diff --git a/haystack/components/converters/azure.py b/haystack/components/converters/azure.py
index d761e2cbb..4c34b6478 100644
--- a/haystack/components/converters/azure.py
+++ b/haystack/components/converters/azure.py
@@ -95,7 +95,7 @@ class AzureOCRDocumentConverter:
 
         self.document_analysis_client = DocumentAnalysisClient(
             endpoint=endpoint, credential=AzureKeyCredential(api_key.resolve_value() or "")
-        )  # type: ignore
+        )
         self.endpoint = endpoint
         self.model_id = model_id
         self.api_key = api_key
@@ -383,10 +383,10 @@ class AzureOCRDocumentConverter:
             if all(line.polygon is not None for line in lines):
                 for i in range(len(lines)):  # pylint: disable=consider-using-enumerate
                     # left_upi, right_upi, right_lowi, left_lowi = lines[i].polygon
-                    left_upi, _, _, _ = lines[i].polygon  # type: ignore
+                    left_upi, _, _, _ = lines[i].polygon
                     pairs_by_page[page_idx].append([i, i])
                     for j in range(i + 1, len(lines)):  # pylint: disable=invalid-name
-                        left_upj, _, _, _ = lines[j].polygon  # type: ignore
+                        left_upj, _, _, _ = lines[j].polygon
                         close_on_y_axis = abs(left_upi[1] - left_upj[1]) < threshold_y
                         if close_on_y_axis:
                             pairs_by_page[page_idx].append([i, j])
@@ -422,13 +422,13 @@ class AzureOCRDocumentConverter:
         for page_idx, _ in enumerate(result.pages):
             sorted_rows = []
             for row_of_lines in merged_lines_by_page[page_idx]:
-                sorted_rows.append(sorted(row_of_lines, key=lambda x: x.polygon[0][0]))  # type: ignore
+                sorted_rows.append(sorted(row_of_lines, key=lambda x: x.polygon[0][0]))
             x_sorted_lines_by_page[page_idx] = sorted_rows
 
         # Sort each row within the page by the y-value of the upper left bounding box coordinate
         y_sorted_lines_by_page = {}
         for page_idx, _ in enumerate(result.pages):
-            sorted_rows = sorted(x_sorted_lines_by_page[page_idx], key=lambda x: x[0].polygon[0][1])  # type: ignore
+            sorted_rows = sorted(x_sorted_lines_by_page[page_idx], key=lambda x: x[0].polygon[0][1])
             y_sorted_lines_by_page[page_idx] = sorted_rows
 
         # Construct the text to write
diff --git a/haystack/components/generators/chat/hugging_face_api.py b/haystack/components/generators/chat/hugging_face_api.py
index 272e84d47..c12a5192b 100644
--- a/haystack/components/generators/chat/hugging_face_api.py
+++ b/haystack/components/generators/chat/hugging_face_api.py
@@ -602,7 +602,7 @@ class HuggingFaceAPIChatGenerator:
                 chunk=chunk, previous_chunks=streaming_chunks, component_info=component_info
             )
             streaming_chunks.append(stream_chunk)
-            await streaming_callback(stream_chunk)  # type: ignore
+            await streaming_callback(stream_chunk)
 
         message = _convert_streaming_chunks_to_chat_message(chunks=streaming_chunks)
         if message.meta.get("usage") is None:
diff --git a/haystack/components/generators/chat/hugging_face_local.py b/haystack/components/generators/chat/hugging_face_local.py
index 7e69b36c7..c256afa67 100644
--- a/haystack/components/generators/chat/hugging_face_local.py
+++ b/haystack/components/generators/chat/hugging_face_local.py
@@ -8,10 +8,10 @@ import re
 import sys
 from concurrent.futures import ThreadPoolExecutor
 from contextlib import asynccontextmanager, suppress
-from typing import Any, Callable, Literal, Optional, Union, cast
+from typing import Any, Callable, Literal, Optional, Union
 
