Merge branch 'dev' into feature/cog-3014-refactor-delete-feature

.env.template

@@ -28,11 +28,10 @@ EMBEDDING_ENDPOINT=""
 EMBEDDING_API_VERSION=""
 EMBEDDING_DIMENSIONS=3072
 EMBEDDING_MAX_TOKENS=8191
+EMBEDDING_BATCH_SIZE=36
 # If embedding key is not provided same key set for LLM_API_KEY will be used
 #EMBEDDING_API_KEY="your_api_key"
-# Note: OpenAI support up to 2048 elements and Gemini supports a maximum of 100 elements in an embedding batch,
+
-#       Cognee sets the optimal batch size for OpenAI and Gemini, but a custom size can be defined if necessary for other models
-#EMBEDDING_BATCH_SIZE=2048
 
 # If using BAML structured output these env variables will be used
 BAML_LLM_PROVIDER=openai
@@ -248,10 +247,10 @@ LITELLM_LOG="ERROR"
 #LLM_PROVIDER="ollama"
 #LLM_ENDPOINT="http://localhost:11434/v1"
 #EMBEDDING_PROVIDER="ollama"
-#EMBEDDING_MODEL="avr/sfr-embedding-mistral:latest"
+#EMBEDDING_MODEL="nomic-embed-text:latest"
 #EMBEDDING_ENDPOINT="http://localhost:11434/api/embeddings"
-#EMBEDDING_DIMENSIONS=4096
+#EMBEDDING_DIMENSIONS=768
-#HUGGINGFACE_TOKENIZER="Salesforce/SFR-Embedding-Mistral"
+#HUGGINGFACE_TOKENIZER="nomic-ai/nomic-embed-text-v1.5"
 
 ########## OpenRouter (also free) #########################################################
 

.github/workflows/examples_tests.yml

@@ -110,6 +110,81 @@ jobs:
           EMBEDDING_API_VERSION: ${{ secrets.EMBEDDING_API_VERSION }}
         run: uv run python ./examples/python/dynamic_steps_example.py
 
+  test-temporal-example:
+    name: Run Temporal Tests
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Cognee Setup
+        uses: ./.github/actions/cognee_setup
+        with:
+          python-version: '3.11.x'
+
+      - name: Run Temporal Example
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_ENDPOINT: ${{ secrets.LLM_ENDPOINT }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_API_VERSION: ${{ secrets.LLM_API_VERSION }}
+          EMBEDDING_MODEL: ${{ secrets.EMBEDDING_MODEL }}
+          EMBEDDING_ENDPOINT: ${{ secrets.EMBEDDING_ENDPOINT }}
+          EMBEDDING_API_KEY: ${{ secrets.EMBEDDING_API_KEY }}
+          EMBEDDING_API_VERSION: ${{ secrets.EMBEDDING_API_VERSION }}
+        run: uv run python ./examples/python/temporal_example.py
+
+  test-ontology-example:
+    name: Run Ontology Tests
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Cognee Setup
+        uses: ./.github/actions/cognee_setup
+        with:
+          python-version: '3.11.x'
+
+      - name: Run Ontology Demo Example
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_ENDPOINT: ${{ secrets.LLM_ENDPOINT }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_API_VERSION: ${{ secrets.LLM_API_VERSION }}
+          EMBEDDING_MODEL: ${{ secrets.EMBEDDING_MODEL }}
+          EMBEDDING_ENDPOINT: ${{ secrets.EMBEDDING_ENDPOINT }}
+          EMBEDDING_API_KEY: ${{ secrets.EMBEDDING_API_KEY }}
+          EMBEDDING_API_VERSION: ${{ secrets.EMBEDDING_API_VERSION }}
+        run: uv run python ./examples/python/ontology_demo_example.py
+
+  test-agentic-reasoning:
+    name: Run Agentic Reasoning Tests
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@v4
+
+      - name: Cognee Setup
+        uses: ./.github/actions/cognee_setup
+        with:
+          python-version: '3.11.x'
+
+      - name: Run Agentic Reasoning Example
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+          LLM_MODEL: ${{ secrets.LLM_MODEL }}
+          LLM_ENDPOINT: ${{ secrets.LLM_ENDPOINT }}
+          LLM_API_KEY: ${{ secrets.LLM_API_KEY }}
+          LLM_API_VERSION: ${{ secrets.LLM_API_VERSION }}
+          EMBEDDING_MODEL: ${{ secrets.EMBEDDING_MODEL }}
+          EMBEDDING_ENDPOINT: ${{ secrets.EMBEDDING_ENDPOINT }}
+          EMBEDDING_API_KEY: ${{ secrets.EMBEDDING_API_KEY }}
+          EMBEDDING_API_VERSION: ${{ secrets.EMBEDDING_API_VERSION }}
+        run: uv run python ./examples/python/agentic_reasoning_procurement_example.py
+
   test-memify:
     name: Run Memify Example
     runs-on: ubuntu-22.04

.github/workflows/test_different_operating_systems.yml

@@ -9,7 +9,7 @@ on:
       python-versions:
         required: false
         type: string
-        default: '["3.10.x", "3.11.x", "3.12.x"]'
+        default: '["3.10.x", "3.12.x", "3.13.x"]'
     secrets:
       LLM_PROVIDER:
         required: true

.github/workflows/test_suites.yml

@@ -85,7 +85,7 @@ jobs:
     needs: [basic-tests, e2e-tests]
     uses: ./.github/workflows/test_different_operating_systems.yml
     with:
-      python-versions: '["3.10.x", "3.11.x", "3.12.x"]'
+      python-versions: '["3.10.x", "3.11.x", "3.12.x", "3.13.x"]'
     secrets: inherit
 
   # Matrix-based vector database tests

README.md

@@ -71,7 +71,7 @@ Build dynamic memory for Agents and replace RAG using scalable, modular ECL (Ext
 
 ## Get Started
 
-Get started quickly with a Google Colab  <a href="https://colab.research.google.com/drive/1jHbWVypDgCLwjE71GSXhRL3YxYhCZzG1?usp=sharing">notebook</a> , <a href="https://deepnote.com/workspace/cognee-382213d0-0444-4c89-8265-13770e333c02/project/cognee-demo-78ffacb9-5832-4611-bb1a-560386068b30/notebook/Notebook-1-75b24cda566d4c24ab348f7150792601?utm_source=share-modal&utm_medium=product-shared-content&utm_campaign=notebook&utm_content=78ffacb9-5832-4611-bb1a-560386068b30">Deepnote notebook</a> or  <a href="https://github.com/topoteretes/cognee/tree/main/cognee-starter-kit">starter repo</a>
+Get started quickly with a Google Colab  <a href="https://colab.research.google.com/drive/12Vi9zID-M3fpKpKiaqDBvkk98ElkRPWy?usp=sharing">notebook</a> , <a href="https://deepnote.com/workspace/cognee-382213d0-0444-4c89-8265-13770e333c02/project/cognee-demo-78ffacb9-5832-4611-bb1a-560386068b30/notebook/Notebook-1-75b24cda566d4c24ab348f7150792601?utm_source=share-modal&utm_medium=product-shared-content&utm_campaign=notebook&utm_content=78ffacb9-5832-4611-bb1a-560386068b30">Deepnote notebook</a> or  <a href="https://github.com/topoteretes/cognee/tree/main/cognee-starter-kit">starter repo</a>
 
