550)
+
+### 🧹 Chores
+
+- Improve consistency of integrations folder structure (#3430)
+- Support sync streaming callbacks in async contexts for Haystack 2.x/3.x compatibility (#3534)
+
+
+## [integrations/transformers-v0.1.0] - 2026-06-08
+
+### 🚀 Features
+
+- Move Transformers components from Haystack (#3409)
+
+<!-- generated by git-cliff -->

{transformers_haystack-0.2.0 → transformers_haystack-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.5
 Name: transformers-haystack
-Version: 0.2.0
+Version: 0.3.0
 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-0.2.0 → transformers_haystack-0.3.0}/pyproject.toml

@@ -140,6 +140,7 @@ ignore = [
     "PLR0912",
     "PLR0913",
     "PLR0915",
+    "PLR0917",
     # Allow `Any` type - used legitimately for dynamic types and SDK boundaries
     "ANN401",
 ]

{transformers_haystack-0.2.0 → transformers_haystack-0.3.0}/src/haystack_integrations/components/generators/transformers/chat/chat_generator.py

@@ -353,7 +353,9 @@ class TransformersChatGenerator:
 
         :param messages: A list of ChatMessage objects representing the input messages. If a string is provided,
             it is converted to a list containing a ChatMessage with user role.
-        :param generation_kwargs: Additional keyword arguments for text generation.
+        :param generation_kwargs: Additional keyword arguments for text generation. These are merged per key with
+            the `generation_kwargs` passed at initialization: keys provided here take precedence, keys set only at
+            initialization are kept.
         :param streaming_callback: An optional callable for handling streaming responses.
         :param tools: A list of Tool and/or Toolset objects, or a single Toolset for which the model can prepare calls.
             If set, it will override the `tools` parameter provided during initialization.
@@ -472,7 +474,9 @@ class TransformersChatGenerator:
         and return values but can be used with `await` in an async code.
 
         :param messages: A list of ChatMessage objects representing the input messages.
-        :param generation_kwargs: Additional keyword arguments for text generation.
+        :param generation_kwargs: Additional keyword arguments for text generation. These are merged per key with
+            the `generation_kwargs` passed at initialization: keys provided here take precedence, keys set only at
+            initialization are kept.
         :param streaming_callback: An optional callable for handling streaming responses.
         :param tools: A list of Tool and/or Toolset objects, or a single Toolset for which the model can prepare calls.
             If set, it will override the `tools` parameter provided during initialization.

{transformers_haystack-0.2.0 → transformers_haystack-0.3.0}/src/haystack_integrations/components/routers/transformers/zero_shot_text_router.py

@@ -156,7 +156,11 @@ class TransformersZeroShotTextRouter:
             Dictionary with serialized data.
         """
         serialization_dict = default_to_dict(
-            self, labels=self.labels, huggingface_pipeline_kwargs=self.huggingface_pipeline_kwargs, token=self.token
+            self,
+            labels=self.labels,
+            multi_label=self.multi_label,
+            huggingface_pipeline_kwargs=self.huggingface_pipeline_kwargs,
+            token=self.token,
         )
 
         huggingface_pipeline_kwargs = serialization_dict["init_parameters"]["huggingface_pipeline_kwargs"]

{transformers_haystack-0.2.0 → transformers_haystack-0.3.0}/tests/test_chat_generator.py

@@ -440,6 +440,19 @@ class TestTransformersChatGenerator:
         assert chat_message.is_from(ChatRole.ASSISTANT)
         assert chat_message.text == "Berlin is cool"
 
+    def test_run_with_generation_kwargs(self, model_info_mock, mock_pipeline_with_tokenizer, chat_messages):
+        generator = TransformersChatGenerator(
+            model="meta-llama/Llama-2-13b-chat-hf",
+            generation_kwargs={"max_new_tokens": 100, "temperature": 0.5},
+        )
+        generator.pipeline = mock_pipeline_with_tokenizer
+
+        generator.run(messages=chat_messages, generation_kwargs={"temperature": 0.9})
+
+        _, kwargs = generator.pipeline.call_args
+        assert kwargs["max_new_tokens"] == 100
+        assert kwargs["temperature"] == 0.9
+
     def test_run_with_streaming_callback(self, model_info_mock, mock_pipeline_with_tokenizer, chat_messages):
         # Define the streaming callback function
         def streaming_callback_fn(chunk: StreamingChunk): ...
@@ -705,6 +718,21 @@ class TestTransformersChatGeneratorAsync:
         assert chat_message.text == "Berlin is cool"
         generator.shutdown()
 
+    @pytest.mark.asyncio
+    async def test_run_async_with_generation_kwargs(self, model_info_mock, mock_pipeline_with_tokenizer, chat_messages):
+        generator = TransformersChatGenerator(
+            model="meta-llama/Llama-2-13b-chat-hf",
+            generation_kwargs={"max_new_tokens": 100, "temperature": 0.5},
+        )
+        generator.pipeline = mock_pipeline_with_tokenizer
+
+        await generator.run_async(messages=chat_messages, generation_kwargs={"temperature": 0.9})
+
+        _, kwargs = generator.pipeline.call_args
+        assert kwargs["max_new_tokens"] == 100
+        assert kwargs["temperature"] == 0.9
+        generator.shutdown()
+
     @pytest.mark.asyncio
     async def test_run_async_with_string_input(self, model_info_mock, mock_pipeline_with_tokenizer):
         generator = TransformersChatGenerator(model="meta-llama/Llama-2-13b-chat-hf")

{transformers_haystack-0.2.0 → transformers_haystack-0.3.0}/tests/test_zero_shot_text_router.py

@@ -22,6 +22,7 @@ class TestTransformersZeroShotTextRouter:
             "type": COMPONENT_TYPE,
             "init_parameters": {
                 "labels": ["query", "passage"],
+                "multi_label": False,
                 "token": {"env_vars": ["HF_API_TOKEN", "HF_TOKEN"], "strict": False, "type": "env_var"},
                 "huggingface_pipeline_kwargs": {
                     "model": "MoritzLaurer/deberta-v3-base-zeroshot-v1.1-all-33",
@@ -31,6 +32,14 @@ class TestTransformersZeroShotTextRouter:
             },
         }
 
+    def test_multi_label_survives_a_serialization_round_trip(self, del_hf_env_vars_if_empty):
+        """`multi_label` changes how the pipeline normalizes scores, so losing it changes routing decisions."""
+        router = TransformersZeroShotTextRouter(labels=["query", "passage"], multi_label=True)
+
+        restored = TransformersZeroShotTextRouter.from_dict(router.to_dict())
+
+        assert restored.multi_label is True
+
     def test_from_dict(self, del_hf_env_vars_if_empty):
         data = {
             "type": COMPONENT_TYPE,

transformers_haystack-0.2.0/CHANGELOG.md

@@ -1,9 +0,0 @@
-# Changelog
-
-## [integrations/transformers-v0.1.0] - 2026-06-08
-
-### 🚀 Features
-
-- Move Transformers components from Haystack (#3409)
-
-<!-- generated by git-cliff -->