 from haystack import component, default_from_dict, default_to_dict, logging
-from haystack.dataclasses import AsyncStreamingCallbackT, ChatMessage, ComponentInfo, StreamingCallbackT, ToolCall
+from haystack.dataclasses import ChatMessage, ComponentInfo, StreamingCallbackT, ToolCall
 from haystack.dataclasses.streaming_chunk import select_streaming_callback
 from haystack.lazy_imports import LazyImport
 from haystack.tools import (
@@ -473,8 +473,7 @@ class HuggingFaceLocalChatGenerator:
         if streaming_callback:
             async_handler = AsyncHFTokenStreamingHandler(
                 tokenizer=prepared_inputs["tokenizer"],
-                # Cast to AsyncStreamingCallbackT since we know streaming_callback is async
-                stream_handler=cast(AsyncStreamingCallbackT, streaming_callback),
+                stream_handler=streaming_callback,
                 stop_words=prepared_inputs["stop_words"],
                 component_info=ComponentInfo.from_component(self),
             )
diff --git a/haystack/components/generators/chat/openai.py b/haystack/components/generators/chat/openai.py
index 679c1a846..12ed1688c 100644
--- a/haystack/components/generators/chat/openai.py
+++ b/haystack/components/generators/chat/openai.py
@@ -416,7 +416,7 @@ class OpenAIChatGenerator:
 
         return {
             "model": self.model,
-            "messages": openai_formatted_messages,  # type: ignore[arg-type] # openai expects list of specific message types
+            "messages": openai_formatted_messages,
             "stream": streaming_callback is not None,
             "n": num_responses,
             **openai_tools,
diff --git a/haystack/components/generators/hugging_face_local.py b/haystack/components/generators/hugging_face_local.py
index 01cc4a23b..2a4bb7336 100644
--- a/haystack/components/generators/hugging_face_local.py
+++ b/haystack/components/generators/hugging_face_local.py
@@ -2,7 +2,7 @@
 #
 # SPDX-License-Identifier: Apache-2.0
 
-from typing import Any, Literal, Optional, cast
+from typing import Any, Literal, Optional
 
 from haystack import component, default_from_dict, default_to_dict, logging
 from haystack.dataclasses import ComponentInfo, StreamingCallbackT, select_streaming_callback
@@ -153,7 +153,7 @@ class HuggingFaceLocalGenerator:
             return
 
         if self.pipeline is None:
-            self.pipeline = cast(HfPipeline, pipeline(**self.huggingface_pipeline_kwargs))
+            self.pipeline = pipeline(**self.huggingface_pipeline_kwargs)
 