 
 ## About cognee
@@ -224,12 +224,12 @@ We now have a paper you can cite:
 
 ```bibtex
 @misc{markovic2025optimizinginterfaceknowledgegraphs,
-      title={Optimizing the Interface Between Knowledge Graphs and LLMs for Complex Reasoning}, 
+      title={Optimizing the Interface Between Knowledge Graphs and LLMs for Complex Reasoning},
       author={Vasilije Markovic and Lazar Obradovic and Laszlo Hajdu and Jovan Pavlovic},
       year={2025},
       eprint={2505.24478},
       archivePrefix={arXiv},
       primaryClass={cs.AI},
-      url={https://arxiv.org/abs/2505.24478}, 
+      url={https://arxiv.org/abs/2505.24478},
 }
 ```

cognee/api/v1/cognify/cognify.py

@@ -44,6 +44,7 @@ async def cognify(
     graph_model: BaseModel = KnowledgeGraph,
     chunker=TextChunker,
     chunk_size: int = None,
+    chunks_per_batch: int = None,
     config: Config = None,
     vector_db_config: dict = None,
     graph_db_config: dict = None,
@@ -106,6 +107,7 @@ async def cognify(
                    Formula: min(embedding_max_completion_tokens, llm_max_completion_tokens // 2)
                    Default limits: ~512-8192 tokens depending on models.
                    Smaller chunks = more granular but potentially fragmented knowledge.
+        chunks_per_batch: Number of chunks to be processed in a single batch in Cognify tasks.
         vector_db_config: Custom vector database configuration for embeddings storage.
         graph_db_config: Custom graph database configuration for relationship storage.
         run_in_background: If True, starts processing asynchronously and returns immediately.
@@ -210,10 +212,18 @@ async def cognify(
             }
 
     if temporal_cognify:
-        tasks = await get_temporal_tasks(user, chunker, chunk_size)
+        tasks = await get_temporal_tasks(
+            user=user, chunker=chunker, chunk_size=chunk_size, chunks_per_batch=chunks_per_batch
+        )
     else:
         tasks = await get_default_tasks(
-            user, graph_model, chunker, chunk_size, config, custom_prompt
+            user=user,
+            graph_model=graph_model,
+            chunker=chunker,
+            chunk_size=chunk_size,
+            config=config,
+            custom_prompt=custom_prompt,
+            chunks_per_batch=chunks_per_batch,
         )
 
     # By calling get pipeline executor we get a function that will have the run_pipeline run in the background or a function that we will need to wait for
@@ -240,6 +250,7 @@ async def get_default_tasks(  # TODO: Find out a better way to do this (Boris's
     chunk_size: int = None,
     config: Config = None,
     custom_prompt: Optional[str] = None,
+    chunks_per_batch: int = 100,
 ) -> list[Task]:
     if config is None:
         ontology_config = get_ontology_env_config()
@@ -258,6 +269,9 @@ async def get_default_tasks(  # TODO: Find out a better way to do this (Boris's
                 "ontology_config": {"ontology_resolver": get_default_ontology_resolver()}
             }
 
+    if chunks_per_batch is None:
+        chunks_per_batch = 100
+
     default_tasks = [
         Task(classify_documents),
         Task(check_permissions_on_dataset, user=user, permissions=["write"]),
@@ -271,20 +285,20 @@ async def get_default_tasks(  # TODO: Find out a better way to do this (Boris's
             graph_model=graph_model,
             config=config,
             custom_prompt=custom_prompt,
-            task_config={"batch_size": 10},
+            task_config={"batch_size": chunks_per_batch},
         ),  # Generate knowledge graphs from the document chunks.
         Task(
             summarize_text,
-            task_config={"batch_size": 10},
+            task_config={"batch_size": chunks_per_batch},
         ),
-        Task(add_data_points, task_config={"batch_size": 10}),
+        Task(add_data_points, task_config={"batch_size": chunks_per_batch}),
     ]
 
     return default_tasks
 
 
 async def get_temporal_tasks(
-    user: User = None, chunker=TextChunker, chunk_size: int = None
+    user: User = None, chunker=TextChunker, chunk_size: int = None, chunks_per_batch: int = 10
 ) -> list[Task]:
     """
     Builds and returns a list of temporal processing tasks to be executed in sequence.
@@ -301,10 +315,14 @@ async def get_temporal_tasks(
         user (User, optional): The user requesting task execution, used for permission checks.
         chunker (Callable, optional): A text chunking function/class to split documents. Defaults to TextChunker.
         chunk_size (int, optional): Maximum token size per chunk. If not provided, uses system default.
+        chunks_per_batch (int, optional): Number of chunks to process in a single batch in Cognify
 
     Returns:
         list[Task]: A list of Task objects representing the temporal processing pipeline.
     """
+    if chunks_per_batch is None:
+        chunks_per_batch = 10
+
     temporal_tasks = [
         Task(classify_documents),
         Task(check_permissions_on_dataset, user=user, permissions=["write"]),
@@ -313,9 +331,9 @@ async def get_temporal_tasks(
             max_chunk_size=chunk_size or get_max_chunk_tokens(),
             chunker=chunker,
         ),
-        Task(extract_events_and_timestamps, task_config={"chunk_size": 10}),
+        Task(extract_events_and_timestamps, task_config={"batch_size": chunks_per_batch}),
         Task(extract_knowledge_graph_from_events),
-        Task(add_data_points, task_config={"batch_size": 10}),
+        Task(add_data_points, task_config={"batch_size": chunks_per_batch}),
     ]
 
     return temporal_tasks

cognee/api/v1/search/search.py

@@ -1,7 +1,6 @@
 from uuid import UUID
 from typing import Union, Optional, List, Type
 
-from cognee.infrastructure.databases.graph import get_graph_engine
 from cognee.modules.engine.models.node_set import NodeSet
 from cognee.modules.users.models import User
 from cognee.modules.search.types import SearchResult, SearchType, CombinedSearchResult
@@ -9,9 +8,6 @@ from cognee.modules.users.methods import get_default_user
 from cognee.modules.search.methods import search as search_function
 from cognee.modules.data.methods import get_authorized_existing_datasets
 from cognee.modules.data.exceptions import DatasetNotFoundError
-from cognee.shared.logging_utils import get_logger
-
-logger = get_logger()
 
 
 async def search(
@@ -179,13 +175,6 @@ async def search(
         if not datasets:
             raise DatasetNotFoundError(message="No datasets found.")
 
