diff --git a/cognee/eval_framework/modal_eval_dashboard.py b/cognee/eval_framework/modal_eval_dashboard.py
index 9ff6f543c..e2ed8e453 100644
--- a/cognee/eval_framework/modal_eval_dashboard.py
+++ b/cognee/eval_framework/modal_eval_dashboard.py
@@ -16,7 +16,7 @@ metrics_volume = modal.Volume.from_name("evaluation_dashboard_results", create_i
 
 image = (
     modal.Image.debian_slim(python_version="3.11")
-    .pip_install("streamlit", "plotly")
+    .pip_install("streamlit", "pandas", "plotly")
     .add_local_file(__file__, "/root/serve_dashboard.py")
 )
 
diff --git a/cognee/infrastructure/data/utils/extract_keywords.py b/cognee/infrastructure/data/utils/extract_keywords.py
deleted file mode 100644
index c3e47c4c8..000000000
--- a/cognee/infrastructure/data/utils/extract_keywords.py
+++ /dev/null
@@ -1,45 +0,0 @@
-from cognee.infrastructure.data.exceptions.exceptions import KeywordExtractionError
-
-
-# def extract_keywords(text: str) -> list[str]:
-#     """
-#     Extract keywords from the provided text string.
-
-#     This function raises an KeyWordExtractionError if the input text is empty. It processes the
-#     text to extract parts of speech, focusing on nouns, and uses TF-IDF to identify the most
-#     relevant keywords based on their frequency. The function returns a list of up to 15
-#     keywords, each having more than 3 characters.
-
-#     Parameters:
-#     -----------
-
-#         - text (str): The input text from which to extract keywords.
-
-#     Returns:
-#     --------
-
-#         - list[str]: A list of keywords extracted from the text, containing up to 15 nouns
-#           with more than 3 characters.
-#     """
-#     if len(text) == 0:
-#         raise KeywordExtractionError()
-
-#     tags = extract_pos_tags(text)
-#     nouns = [word for (word, tag) in tags if tag == "NN"]
-
-#     vectorizer = TfidfVectorizer()
-#     tfidf = vectorizer.fit_transform(nouns)
-
-#     top_nouns = sorted(
-#         vectorizer.vocabulary_, key=lambda x: tfidf[0, vectorizer.vocabulary_[x]], reverse=True
-#     )
-
-#     keywords = []
-
-#     for word in top_nouns:
-#         if len(word) > 3:
-#             keywords.append(word)
-#         if len(keywords) >= 15:
-#             break
-
-#     return keywords
diff --git a/cognee/infrastructure/llm/config.py b/cognee/infrastructure/llm/config.py
index c5240935c..9e5120456 100644
--- a/cognee/infrastructure/llm/config.py
+++ b/cognee/infrastructure/llm/config.py
@@ -75,6 +75,13 @@ class LLMConfig(BaseSettings):
 
     def model_post_init(self, __context) -> None:
         """Initialize the BAML registry after the model is created."""
+        # Check if BAML is selected as structured output framework but not available
+        if self.structured_output_framework == "baml" and ClientRegistry is None:
+            raise ImportError(
+                "BAML is selected as structured output framework but not available. "
+                "Please install with 'pip install cognee[baml]' to use BAML extraction features."
+            )
+
         if ClientRegistry is not None:
             self.baml_registry = ClientRegistry()
             self.baml_registry.add_llm_client(
diff --git a/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/extract_summary.py b/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/extract_summary.py
index 89889d294..697a52a45 100644
--- a/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/extract_summary.py
+++ b/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/extract_summary.py
@@ -37,11 +37,6 @@ async def extract_summary(content: str, response_model: Type[BaseModel]):
     """
     config = get_llm_config()
 
-    if config.baml_registry is None:
-        raise ImportError(
-            "BAML is not available. Please install with 'pip install cognee[baml]' to use BAML extraction features."
-        )
-
     # Use BAML's SummarizeContent function
     summary_result = await b.SummarizeContent(
         content, baml_options={"client_registry": config.baml_registry}
@@ -82,11 +77,6 @@ async def extract_code_summary(content: str):
         try:
             config = get_llm_config()
 
