diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 1bfad81..3510c5d 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -24,7 +24,7 @@ repos:
     - id: ruff-format
 
 - repo: https://github.com/pre-commit/mirrors-mypy
-  rev: v1.11.2
+  rev: v1.14.1
   hooks:
   -   id: mypy
       args: [--install-types, --non-interactive, --ignore-missing-imports]
diff --git a/pyproject.toml b/pyproject.toml
index c3aae10..ddbcd31 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -5,6 +5,14 @@ build-backend = "setuptools.build_meta"
 [tool.black]
 line-length = 79
 
+[[tool.mypy.overrides]]
+module = ["beancount.*"]
+follow_untyped_imports = true
+
+[[tool.mypy.overrides]]
+module = ["beangulp.*"]
+follow_untyped_imports = true
+
 [tool.ruff]
 target-version = "py38"
 line-length = 79
diff --git a/smart_importer/hooks.py b/smart_importer/hooks.py
index f2ca2e8..1e3517a 100644
--- a/smart_importer/hooks.py
+++ b/smart_importer/hooks.py
@@ -1,7 +1,13 @@
 """Importer decorators."""
 
+from __future__ import annotations
+
 import logging
 from functools import wraps
+from typing import Callable, Sequence
+
+from beancount.core import data
+from beangulp import Adapter, Importer, ImporterProtocol
 
 logger = logging.getLogger(__name__)  # pylint: disable=invalid-name
 
@@ -9,7 +15,13 @@
 class ImporterHook:
     """Interface for an importer hook."""
 
-    def __call__(self, importer, file, imported_entries, existing):
+    def __call__(
+        self,
+        importer: Importer,
+        file: str,
+        imported_entries: data.Directives,
+        existing: data.Directives,
+    ) -> data.Directives:
         """Apply the hook and modify the imported entries.
 
         Args:
@@ -25,7 +37,14 @@ def __call__(self, importer, file, imported_entries, existing):
         raise NotImplementedError
 
 
-def apply_hooks(importer, hooks):
+def apply_hooks(
+    importer: Importer | ImporterProtocol,
+    hooks: Sequence[
+        Callable[
+            [Importer, str, data.Directives, data.Directives], data.Directives
+        ]
+    ],
+) -> Importer:
     """Apply a list of importer hooks to an importer.
 
     Args:
@@ -33,12 +52,16 @@ def apply_hooks(importer, hooks):
         hooks: A list of hooks, each a callable object.
     """
 
+    if not isinstance(importer, Importer):
+        importer = Adapter(importer)
     unpatched_extract = importer.extract
 
     @wraps(unpatched_extract)
-    def patched_extract_method(filepath, existing=None):
+    def patched_extract_method(
+        filepath: str, existing: data.Directives
+    ) -> data.Directives:
         logger.debug("Calling the importer's extract method.")
-        imported_entries = unpatched_extract(filepath, existing=existing)
+        imported_entries = unpatched_extract(filepath, existing)
 
         for hook in hooks:
             imported_entries = hook(
diff --git a/smart_importer/predictor.py b/smart_importer/predictor.py
index eeec34b..c3999d1 100644
--- a/smart_importer/predictor.py
+++ b/smart_importer/predictor.py
@@ -6,10 +6,10 @@
 
 import logging
 import threading
-from typing import TYPE_CHECKING, Callable
+from typing import TYPE_CHECKING, Any, Callable
 
+from beancount.core import data
 from beancount.core.data import (
-    ALL_DIRECTIVES,
     Close,
     Open,
     Transaction,
@@ -27,6 +27,7 @@
 from smart_importer.pipelines import get_pipeline
 
 if TYPE_CHECKING:
+    from beangulp import Importer
     from sklearn import Pipeline
 
 logger = logging.getLogger(__name__)  # pylint: disable=invalid-name
@@ -53,11 +54,11 @@ class EntryPredictor(ImporterHook):
 
     def __init__(
         self,
-        predict=True,
-        overwrite=False,
+        predict: bool = True,
+        overwrite: bool = False,
         string_tokenizer: Callable[[str], list] | None = None,
         denylist_accounts: list[str] | None = None,
-    ):
+    ) -> None:
         super().__init__()
         self.training_data = None
         self.open_accounts: dict[str, str] = {}
@@ -65,13 +66,19 @@ def __init__(
         self.pipeline: Pipeline | None = None
         self.is_fitted = False
         self.lock = threading.Lock()
-        self.account = None
+        self.account: str | None = None
 
         self.predict = predict
         self.overwrite = overwrite
         self.string_tokenizer = string_tokenizer
 
-    def __call__(self, importer, file, imported_entries, existing_entries):
+    def __call__(
+        self,
+        importer: Importer,
+        file: str,
+        imported_entries: data.Directives,
+        existing_entries: data.Directives,
+    ) -> data.Directives:
         """Predict attributes for imported transactions.
 