-    graph_engine = await get_graph_engine()
-    is_empty = await graph_engine.is_empty()
-
-    if is_empty:
-        logger.warning("Search attempt on an empty knowledge graph")
-        return []
-
     filtered_search_results = await search_function(
         query_text=query_text,
         query_type=query_type,

cognee/infrastructure/databases/graph/get_graph_engine.py

@@ -162,5 +162,5 @@ def create_graph_engine(
 
     raise EnvironmentError(
         f"Unsupported graph database provider: {graph_database_provider}. "
-        f"Supported providers are: {', '.join(list(supported_databases.keys()) + ['neo4j', 'kuzu', 'kuzu-remote', 'memgraph', 'neptune', 'neptune_analytics'])}"
+        f"Supported providers are: {', '.join(list(supported_databases.keys()) + ['neo4j', 'kuzu', 'kuzu-remote', 'neptune', 'neptune_analytics'])}"
     )

cognee/infrastructure/databases/graph/graph_db_interface.py

@@ -39,11 +39,6 @@ class GraphDBInterface(ABC):
     - get_connections
     """
 
-    @abstractmethod
-    async def is_empty(self) -> bool:
-        logger.warning("is_empty() is not implemented")
-        return True
-
     @abstractmethod
     async def query(self, query: str, params: dict) -> List[Any]:
         """

cognee/infrastructure/databases/graph/kuzu/adapter.py

@@ -197,15 +197,6 @@ class KuzuAdapter(GraphDBInterface):
         except FileNotFoundError:
             logger.warning(f"Kuzu S3 storage file not found: {self.db_path}")
 
-    async def is_empty(self) -> bool:
-        query = """
-        MATCH (n)
-        RETURN true
-        LIMIT 1;
-        """
-        query_result = await self.query(query)
-        return len(query_result) == 0
-
     async def query(self, query: str, params: Optional[dict] = None) -> List[Tuple]:
         """
         Execute a Kuzu query asynchronously with automatic reconnection.

cognee/infrastructure/databases/graph/neo4j_driver/adapter.py

@@ -86,15 +86,6 @@ class Neo4jAdapter(GraphDBInterface):
         async with self.driver.session(database=self.graph_database_name) as session:
             yield session
 
-    async def is_empty(self) -> bool:
-        query = """
-        RETURN EXISTS {
-        MATCH (n)
-        } AS node_exists;
-        """
-        query_result = await self.query(query)
-        return not query_result[0]["node_exists"]
-
     @deadlock_retry()
     async def query(
         self,
@@ -1073,7 +1064,7 @@ class Neo4jAdapter(GraphDBInterface):
         query_nodes = f"""
         MATCH (n)
         WHERE {where_clause}
-        RETURN ID(n) AS id, labels(n) AS labels, properties(n) AS properties
+        RETURN n.id AS id, labels(n) AS labels, properties(n) AS properties
         """
         result_nodes = await self.query(query_nodes)
 
@@ -1088,7 +1079,7 @@ class Neo4jAdapter(GraphDBInterface):
         query_edges = f"""
         MATCH (n)-[r]->(m)
         WHERE {where_clause} AND {where_clause.replace("n.", "m.")}
-        RETURN ID(n) AS source, ID(m) AS target, TYPE(r) AS type, properties(r) AS properties
+        RETURN n.id AS source, n.id AS target, TYPE(r) AS type, properties(r) AS properties
         """
         result_edges = await self.query(query_edges)
 

cognee/infrastructure/databases/vector/embeddings/FastembedEmbeddingEngine.py

@@ -1,8 +1,17 @@
-from cognee.shared.logging_utils import get_logger
+import os
+import logging
 from typing import List, Optional
 from fastembed import TextEmbedding
 import litellm
-import os
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
+
+from cognee.shared.logging_utils import get_logger
 from cognee.infrastructure.databases.vector.embeddings.EmbeddingEngine import EmbeddingEngine
 from cognee.infrastructure.databases.exceptions import EmbeddingException
 from cognee.infrastructure.llm.tokenizer.TikToken import (
@@ -57,6 +66,13 @@ class FastembedEmbeddingEngine(EmbeddingEngine):
             enable_mocking = str(enable_mocking).lower()
         self.mock = enable_mocking in ("true", "1", "yes")
 
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def embed_text(self, text: List[str]) -> List[List[float]]:
         """
         Embed the given text into numerical vectors.

cognee/infrastructure/databases/vector/embeddings/LiteLLMEmbeddingEngine.py

@@ -1,15 +1,21 @@
 import asyncio
+import logging
+
 from cognee.shared.logging_utils import get_logger
 from typing import List, Optional
 import numpy as np
 import math
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
 import litellm
 import os
 from cognee.infrastructure.databases.vector.embeddings.EmbeddingEngine import EmbeddingEngine
 from cognee.infrastructure.databases.exceptions import EmbeddingException
-from cognee.infrastructure.llm.tokenizer.Gemini import (
-    GeminiTokenizer,
-)
 from cognee.infrastructure.llm.tokenizer.HuggingFace import (
     HuggingFaceTokenizer,
 )
@@ -19,10 +25,6 @@ from cognee.infrastructure.llm.tokenizer.Mistral import (
 from cognee.infrastructure.llm.tokenizer.TikToken import (
     TikTokenTokenizer,
 )
-from cognee.infrastructure.databases.vector.embeddings.embedding_rate_limiter import (
-    embedding_rate_limit_async,
-    embedding_sleep_and_retry_async,
-)
 
 litellm.set_verbose = False
 logger = get_logger("LiteLLMEmbeddingEngine")
@@ -76,8 +78,13 @@ class LiteLLMEmbeddingEngine(EmbeddingEngine):
             enable_mocking = str(enable_mocking).lower()
         self.mock = enable_mocking in ("true", "1", "yes")
 
-    @embedding_sleep_and_retry_async()
+    @retry(
-    @embedding_rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def embed_text(self, text: List[str]) -> List[List[float]]:
         """
         Embed a list of text strings into vector representations.

cognee/infrastructure/databases/vector/embeddings/OllamaEmbeddingEngine.py

@@ -3,8 +3,16 @@ from cognee.shared.logging_utils import get_logger
 import aiohttp
 from typing import List, Optional
 import os
-
+import litellm
+import logging
 import aiohttp.http_exceptions
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
 
 from cognee.infrastructure.databases.vector.embeddings.EmbeddingEngine import EmbeddingEngine
 from cognee.infrastructure.llm.tokenizer.HuggingFace import (
@@ -69,7 +77,6 @@ class OllamaEmbeddingEngine(EmbeddingEngine):
             enable_mocking = str(enable_mocking).lower()
         self.mock = enable_mocking in ("true", "1", "yes")
 