-            if config.baml_registry is None:
-                raise ImportError(
-                    "BAML is not available. Please install with 'pip install cognee[baml]' to use BAML extraction features."
-                )
-
             result = await b.SummarizeCode(
                 content, baml_options={"client_registry": config.baml_registry}
             )
diff --git a/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/knowledge_graph/extract_content_graph.py b/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/knowledge_graph/extract_content_graph.py
index f87d87d1b..abff07e09 100644
--- a/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/knowledge_graph/extract_content_graph.py
+++ b/cognee/infrastructure/llm/structured_output_framework/baml/baml_src/extraction/knowledge_graph/extract_content_graph.py
@@ -16,11 +16,6 @@ async def extract_content_graph(
 
     get_logger(level="INFO")
 
-    if config.baml_registry is None:
-        raise ImportError(
-            "BAML is not available. Please install with 'pip install cognee[baml]' to use BAML extraction features."
-        )
-
     # if response_model:
     #     # tb = TypeBuilder()
     #     # country = tb.union \
diff --git a/cognee/shared/utils.py b/cognee/shared/utils.py
index 16d0961cc..3071a82cb 100644
--- a/cognee/shared/utils.py
+++ b/cognee/shared/utils.py
@@ -17,46 +17,6 @@ from cognee.infrastructure.databases.graph import get_graph_engine
 proxy_url = "https://test.prometh.ai"
 
 
-def get_entities(tagged_tokens):
-    try:
-        import nltk
-
-        nltk.download("maxent_ne_chunker", quiet=True)
-        from nltk.chunk import ne_chunk
-
-        return ne_chunk(tagged_tokens)
-    except ImportError:
-        raise ImportError(
-            "NLTK is required for entity extraction. Install with 'pip install cognee[nlp]' to use this feature."
-        )
-
-
-def extract_pos_tags(sentence):
-    """Extract Part-of-Speech (POS) tags for words in a sentence."""
-    try:
-        import nltk
-
-        # Ensure that the necessary NLTK resources are downloaded
-        nltk.download("words", quiet=True)
-        nltk.download("punkt", quiet=True)
-        nltk.download("averaged_perceptron_tagger", quiet=True)
-
-        from nltk.tag import pos_tag
-        from nltk.tokenize import word_tokenize
-
-        # Tokenize the sentence into words
-        tokens = word_tokenize(sentence)
-
-        # Tag each word with its corresponding POS tag
-        pos_tags = pos_tag(tokens)
-
-        return pos_tags
-    except ImportError:
-        raise ImportError(
-            "NLTK is required for POS tagging. Install with 'pip install cognee[nlp]' to use this feature."
-        )
-
-
 def get_anonymous_id():
     """Creates or reads a anonymous user id"""
     tracking_id = os.getenv("TRACKING_ID", None)
diff --git a/pyproject.toml b/pyproject.toml
index 6667819c8..2a4c36132 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -49,7 +49,8 @@ dependencies = [
     "python-magic-bin<0.5 ; platform_system == 'Windows'", # Only needed for Windows
     "fastembed<=0.6.0 ",
     "networkx>=3.4.2,<4",
-    "matplotlib>=3.8.3,<4"
+    "matplotlib>=3.8.3,<4",
+    "baml-py>=0.201.0"
 
 ]
 
@@ -66,18 +67,12 @@ distributed = [
 # Database backends
 neo4j = ["neo4j>=5.28.0,<6"]
 neptune = ["langchain_aws>=0.2.22"]
-# PostgreSQL support (binary - no compilation required)
 postgres = [
+    "psycopg2>=2.9.10,<3",
     "psycopg2-binary>=2.9.10,<3.0.0",  # Pre-compiled binary, no PostgreSQL headers needed
     "pgvector>=0.3.5,<0.4",
     "asyncpg>=0.30.0,<1.0.0",
 ]
-# PostgreSQL support (source - requires PostgreSQL development headers)
-postgres-source = [
-    "psycopg2>=2.9.10,<3 ; platform_system != 'Windows'",  # Requires libpq-dev, build tools
-    "pgvector>=0.3.5,<0.4",
-    "asyncpg>=0.30.0,<1.0.0",
-]
 notebook = ["notebook>=7.1.0,<8"]
 langchain = [
     "langsmith>=0.2.3,<1.0.0",