         if self.stop_words:
             # text-generation and text2text-generation pipelines always have a non-None tokenizer
diff --git a/haystack/components/generators/openai_dalle.py b/haystack/components/generators/openai_dalle.py
index 8d6e52bb3..7a907e2ec 100644
--- a/haystack/components/generators/openai_dalle.py
+++ b/haystack/components/generators/openai_dalle.py
@@ -143,7 +143,7 @@ class DALLEImageGenerator:
         :returns:
             The serialized component as a dictionary.
         """
-        return default_to_dict(  # type: ignore
+        return default_to_dict(
             self,
             model=self.model,
             quality=self.quality,
@@ -167,4 +167,4 @@ class DALLEImageGenerator:
         """
         init_params = data.get("init_parameters", {})
         deserialize_secrets_inplace(init_params, keys=["api_key"])
-        return default_from_dict(cls, data)  # type: ignore
+        return default_from_dict(cls, data)
diff --git a/haystack/components/preprocessors/csv_document_cleaner.py b/haystack/components/preprocessors/csv_document_cleaner.py
index 981f87b9b..d5b2a53f1 100644
--- a/haystack/components/preprocessors/csv_document_cleaner.py
+++ b/haystack/components/preprocessors/csv_document_cleaner.py
@@ -80,7 +80,7 @@ class CSVDocumentCleaner:
         cleaned_documents = []
         for document in documents:
             try:
-                df = pd.read_csv(StringIO(document.content), header=None, dtype=object)  # type: ignore
+                df = pd.read_csv(StringIO(document.content), header=None, dtype=object)
             except Exception as e:
                 logger.error(
                     "Error processing document {id}. Keeping it, but skipping cleaning. Error: {error}",
diff --git a/haystack/components/preprocessors/csv_document_splitter.py b/haystack/components/preprocessors/csv_document_splitter.py
index 5cbd6ddbb..4968506d7 100644
--- a/haystack/components/preprocessors/csv_document_splitter.py
+++ b/haystack/components/preprocessors/csv_document_splitter.py
@@ -107,7 +107,7 @@ class CSVDocumentSplitter:
         split_dfs = []
         for document in documents:
             try:
-                df = pd.read_csv(StringIO(document.content), **resolved_read_csv_kwargs)  # type: ignore
+                df = pd.read_csv(StringIO(document.content), **resolved_read_csv_kwargs)
             except Exception as e:
                 logger.error(f"Error processing document {document.id}. Keeping it, but skipping splitting. Error: {e}")
                 split_documents.append(document)
diff --git a/haystack/components/preprocessors/recursive_splitter.py b/haystack/components/preprocessors/recursive_splitter.py
index 2511d5263..c4f56e96c 100644
--- a/haystack/components/preprocessors/recursive_splitter.py
+++ b/haystack/components/preprocessors/recursive_splitter.py
@@ -289,7 +289,7 @@ class RecursiveDocumentSplitter:
         if self._chunk_length(text) <= self.split_length:
             return [text]
 
-        for curr_separator in self.separators:  # type: ignore # the caller already checked that separators is not None
+        for curr_separator in self.separators:
             if curr_separator == "sentence":
                 # re. ignore: correct SentenceSplitter initialization is checked at the initialization of the component
                 sentence_with_spans = self.nltk_tokenizer.split_sentences(text)  # type: ignore
diff --git a/haystack/components/rankers/meta_field_grouping_ranker.py b/haystack/components/rankers/meta_field_grouping_ranker.py
index 6a286726f..eb3265871 100644
--- a/haystack/components/rankers/meta_field_grouping_ranker.py
+++ b/haystack/components/rankers/meta_field_grouping_ranker.py
@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 from collections import defaultdict
-from typing import Any, Optional, cast
+from typing import Any, Optional
 
 from haystack import Document, component
 
@@ -107,7 +107,7 @@ class MetaFieldGroupingRanker:
         for subgroups in document_groups.values():
             for docs in subgroups.values():
                 if self.sort_docs_by:
-                    docs.sort(key=lambda d: d.meta.get(cast(str, self.sort_docs_by), float("inf")))
+                    docs.sort(key=lambda d: d.meta.get(self.sort_docs_by or "", float("inf")))
                 ordered_docs.extend(docs)
 
         ordered_docs.extend(no_group_docs)
diff --git a/haystack/components/rankers/transformers_similarity.py b/haystack/components/rankers/transformers_similarity.py
index 23ed954c4..77958b411 100644
--- a/haystack/components/rankers/transformers_similarity.py
+++ b/haystack/components/rankers/transformers_similarity.py
@@ -292,7 +292,7 @@ class TransformersSimilarityRanker:
             def __getitem__(self, item):
                 return {key: self.batch_encoding.data[key][item] for key in self.batch_encoding.data.keys()}
 
-        batch_enc = self.tokenizer(query_doc_pairs, padding=True, truncation=True, return_tensors="pt").to(  # type: ignore
+        batch_enc = self.tokenizer(query_doc_pairs, padding=True, truncation=True, return_tensors="pt").to(
             self.device.first_device.to_torch()
         )
         dataset = _Dataset(batch_enc)
@@ -301,7 +301,7 @@ class TransformersSimilarityRanker:
         similarity_scores = []
         with torch.inference_mode():
             for features in inp_dataloader:
-                model_preds = self.model(**features).logits.squeeze(dim=1)  # type: ignore
+                model_preds = self.model(**features).logits.squeeze(dim=1)
                 similarity_scores.extend(model_preds)
         similarity_scores = torch.stack(similarity_scores)
 