-    @embedding_rate_limit_async
     async def embed_text(self, text: List[str]) -> List[List[float]]:
         """
         Generate embedding vectors for a list of text prompts.
@@ -92,7 +99,13 @@ class OllamaEmbeddingEngine(EmbeddingEngine):
         embeddings = await asyncio.gather(*[self._get_embedding(prompt) for prompt in text])
         return embeddings
 
-    @embedding_sleep_and_retry_async()
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def _get_embedding(self, prompt: str) -> List[float]:
         """
         Internal method to call the Ollama embeddings endpoint for a single prompt.

cognee/infrastructure/databases/vector/embeddings/config.py

@@ -24,11 +24,10 @@ class EmbeddingConfig(BaseSettings):
     model_config = SettingsConfigDict(env_file=".env", extra="allow")
 
     def model_post_init(self, __context) -> None:
-        # If embedding batch size is not defined use 2048 as default for OpenAI and 100 for all other embedding models
         if not self.embedding_batch_size and self.embedding_provider.lower() == "openai":
-            self.embedding_batch_size = 2048
+            self.embedding_batch_size = 36
         elif not self.embedding_batch_size:
-            self.embedding_batch_size = 100
+            self.embedding_batch_size = 36
 
     def to_dict(self) -> dict:
         """

cognee/infrastructure/files/utils/guess_file_type.py

@@ -124,6 +124,12 @@ def guess_file_type(file: BinaryIO) -> filetype.Type:
     """
     file_type = filetype.guess(file)
 
+    # If file type could not be determined consider it a plain text file as they don't have magic number encoding
+    if file_type is None:
+        from filetype.types.base import Type
+
+        file_type = Type("text/plain", "txt")
+
     if file_type is None:
         raise FileTypeException(f"Unknown file detected: {file.name}.")
 

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.py

@@ -1,19 +1,24 @@
+import logging
 from typing import Type
 from pydantic import BaseModel
+import litellm
 import instructor
+from cognee.shared.logging_utils import get_logger
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
 
-from cognee.infrastructure.llm.exceptions import MissingSystemPromptPathError
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
-    rate_limit_async,
-    sleep_and_retry_async,
-)
-
-from cognee.infrastructure.llm.LLMGateway import LLMGateway
 from cognee.infrastructure.llm.config import get_llm_config
 
+logger = get_logger()
+
 
 class AnthropicAdapter(LLMInterface):
     """
@@ -35,8 +40,13 @@ class AnthropicAdapter(LLMInterface):
         self.model = model
         self.max_completion_tokens = max_completion_tokens
 
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.py

@@ -12,11 +12,18 @@ from cognee.infrastructure.llm.exceptions import ContentPolicyFilterError
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
+import logging
-    rate_limit_async,
+from cognee.shared.logging_utils import get_logger
-    sleep_and_retry_async,
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
 )
 
+logger = get_logger()
+
 
 class GeminiAdapter(LLMInterface):
     """
@@ -58,8 +65,13 @@ class GeminiAdapter(LLMInterface):
 
         self.aclient = instructor.from_litellm(litellm.acompletion, mode=instructor.Mode.JSON)
 
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py

@@ -12,11 +12,18 @@ from cognee.infrastructure.llm.exceptions import ContentPolicyFilterError
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
+import logging
-    rate_limit_async,
+from cognee.shared.logging_utils import get_logger
-    sleep_and_retry_async,
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
 )
 
+logger = get_logger()
+
 
 class GenericAPIAdapter(LLMInterface):
     """
@@ -58,8 +65,13 @@ class GenericAPIAdapter(LLMInterface):
 
         self.aclient = instructor.from_litellm(litellm.acompletion, mode=instructor.Mode.JSON)
 
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/mistral/adapter.py

@@ -1,20 +1,23 @@
 import litellm
 import instructor
 from pydantic import BaseModel
-from typing import Type, Optional
+from typing import Type
-from litellm import acompletion, JSONSchemaValidationError
+from litellm import JSONSchemaValidationError
 
 from cognee.shared.logging_utils import get_logger
 from cognee.modules.observability.get_observe import get_observe
-from cognee.infrastructure.llm.exceptions import MissingSystemPromptPathError
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
-from cognee.infrastructure.llm.LLMGateway import LLMGateway
 from cognee.infrastructure.llm.config import get_llm_config
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
+
-    rate_limit_async,
+import logging
-    sleep_and_retry_async,
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
 )
 
 logger = get_logger()
@@ -47,8 +50,13 @@ class MistralAdapter(LLMInterface):
             api_key=get_llm_config().llm_api_key,
         )
 
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:
@@ -99,31 +107,3 @@ class MistralAdapter(LLMInterface):
             logger.error(f"Schema validation failed: {str(e)}")
             logger.debug(f"Raw response: {e.raw_response}")
             raise ValueError(f"Response failed schema validation: {str(e)}")
-
-    def show_prompt(self, text_input: str, system_prompt: str) -> str:
-        """
-        Format and display the prompt for a user query.
-
-        Parameters:
-        -----------
-            - text_input (str): Input text from the user to be included in the prompt.
-            - system_prompt (str): The system prompt that will be shown alongside the user input.
-
-        Returns:
-        --------
-            - str: The formatted prompt string combining system prompt and user input.
-        """
-        if not text_input:
-            text_input = "No user input provided."
-        if not system_prompt:
-            raise MissingSystemPromptPathError()
-
-        system_prompt = LLMGateway.read_query_prompt(system_prompt)
-
-        formatted_prompt = (
-            f"""System Prompt:\n{system_prompt}\n\nUser Input:\n{text_input}\n"""
-            if system_prompt
-            else None
-        )
-
-        return formatted_prompt

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/ollama/adapter.py

@@ -1,4 +1,6 @@
 import base64
+import litellm
+import logging
 import instructor
 from typing import Type
 from openai import OpenAI
@@ -7,11 +9,17 @@ from pydantic import BaseModel
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
-    rate_limit_async,
-    sleep_and_retry_async,
-)
 from cognee.infrastructure.files.utils.open_data_file import open_data_file
+from cognee.shared.logging_utils import get_logger
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
+
+logger = get_logger()
 
 
 class OllamaAPIAdapter(LLMInterface):
@@ -47,8 +55,13 @@ class OllamaAPIAdapter(LLMInterface):
             OpenAI(base_url=self.endpoint, api_key=self.api_key), mode=instructor.Mode.JSON
         )
 
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:
@@ -90,7 +103,13 @@ class OllamaAPIAdapter(LLMInterface):
 
         return response
 
-    @rate_limit_async
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def create_transcript(self, input_file: str) -> str:
         """
         Generate an audio transcript from a user query.
@@ -123,7 +142,13 @@ class OllamaAPIAdapter(LLMInterface):
 
         return transcription.text
 
-    @rate_limit_async
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def transcribe_image(self, input_file: str) -> str:
         """
         Transcribe content from an image using base64 encoding.

cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/openai/adapter.py

@@ -7,6 +7,15 @@ from openai import ContentFilterFinishReasonError
 from litellm.exceptions import ContentPolicyViolationError
 from instructor.core import InstructorRetryException
 
+import logging
+from tenacity import (
+    retry,
+    stop_after_delay,
+    wait_exponential_jitter,
+    retry_if_not_exception_type,
+    before_sleep_log,
+)
+
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
@@ -14,19 +23,13 @@ from cognee.infrastructure.llm.exceptions import (
     ContentPolicyFilterError,
 )
 from cognee.infrastructure.files.utils.open_data_file import open_data_file
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.rate_limiter import (
-    rate_limit_async,
-    rate_limit_sync,
-    sleep_and_retry_async,
-    sleep_and_retry_sync,
-)
 from cognee.modules.observability.get_observe import get_observe
 from cognee.shared.logging_utils import get_logger
 
-observe = get_observe()
-
 logger = get_logger()
 
+observe = get_observe()
+
 
 class OpenAIAdapter(LLMInterface):
     """
@@ -97,8 +100,13 @@ class OpenAIAdapter(LLMInterface):
         self.fallback_endpoint = fallback_endpoint
 
     @observe(as_type="generation")
-    @sleep_and_retry_async()
+    @retry(
-    @rate_limit_async
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def acreate_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:
@@ -148,10 +156,7 @@ class OpenAIAdapter(LLMInterface):
             InstructorRetryException,
         ) as e:
             if not (self.fallback_model and self.fallback_api_key):
-                raise ContentPolicyFilterError(
+                raise e
-                    f"The provided input contains content that is not aligned with our content policy: {text_input}"
-                ) from e
-
             try:
                 return await self.aclient.chat.completions.create(
                     model=self.fallback_model,
@@ -186,8 +191,13 @@ class OpenAIAdapter(LLMInterface):
                     ) from error
 
     @observe
-    @sleep_and_retry_sync()
+    @retry(
-    @rate_limit_sync
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     def create_structured_output(
         self, text_input: str, system_prompt: str, response_model: Type[BaseModel]
     ) -> BaseModel:
@@ -231,7 +241,13 @@ class OpenAIAdapter(LLMInterface):
             max_retries=self.MAX_RETRIES,
         )
 
-    @rate_limit_async
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def create_transcript(self, input):
         """
         Generate an audio transcript from a user query.
@@ -263,7 +279,13 @@ class OpenAIAdapter(LLMInterface):
 
         return transcription
 
-    @rate_limit_async
+    @retry(
+        stop=stop_after_delay(128),
+        wait=wait_exponential_jitter(2, 128),
+        retry=retry_if_not_exception_type(litellm.exceptions.NotFoundError),
+        before_sleep=before_sleep_log(logger, logging.DEBUG),
+        reraise=True,
+    )
     async def transcribe_image(self, input) -> BaseModel:
         """
         Generate a transcription of an image from a user query.

cognee/infrastructure/loaders/LoaderEngine.py

@@ -105,7 +105,6 @@ class LoaderEngine:
     async def load_file(
         self,
         file_path: str,
-        file_stream: Optional[Any],
         preferred_loaders: Optional[List[str]] = None,
         **kwargs,
     ):

cognee/infrastructure/loaders/external/advanced_pdf_loader.py

@@ -14,14 +14,6 @@ from cognee.infrastructure.loaders.external.pypdf_loader import PyPdfLoader
 
 logger = get_logger(__name__)
 
-try:
-    from unstructured.partition.pdf import partition_pdf
-except ImportError as e:
-    logger.info(
-        "unstructured[pdf] not installed, can't use AdvancedPdfLoader, will use PyPdfLoader instead."
-    )
-    raise ImportError from e
-
 
 @dataclass
 class _PageBuffer:
@@ -88,6 +80,8 @@ class AdvancedPdfLoader(LoaderInterface):
                 **kwargs,
             }
             # Use partition to extract elements
+            from unstructured.partition.pdf import partition_pdf
+
             elements = partition_pdf(**partition_kwargs)
 
             # Process elements into text content

cognee/modules/pipelines/operations/run_tasks_distributed.py

@@ -88,6 +88,7 @@ async def run_tasks_distributed(
     pipeline_name: str = "unknown_pipeline",
     context: Optional[dict] = None,
     incremental_loading: bool = False,
+    data_per_batch: int = 20,
 ):
     if not user:
         user = await get_default_user()

cognee/tasks/storage/index_data_points.py

@@ -1,6 +1,6 @@
-from cognee.shared.logging_utils import get_logger
+import asyncio
 
-from cognee.infrastructure.databases.exceptions import EmbeddingException
+from cognee.shared.logging_utils import get_logger
 from cognee.infrastructure.databases.vector import get_vector_engine
 from cognee.infrastructure.engine import DataPoint
 
@@ -33,18 +33,23 @@ async def index_data_points(data_points: list[DataPoint]):
             indexed_data_point.metadata["index_fields"] = [field_name]
             index_points[index_name].append(indexed_data_point)
 
-    for index_name_and_field, indexable_points in index_points.items():
+    tasks: list[asyncio.Task] = []
-        first_occurence = index_name_and_field.index("_")
+    batch_size = vector_engine.embedding_engine.get_batch_size()
-        index_name = index_name_and_field[:first_occurence]
+
-        field_name = index_name_and_field[first_occurence + 1 :]
+    for index_name_and_field, points in index_points.items():
-        try:
+        first = index_name_and_field.index("_")
-            # In case the amount of indexable points is too large we need to send them in batches
+        index_name = index_name_and_field[:first]
-            batch_size = vector_engine.embedding_engine.get_batch_size()
+        field_name = index_name_and_field[first + 1 :]
-            for i in range(0, len(indexable_points), batch_size):
+
-                batch = indexable_points[i : i + batch_size]
+        # Create embedding requests per batch to run in parallel later
-                await vector_engine.index_data_points(index_name, field_name, batch)
+        for i in range(0, len(points), batch_size):
-        except EmbeddingException as e:
+            batch = points[i : i + batch_size]
-            logger.warning(f"Failed to index data points for {index_name}.{field_name}: {e}")
+            tasks.append(
+                asyncio.create_task(vector_engine.index_data_points(index_name, field_name, batch))
+            )
+
+    # Run all embedding requests in parallel
+    await asyncio.gather(*tasks)
 
     return data_points
 

cognee/tasks/storage/index_graph_edges.py

@@ -1,3 +1,5 @@
+import asyncio
+
 from cognee.modules.engine.utils.generate_edge_id import generate_edge_id
 from cognee.shared.logging_utils import get_logger
 from collections import Counter
@@ -76,15 +78,20 @@ async def index_graph_edges(
             indexed_data_point.metadata["index_fields"] = [field_name]
             index_points[index_name].append(indexed_data_point)
 
+    # Get maximum batch size for embedding model
+    batch_size = vector_engine.embedding_engine.get_batch_size()
+    tasks: list[asyncio.Task] = []
+
     for index_name, indexable_points in index_points.items():
         index_name, field_name = index_name.split(".")
 