         Args:
@@ -157,7 +164,7 @@ def targets(self):
             for entry in self.training_data
         ]
 
-    def define_pipeline(self):
+    def define_pipeline(self) -> None:
         """Defines the machine learning pipeline based on given weights."""
 
         transformers = [
@@ -172,7 +179,7 @@ def define_pipeline(self):
             SVC(kernel="linear"),
         )
 
-    def train_pipeline(self):
+    def train_pipeline(self) -> None:
         """Train the machine learning pipeline."""
 
         self.is_fitted = False
@@ -187,11 +194,14 @@ def train_pipeline(self):
             self.is_fitted = True
             logger.debug("Only one target possible.")
         else:
+            assert self.pipeline is not None
             self.pipeline.fit(self.training_data, self.targets)
             self.is_fitted = True
             logger.debug("Trained the machine learning model.")
 
-    def process_entries(self, imported_entries) -> list[ALL_DIRECTIVES]:
+    def process_entries(
+        self, imported_entries: data.Directives
+    ) -> data.Directives:
         """Process imported entries.
 
         Transactions might be modified, all other entries are left as is.
@@ -206,7 +216,9 @@ def process_entries(self, imported_entries) -> list[ALL_DIRECTIVES]:
             imported_entries, enhanced_transactions
         )
 
-    def apply_prediction(self, entry, prediction):
+    def apply_prediction(
+        self, entry: data.Transaction, prediction: Any
+    ) -> data.Transaction:
         """Apply a single prediction to an entry.
 
         Args:
diff --git a/tests/data_test.py b/tests/data_test.py
index f219f04..f21705b 100644
--- a/tests/data_test.py
+++ b/tests/data_test.py
@@ -1,11 +1,15 @@
 """Tests for the `PredictPostings` decorator"""
 
+from __future__ import annotations
+
 # pylint: disable=missing-docstring
 import os
 import pprint
 import re
+from typing import Callable
 
 import pytest
+from beancount.core import data
 from beancount.core.compare import stable_hash_namedtuple
 from beancount.parser import parser
 from beangulp import Importer
@@ -13,17 +17,19 @@
 from smart_importer import PredictPostings, apply_hooks
 
 
-def chinese_string_tokenizer(pre_tokenizer_string):
+def chinese_string_tokenizer(pre_tokenizer_string: str) -> list[str]:
     jieba = pytest.importorskip("jieba")
     jieba.initialize()
     return list(jieba.cut(pre_tokenizer_string))
 
 
-def _hash(entry):
+def _hash(entry: data.Directive) -> str:
     return stable_hash_namedtuple(entry, ignore={"meta", "units"})
 
 
-def _load_testset(testset):
+def _load_testset(
+    testset: str,
+) -> tuple[data.Directives, data.Directives, data.Directives]:
     path = os.path.join(
         os.path.dirname(__file__), "data", testset + ".beancount"
     )
@@ -35,7 +41,7 @@ def _load_testset(testset):
         assert not errors
         parsed_sections.append(entries)
     assert len(parsed_sections) == 3
-    return parsed_sections
+    return tuple(parsed_sections)
 
 
 @pytest.mark.parametrize(
@@ -47,25 +53,27 @@ def _load_testset(testset):
         ("chinese", chinese_string_tokenizer),
     ],
 )
-def test_testset(testset, string_tokenizer):
+def test_testset(
+    testset: str, string_tokenizer: Callable[[str], list[str]]
+) -> None:
     # pylint: disable=unbalanced-tuple-unpacking
     imported, training_data, expected = _load_testset(testset)
 
     class DummyImporter(Importer):
-        def extract(self, filepath, existing=None):
+        def extract(
+            self, filepath: str, existing: data.Directives
+        ) -> data.Directives:
             return imported
 
-        def account(self, filepath):
+        def account(self, filepath: str) -> str:
             return ""
 
-        def identify(self, filepath):
+        def identify(self, filepath: str) -> bool:
             return True
 
     importer = DummyImporter()
     apply_hooks(importer, [PredictPostings(string_tokenizer=string_tokenizer)])
-    imported_transactions = importer.extract(
-        "dummy-data", existing=training_data
-    )
+    imported_transactions = importer.extract("dummy-data", training_data)
 
     for txn1, txn2 in zip(imported_transactions, expected):
         if _hash(txn1) != _hash(txn2):
diff --git a/tests/predictors_test.py b/tests/predictors_test.py
index dd3de51..4162d0a 100644
--- a/tests/predictors_test.py
+++ b/tests/predictors_test.py
@@ -1,6 +1,7 @@
 """Tests for the `PredictPayees` and the `PredictPostings` decorator"""
 