@@ -310,7 +310,7 @@ class TransformersSimilarityRanker:
 
         _, sorted_indices = torch.sort(similarity_scores, descending=True)
 
-        sorted_indices = sorted_indices.cpu().tolist()  # type: ignore
+        sorted_indices = sorted_indices.cpu().tolist()
         similarity_scores = similarity_scores.cpu().tolist()
         ranked_docs = []
         for sorted_index in sorted_indices:
diff --git a/haystack/components/tools/tool_invoker.py b/haystack/components/tools/tool_invoker.py
index 91646b3d1..558f9dd60 100644
--- a/haystack/components/tools/tool_invoker.py
+++ b/haystack/components/tools/tool_invoker.py
@@ -529,7 +529,7 @@ class ToolInvoker:
             with ThreadPoolExecutor(max_workers=self.max_workers) as executor:
                 futures = []
                 for params in tool_call_params:
-                    future = executor.submit(self._execute_single_tool_call, **params)  # type: ignore[arg-type]
+                    future = executor.submit(self._execute_single_tool_call, **params)
                     futures.append(future)
 
                 # 3) Process results in the order they are submitted
diff --git a/haystack/components/websearch/serper_dev.py b/haystack/components/websearch/serper_dev.py
index 930a9af34..be679cf10 100644
--- a/haystack/components/websearch/serper_dev.py
+++ b/haystack/components/websearch/serper_dev.py
@@ -110,7 +110,7 @@ class SerperDevWebSearch:
         headers = {"X-API-KEY": self.api_key.resolve_value(), "Content-Type": "application/json"}
 
         try:
-            response = requests.post(SERPERDEV_BASE_URL, headers=headers, data=payload, timeout=30)  # type: ignore
+            response = requests.post(SERPERDEV_BASE_URL, headers=headers, data=payload, timeout=30)
             response.raise_for_status()  # Will raise an HTTPError for bad responses
         except requests.Timeout as error:
             raise TimeoutError(f"Request to {self.__class__.__name__} timed out.") from error
diff --git a/haystack/core/pipeline/base.py b/haystack/core/pipeline/base.py
index 1cd01339f..27c3c0c1c 100644
--- a/haystack/core/pipeline/base.py
+++ b/haystack/core/pipeline/base.py
@@ -10,7 +10,7 @@ from enum import IntEnum
 from pathlib import Path
 from typing import Any, ContextManager, Iterator, Mapping, Optional, TextIO, TypeVar, Union
 
-import networkx  # type:ignore
+import networkx
 
 from haystack import logging, tracing
 from haystack.core.component import Component, InputSocket, OutputSocket, component
@@ -123,7 +123,7 @@ class PipelineBase:  # noqa: PLW1641
                 res += f"  - {k}: {v}\n"
 
         res += "🚅 Components\n"
-        for name, instance in self.graph.nodes(data="instance"):  # type: ignore # type wrongly defined in networkx
+        for name, instance in self.graph.nodes(data="instance"):
             res += f"  - {name}: {instance.__class__.__name__}\n"
 
         res += "🛤️ Connections\n"
@@ -144,7 +144,7 @@ class PipelineBase:  # noqa: PLW1641
             Dictionary with serialized data.
         """
         components = {}
-        for name, instance in self.graph.nodes(data="instance"):  # type:ignore
+        for name, instance in self.graph.nodes(data="instance"):
             components[name] = component_to_dict(instance, name)
 