-        # Get maximum batch size for embedding model
+        # Create embedding tasks to run in parallel later
-        batch_size = vector_engine.embedding_engine.get_batch_size()
-        # We save the data in batches of {batch_size} to not put a lot of pressure on the database
         for start in range(0, len(indexable_points), batch_size):
             batch = indexable_points[start : start + batch_size]
 
-            await vector_engine.index_data_points(index_name, field_name, batch)
+            tasks.append(vector_engine.index_data_points(index_name, field_name, batch))
+
+    # Start all embedding tasks and wait for completion
+    await asyncio.gather(*tasks)
 
     return None

cognee/tests/tasks/descriptive_metrics/metrics_test_utils.py

@@ -1,7 +1,6 @@
 from typing import List
 from cognee.infrastructure.engine import DataPoint
 from cognee.tasks.storage.add_data_points import add_data_points
-from cognee.infrastructure.databases.graph.get_graph_engine import create_graph_engine
 import cognee
 from cognee.infrastructure.databases.graph import get_graph_engine
 import json
@@ -64,7 +63,6 @@ async def create_connected_test_graph():
 
 
 async def get_metrics(provider: str, include_optional=True):
-    create_graph_engine.cache_clear()
     cognee.config.set_graph_database_provider(provider)
     graph_engine = await get_graph_engine()
     await graph_engine.delete_graph()

cognee/tests/tasks/descriptive_metrics/neo4j_metrics_test.py

@@ -1,7 +1,12 @@
-from cognee.tests.tasks.descriptive_metrics.metrics_test_utils import assert_metrics
 import asyncio
 
 
+async def main():
+    from cognee.tests.tasks.descriptive_metrics.metrics_test_utils import assert_metrics
+
+    await assert_metrics(provider="neo4j", include_optional=False)
+    await assert_metrics(provider="neo4j", include_optional=True)
+
+
 if __name__ == "__main__":
-    asyncio.run(assert_metrics(provider="neo4j", include_optional=False))
+    asyncio.run(main())
-    asyncio.run(assert_metrics(provider="neo4j", include_optional=True))

cognee/tests/test_kuzu.py

@@ -47,26 +47,10 @@ async def main():
             pathlib.Path(__file__).parent, "test_data/Quantum_computers.txt"
         )
 
-        from cognee.infrastructure.databases.graph import get_graph_engine
-
-        graph_engine = await get_graph_engine()
-
-        is_empty = await graph_engine.is_empty()
-
-        assert is_empty, "Kuzu graph database is not empty"
-
         await cognee.add([explanation_file_path_quantum], dataset_name)
 
-        is_empty = await graph_engine.is_empty()
-
-        assert is_empty, "Kuzu graph database should be empty before cognify"
-
         await cognee.cognify([dataset_name])
 
-        is_empty = await graph_engine.is_empty()
-
-        assert not is_empty, "Kuzu graph database should not be empty"
-
         from cognee.infrastructure.databases.vector import get_vector_engine
 
         vector_engine = get_vector_engine()
@@ -130,10 +114,11 @@ async def main():
         assert not os.path.isdir(data_root_directory), "Local data files are not deleted"
 
         await cognee.prune.prune_system(metadata=True)
+        from cognee.infrastructure.databases.graph import get_graph_engine
 
-        is_empty = await graph_engine.is_empty()
+        graph_engine = await get_graph_engine()
-
+        nodes, edges = await graph_engine.get_graph_data()
-        assert is_empty, "Kuzu graph database is not empty"
+        assert len(nodes) == 0 and len(edges) == 0, "Kuzu graph database is not empty"
 
     finally:
         # Ensure cleanup even if tests fail

cognee/tests/test_memgraph.py

@@ -1,105 +0,0 @@
-import os
-
-import pathlib
-import cognee
-from cognee.infrastructure.files.storage import get_storage_config
-from cognee.modules.search.operations import get_history
-from cognee.modules.users.methods import get_default_user
-from cognee.shared.logging_utils import get_logger
-from cognee.modules.search.types import SearchType
-
-
-logger = get_logger()
-
-
-async def main():
-    cognee.config.set_graph_database_provider("memgraph")
-    data_directory_path = str(
-        pathlib.Path(
-            os.path.join(pathlib.Path(__file__).parent, ".data_storage/test_memgraph")
-        ).resolve()
-    )
-    cognee.config.data_root_directory(data_directory_path)
-    cognee_directory_path = str(
-        pathlib.Path(
-            os.path.join(pathlib.Path(__file__).parent, ".cognee_system/test_memgraph")
-        ).resolve()
-    )
-    cognee.config.system_root_directory(cognee_directory_path)
-
-    await cognee.prune.prune_data()
-    await cognee.prune.prune_system(metadata=True)
-
-    dataset_name = "cs_explanations"
-
-    explanation_file_path_nlp = os.path.join(
-        pathlib.Path(__file__).parent, "test_data/Natural_language_processing.txt"
-    )
-    await cognee.add([explanation_file_path_nlp], dataset_name)
-
-    explanation_file_path_quantum = os.path.join(
-        pathlib.Path(__file__).parent, "test_data/Quantum_computers.txt"
-    )
-
-    await cognee.add([explanation_file_path_quantum], dataset_name)
-
-    await cognee.cognify([dataset_name])
-
-    from cognee.infrastructure.databases.vector import get_vector_engine
-
-    vector_engine = get_vector_engine()
-    random_node = (await vector_engine.search("Entity_name", "Quantum computer"))[0]
-    random_node_name = random_node.payload["text"]
-
-    search_results = await cognee.search(
-        query_type=SearchType.GRAPH_COMPLETION, query_text=random_node_name
-    )
-    assert len(search_results) != 0, "The search results list is empty."
-    print("\n\nExtracted sentences are:\n")
-    for result in search_results:
-        print(f"{result}\n")
-
-    search_results = await cognee.search(query_type=SearchType.CHUNKS, query_text=random_node_name)
-    assert len(search_results) != 0, "The search results list is empty."
-    print("\n\nExtracted chunks are:\n")
-    for result in search_results:
-        print(f"{result}\n")
-
-    search_results = await cognee.search(
-        query_type=SearchType.SUMMARIES, query_text=random_node_name
-    )
-    assert len(search_results) != 0, "Query related summaries don't exist."
-    print("\nExtracted results are:\n")
-    for result in search_results:
-        print(f"{result}\n")
-
-    search_results = await cognee.search(
-        query_type=SearchType.NATURAL_LANGUAGE,
-        query_text=f"Find nodes connected to node with name {random_node_name}",
-    )
-    assert len(search_results) != 0, "Query related natural language don't exist."
-    print("\nExtracted results are:\n")
-    for result in search_results:
-        print(f"{result}\n")
-
-    user = await get_default_user()
-    history = await get_history(user.id)
-
-    assert len(history) == 8, "Search history is not correct."
-
-    await cognee.prune.prune_data()
-    data_root_directory = get_storage_config()["data_root_directory"]
-    assert not os.path.isdir(data_root_directory), "Local data files are not deleted"
-
-    await cognee.prune.prune_system(metadata=True)
-    from cognee.infrastructure.databases.graph import get_graph_engine
-
-    graph_engine = await get_graph_engine()
-    nodes, edges = await graph_engine.get_graph_data()
-    assert len(nodes) == 0 and len(edges) == 0, "Memgraph graph database is not empty"
-
-
-if __name__ == "__main__":
-    import asyncio
-
-    asyncio.run(main())