 # pylint: disable=missing-docstring
+from beancount.core import data
 from beancount.parser import parser
 from beangulp import Importer
 
@@ -133,7 +134,9 @@
 
 
 class BasicTestImporter(Importer):
-    def extract(self, filepath, existing=None):
+    def extract(
+        self, filepath: str, existing: data.Directives
+    ) -> data.Directives:
         if filepath == "dummy-data":
             return TEST_DATA
         if filepath == "empty":
@@ -141,10 +144,10 @@ def extract(self, filepath, existing=None):
         assert False
         return []
 
-    def account(self, filepath):
+    def account(self, filepath: str) -> str:
         return "Assets:US:BofA:Checking"
 
-    def identify(self, filepath):
+    def identify(self, filepath: str) -> bool:
         return True
 
 
@@ -155,39 +158,38 @@ def identify(self, filepath):
 )
 
 
-def test_empty_training_data():
+def test_empty_training_data() -> None:
     """
     Verifies that the decorator leaves the narration intact.
     """
-    assert POSTING_IMPORTER.extract("dummy-data") == TEST_DATA
-    assert PAYEE_IMPORTER.extract("dummy-data") == TEST_DATA
+    assert POSTING_IMPORTER.extract("dummy-data", []) == TEST_DATA
+    assert PAYEE_IMPORTER.extract("dummy-data", []) == TEST_DATA
 
 
-def test_no_transactions():
+def test_no_transactions() -> None:
     """
     Should not crash when passed empty list of transactions.
     """
-    POSTING_IMPORTER.extract("empty")
-    PAYEE_IMPORTER.extract("empty")
-    POSTING_IMPORTER.extract("empty", existing=TRAINING_DATA)
-    PAYEE_IMPORTER.extract("empty", existing=TRAINING_DATA)
+    POSTING_IMPORTER.extract("empty", [])
+    PAYEE_IMPORTER.extract("empty", [])
+    POSTING_IMPORTER.extract("empty", TRAINING_DATA)
+    PAYEE_IMPORTER.extract("empty", TRAINING_DATA)
 
 
-def test_unchanged_narrations():
+def test_unchanged_narrations() -> None:
     """
     Verifies that the decorator leaves the narration intact
     """
     correct_narrations = [transaction.narration for transaction in TEST_DATA]
     extracted_narrations = [
         transaction.narration
-        for transaction in PAYEE_IMPORTER.extract(
-            "dummy-data", existing=TRAINING_DATA
-        )
+        for transaction in PAYEE_IMPORTER.extract("dummy-data", TRAINING_DATA)
+        if isinstance(transaction, data.Transaction)
     ]
     assert extracted_narrations == correct_narrations
 
 
-def test_unchanged_first_posting():
+def test_unchanged_first_posting() -> None:
     """
     Verifies that the decorator leaves the first posting intact
     """
@@ -196,30 +198,32 @@ def test_unchanged_first_posting():
     ]
     extracted_first_postings = [
         transaction.postings[0]
-        for transaction in PAYEE_IMPORTER.extract(
-            "dummy-data", existing=TRAINING_DATA
-        )
+        for transaction in PAYEE_IMPORTER.extract("dummy-data", TRAINING_DATA)
+        if isinstance(transaction, data.Transaction)
     ]
     assert extracted_first_postings == correct_first_postings
 
 
-def test_payee_predictions():
+def test_payee_predictions() -> None:
     """
     Verifies that the decorator adds predicted postings.
     """
-    transactions = PAYEE_IMPORTER.extract("dummy-data", existing=TRAINING_DATA)
-    predicted_payees = [transaction.payee for transaction in transactions]
+    transactions = PAYEE_IMPORTER.extract("dummy-data", TRAINING_DATA)
+    predicted_payees = [
+        transaction.payee
+        for transaction in transactions
+        if isinstance(transaction, data.Transaction)
+    ]
     assert predicted_payees == PAYEE_PREDICTIONS
 
 
-def test_account_predictions():
+def test_account_predictions() -> None:
     """
     Verifies that the decorator adds predicted postings.
     """
     predicted_accounts = [
         entry.postings[-1].account
-        for entry in POSTING_IMPORTER.extract(
-            "dummy-data", existing=TRAINING_DATA
-        )
+        for entry in POSTING_IMPORTER.extract("dummy-data", TRAINING_DATA)
+        if isinstance(entry, data.Transaction)
     ]
     assert predicted_accounts == ACCOUNT_PREDICTIONS