         connections = []
@@ -619,7 +619,7 @@ class PipelineBase:  # noqa: PLW1641
         :returns:
             The name of the Component instance.
         """
-        for name, inst in self.graph.nodes(data="instance"):  # type: ignore # type wrongly defined in networkx
+        for name, inst in self.graph.nodes(data="instance"):
             if inst == instance:
                 return name
         return ""
@@ -717,7 +717,7 @@ class PipelineBase:  # noqa: PLW1641
         """
 
         if is_in_jupyter():
-            from IPython.display import Image, display  # type: ignore
+            from IPython.display import Image, display
 
             if super_component_expansion:
                 graph, super_component_mapping = self._merge_super_component_pipelines()
@@ -811,7 +811,7 @@ class PipelineBase:  # noqa: PLW1641
         :returns:
             An iterator of tuples of component name and component instance.
         """
-        for component_name, instance in self.graph.nodes(data="instance"):  # type: ignore # type is wrong in networkx
+        for component_name, instance in self.graph.nodes(data="instance"):
             yield component_name, instance
 
     def warm_up(self) -> None:
diff --git a/haystack/core/pipeline/descriptions.py b/haystack/core/pipeline/descriptions.py
index b229f3082..a167aa4e9 100644
--- a/haystack/core/pipeline/descriptions.py
+++ b/haystack/core/pipeline/descriptions.py
@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 
-import networkx  # type:ignore
+import networkx
 
 from haystack.core.component.types import InputSocket, InputSocketTypeDescriptor, OutputSocket
 from haystack.core.type_utils import _type_name
diff --git a/haystack/core/pipeline/draw.py b/haystack/core/pipeline/draw.py
index 9703ba771..39a685ee3 100644
--- a/haystack/core/pipeline/draw.py
+++ b/haystack/core/pipeline/draw.py
@@ -9,7 +9,7 @@ import random
 import zlib
 from typing import Any, Optional
 
-import networkx  # type:ignore
+import networkx
 import requests
 
 from haystack import logging
diff --git a/haystack/testing/sample_components/accumulate.py b/haystack/testing/sample_components/accumulate.py
index e9e30e792..670ed8017 100644
--- a/haystack/testing/sample_components/accumulate.py
+++ b/haystack/testing/sample_components/accumulate.py
@@ -37,7 +37,7 @@ class Accumulate:
             import it at need. This is also a parameter.
         """
         self.state = 0
-        self.function: Callable = _default_function if function is None else function  # type: ignore
+        self.function: Callable = _default_function if function is None else function
 
     def to_dict(self) -> dict[str, Any]:
         """Converts the component to a dictionary"""
diff --git a/haystack/tools/parameters_schema_utils.py b/haystack/tools/parameters_schema_utils.py
index 9feea0005..707671d14 100644
--- a/haystack/tools/parameters_schema_utils.py
+++ b/haystack/tools/parameters_schema_utils.py
@@ -161,7 +161,7 @@ def _resolve_type(_type: Any) -> Any:
         return Sequence[_resolve_type(args[0]) if args else Any]  # type: ignore[misc]
 
     if origin is Union:
-        return Union[tuple(_resolve_type(a) for a in args)]  # type: ignore[misc]
+        return Union[tuple(_resolve_type(a) for a in args)]
 
     if origin is dict:
         return dict[args[0] if args else Any, _resolve_type(args[1]) if args else Any]  # type: ignore[misc]
diff --git a/pyproject.toml b/pyproject.toml
index 4d8c50adb..5c3c21cd4 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -276,10 +276,12 @@ asyncio_default_fixture_loop_scope = "class"
 
 [tool.mypy]
 python_version = "3.9"
-disallow_incomplete_defs = true
-warn_return_any = false
-ignore_missing_imports = true
 check_untyped_defs = true
+disallow_incomplete_defs = true
+ignore_missing_imports = true
+warn_unused_ignores = true
+warn_redundant_casts = true
+warn_return_any = false
 
 [[tool.mypy.overrides]]
 # TODO: Fix component typings