cognee/tests/test_neo4j.py

@@ -35,14 +35,6 @@ async def main():
     explanation_file_path_nlp = os.path.join(
         pathlib.Path(__file__).parent, "test_data/Natural_language_processing.txt"
     )
-    from cognee.infrastructure.databases.graph import get_graph_engine
-
-    graph_engine = await get_graph_engine()
-
-    is_empty = await graph_engine.is_empty()
-
-    assert is_empty, "Graph has to be empty"
-
     await cognee.add([explanation_file_path_nlp], dataset_name)
 
     explanation_file_path_quantum = os.path.join(
@@ -50,16 +42,9 @@ async def main():
     )
 
     await cognee.add([explanation_file_path_quantum], dataset_name)
-    is_empty = await graph_engine.is_empty()
-
-    assert is_empty, "Graph has to be empty before cognify"
 
     await cognee.cognify([dataset_name])
 
-    is_empty = await graph_engine.is_empty()
-
-    assert not is_empty, "Graph shouldn't be empty"
-
     from cognee.infrastructure.databases.vector import get_vector_engine
 
     vector_engine = get_vector_engine()
@@ -132,8 +117,11 @@ async def main():
     assert not os.path.isdir(data_root_directory), "Local data files are not deleted"
 
     await cognee.prune.prune_system(metadata=True)
-    is_empty = await graph_engine.is_empty()
+    from cognee.infrastructure.databases.graph import get_graph_engine
-    assert is_empty, "Neo4j graph database is not empty"
+
+    graph_engine = await get_graph_engine()
+    nodes, edges = await graph_engine.get_graph_data()
+    assert len(nodes) == 0 and len(edges) == 0, "Neo4j graph database is not empty"
 
 
 if __name__ == "__main__":

cognee/tests/unit/api/test_search.py

@@ -1,21 +0,0 @@
-import pytest
-import cognee
-
-
-@pytest.mark.asyncio
-async def test_empty_search_raises_SearchOnEmptyGraphError_on_empty_graph():
-    await cognee.prune.prune_data()
-    await cognee.prune.prune_system(metadata=True)
-    await cognee.add("Sample input")
-    result = await cognee.search("Sample query")
-    assert result == []
-
-
-@pytest.mark.asyncio
-async def test_empty_search_doesnt_raise_SearchOnEmptyGraphError():
-    await cognee.prune.prune_data()
-    await cognee.prune.prune_system(metadata=True)
-    await cognee.add("Sample input")
-    await cognee.cognify()
-    result = await cognee.search("Sample query")
-    assert result != []

notebooks/neptune-analytics-example.ipynb

@@ -83,16 +83,16 @@
    ]
   },
   {
-   "metadata": {},
    "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "import os\n",
     "import pathlib\n",
     "from cognee import config, add, cognify, search, SearchType, prune, visualize_graph\n",
     "from dotenv import load_dotenv"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
    "cell_type": "markdown",
@@ -106,7 +106,9 @@
   },
   {
    "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
+   "outputs": [],
    "source": [
     "# load environment variables from file .env\n",
     "load_dotenv()\n",
@@ -145,9 +147,7 @@
     "        \"vector_db_url\": f\"neptune-graph://{graph_identifier}\",  # Neptune Analytics endpoint with the format neptune-graph://<GRAPH_ID>\n",
     "    }\n",
     ")"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
    "cell_type": "markdown",
@@ -159,19 +159,19 @@
    ]
   },
   {
-   "metadata": {},
    "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "# Prune data and system metadata before running, only if we want \"fresh\" state.\n",
     "await prune.prune_data()\n",
     "await prune.prune_system(metadata=True)"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
-   "metadata": {},
    "cell_type": "markdown",
+   "metadata": {},
    "source": [
     "## Setup data and cognify\n",
     "\n",
@@ -180,7 +180,9 @@
   },
   {
    "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
+   "outputs": [],
    "source": [
     "# Add sample text to the dataset\n",
     "sample_text_1 = \"\"\"Neptune Analytics is a memory-optimized graph database engine for analytics. With Neptune\n",
@@ -205,9 +207,7 @@
     "\n",
     "# Cognify the text data.\n",
     "await cognify([dataset_name])"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
    "cell_type": "markdown",
@@ -215,14 +215,16 @@
    "source": [
     "## Graph Memory visualization\n",
     "\n",
-    "Initialize Memgraph as a Graph Memory store and save to .artefacts/graph_visualization.html\n",
+    "Initialize Neptune as a Graph Memory store and save to .artefacts/graph_visualization.html\n",
     "\n",
     "![visualization](./neptune_analytics_demo.png)"
    ]
   },
   {
    "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
+   "outputs": [],
    "source": [
     "# Get a graphistry url (Register for a free account at https://www.graphistry.com)\n",
     "# url = await render_graph()\n",
@@ -235,9 +237,7 @@
     "    ).resolve()\n",
     ")\n",
     "await visualize_graph(graph_file_path)"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
    "cell_type": "markdown",
@@ -250,19 +250,19 @@
   },
   {
    "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
+   "outputs": [],
    "source": [
     "# Completion query that uses graph data to form context.\n",
     "graph_completion = await search(query_text=\"What is Neptune Analytics?\", query_type=SearchType.GRAPH_COMPLETION)\n",
     "print(\"\\nGraph completion result is:\")\n",
     "print(graph_completion)"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
-   "metadata": {},
    "cell_type": "markdown",
+   "metadata": {},
    "source": [
     "## SEARCH: RAG Completion\n",
     "\n",
@@ -271,19 +271,19 @@
   },
   {
    "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
+   "outputs": [],
    "source": [
     "# Completion query that uses document chunks to form context.\n",
     "rag_completion = await search(query_text=\"What is Neptune Analytics?\", query_type=SearchType.RAG_COMPLETION)\n",
     "print(\"\\nRAG Completion result is:\")\n",
     "print(rag_completion)"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
-   "metadata": {},
    "cell_type": "markdown",
+   "metadata": {},
    "source": [
     "## SEARCH: Graph Insights\n",
     "\n",
@@ -291,8 +291,10 @@
    ]
   },
   {
-   "metadata": {},
    "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "# Search graph insights\n",
     "insights_results = await search(query_text=\"Neptune Analytics\", query_type=SearchType.GRAPH_COMPLETION)\n",
@@ -302,13 +304,11 @@
     "    tgt_node = result[2].get(\"name\", result[2][\"type\"])\n",
     "    relationship = result[1].get(\"relationship_name\", \"__relationship__\")\n",
     "    print(f\"- {src_node} -[{relationship}]-> {tgt_node}\")"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
-   "metadata": {},
    "cell_type": "markdown",
+   "metadata": {},
    "source": [
     "## SEARCH: Entity Summaries\n",
     "\n",
@@ -316,8 +316,10 @@
    ]
   },
   {
-   "metadata": {},
    "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "# Query all summaries related to query.\n",
     "summaries = await search(query_text=\"Neptune Analytics\", query_type=SearchType.SUMMARIES)\n",
@@ -326,13 +328,11 @@
     "    type = summary[\"type\"]\n",
     "    text = summary[\"text\"]\n",
     "    print(f\"- {type}: {text}\")"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   },
   {
-   "metadata": {},
    "cell_type": "markdown",
+   "metadata": {},
    "source": [
     "## SEARCH: Chunks\n",
     "\n",
@@ -340,8 +340,10 @@
    ]
   },
   {
-   "metadata": {},
    "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
    "source": [
     "chunks = await search(query_text=\"Neptune Analytics\", query_type=SearchType.CHUNKS)\n",
     "print(\"\\nChunk results are:\")\n",
@@ -349,9 +351,7 @@
     "    type = chunk[\"type\"]\n",
     "    text = chunk[\"text\"]\n",
     "    print(f\"- {type}: {text}\")"
-   ],
+   ]
-   "outputs": [],
-   "execution_count": null
   }
  ],
  "metadata": {

poetry.lock

@@ -1,4 +1,4 @@
-# This file is automatically @generated by Poetry 2.1.3 and should not be changed by hand.
+# This file is automatically @generated by Poetry 2.1.2 and should not be changed by hand.
 
 [[package]]
 name = "accelerate"
@@ -6631,7 +6631,7 @@ description = "Fundamental package for array computing in Python"
 optional = false
 python-versions = ">=3.11"
 groups = ["main"]
-markers = "python_version == \"3.12\" or python_full_version == \"3.13.0\""
+markers = "python_version >= \"3.12\""
 files = [
     {file = "numpy-2.3.3-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:0ffc4f5caba7dfcbe944ed674b7eef683c7e94874046454bb79ed7ee0236f59d"},
     {file = "numpy-2.3.3-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:e7e946c7170858a0295f79a60214424caac2ffdb0063d4d79cb681f9aa0aa569"},
@@ -11199,7 +11199,7 @@ description = "Easily download, build, install, upgrade, and uninstall Python pa
 optional = true
 python-versions = ">=3.9"
 groups = ["main"]
-markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\" and (extra == \"docs\" or extra == \"docling\" or extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\") or python_version == \"3.12\" and (extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\" or extra == \"docs\" or extra == \"docling\") or python_full_version == \"3.13.0\" and (extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\" or extra == \"docs\" or extra == \"docling\") or extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\""
+markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\" and (extra == \"docs\" or extra == \"docling\" or extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\") or python_version >= \"3.12\" and (extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\" or extra == \"docs\" or extra == \"docling\") or extra == \"notebook\" or extra == \"dev\" or extra == \"llama-index\" or extra == \"deepeval\" or extra == \"dlt\""
 files = [
     {file = "setuptools-80.9.0-py3-none-any.whl", hash = "sha256:062d34222ad13e0cc312a4c02d73f059e86a4acbfbdea8f8f76b28c99f306922"},
     {file = "setuptools-80.9.0.tar.gz", hash = "sha256:f36b47402ecde768dbfafc46e8e4207b4360c654f1f3bb84475f0a28628fb19c"},
@@ -13506,7 +13506,7 @@ dev = ["coverage", "deptry", "gitpython", "mkdocs-material", "mkdocs-minify-plug
 distributed = ["modal"]
 dlt = ["dlt"]
 docling = ["docling", "transformers"]
-docs = ["unstructured"]
+docs = ["lxml", "unstructured"]
 evals = ["gdown", "matplotlib", "pandas", "plotly", "scikit-learn"]
 graphiti = ["graphiti-core"]
 groq = ["groq"]
@@ -13527,5 +13527,5 @@ scraping = ["APScheduler", "beautifulsoup4", "lxml", "playwright", "protego", "t
 
 [metadata]
 lock-version = "2.1"
-python-versions = ">=3.10,<=3.13"
+python-versions = ">=3.10,<3.14"
-content-hash = "8d8172ac8ddc3c30ca79a1677ecf2a28897d52c0a564d8fb5646c8565c313a0f"
+content-hash = "bcab5420339473ec08b89cde588899b60999762fb8ca9a011240d47ea86198e3"

pyproject.toml

@@ -7,7 +7,7 @@ authors = [
     { name = "Vasilije Markovic" },
     { name = "Boris Arzentar" },
 ]
-requires-python = ">=3.10,<=3.13"
+requires-python = ">=3.10,<3.14"
 readme = "README.md"
 license = "Apache-2.0"
 classifiers = [
@@ -56,6 +56,7 @@ dependencies = [
     "gunicorn>=20.1.0,<24",
     "websockets>=15.0.1,<16.0.0",
     "mistralai>=1.9.10",
+    "tenacity>=9.0.0",
 ]
 
 [project.optional-dependencies]
@@ -64,14 +65,16 @@ api=[]
 distributed = [
     "modal>=1.0.5,<2.0.0",
 ]
+
 scraping = [
-    "tavily-python>=0.7.0",
+    "tavily-python>=0.7.12",
     "beautifulsoup4>=4.13.1",
     "playwright>=1.9.0",
-    "lxml>=4.9.3,<5.0.0",
+    "lxml>=4.9.3",
     "protego>=0.1",
     "APScheduler>=3.10.0,<=3.11.0"
 ]
+
 neo4j = ["neo4j>=5.28.0,<6"]
 neptune = ["langchain_aws>=0.2.22"]
 postgres = [
@@ -101,7 +104,7 @@ chromadb = [
     "chromadb>=0.6,<0.7",
     "pypika==0.48.9",
 ]
-docs = ["unstructured[csv, doc, docx, epub, md, odt, org, ppt, pptx, rst, rtf, tsv, xlsx, pdf]>=0.18.1,<19"]
+docs = ["lxml<6.0.0", "unstructured[csv, doc, docx, epub, md, odt, org, ppt, pptx, rst, rtf, tsv, xlsx, pdf]>=0.18.1,<19"]
 codegraph = [
     "fastembed<=0.6.0 ; python_version < '3.13'",
     "transformers>=4.46.3,<5",