cognee 0.4.0__py3-none-any.whl → 0.5.0__py3-none-any.whl

cognee/context_global_variables.py

@@ -4,7 +4,10 @@ from typing import Union
 from uuid import UUID
 
 from cognee.base_config import get_base_config
+from cognee.infrastructure.databases.vector.config import get_vectordb_config
+from cognee.infrastructure.databases.graph.config import get_graph_config
 from cognee.infrastructure.databases.utils import get_or_create_dataset_database
+from cognee.infrastructure.databases.utils import resolve_dataset_database_connection_info
 from cognee.infrastructure.files.storage.config import file_storage_config
 from cognee.modules.users.methods import get_user
 
@@ -19,6 +22,67 @@ async def set_session_user_context_variable(user):
     session_user.set(user)
 
 
+def multi_user_support_possible():
+    graph_db_config = get_graph_config()
+    vector_db_config = get_vectordb_config()
+
+    graph_handler = graph_db_config.graph_dataset_database_handler
+    vector_handler = vector_db_config.vector_dataset_database_handler
+    from cognee.infrastructure.databases.dataset_database_handler import (
+        supported_dataset_database_handlers,
+    )
+
+    if graph_handler not in supported_dataset_database_handlers:
+        raise EnvironmentError(
+            "Unsupported graph dataset to database handler configured. Cannot add support for multi-user access control mode. Please use a supported graph dataset to database handler or set the environment variables ENABLE_BACKEND_ACCESS_CONTROL to false to switch off multi-user access control mode.\n"
+            f"Selected graph dataset to database handler: {graph_handler}\n"
+            f"Supported dataset to database handlers: {list(supported_dataset_database_handlers.keys())}\n"
+        )
+
+    if vector_handler not in supported_dataset_database_handlers:
+        raise EnvironmentError(
+            "Unsupported vector dataset to database handler configured. Cannot add support for multi-user access control mode. Please use a supported vector dataset to database handler or set the environment variables ENABLE_BACKEND_ACCESS_CONTROL to false to switch off multi-user access control mode.\n"
+            f"Selected vector dataset to database handler: {vector_handler}\n"
+            f"Supported dataset to database handlers: {list(supported_dataset_database_handlers.keys())}\n"
+        )
+
+    if (
+        supported_dataset_database_handlers[graph_handler]["handler_provider"]
+        != graph_db_config.graph_database_provider
+    ):
+        raise EnvironmentError(
+            "The selected graph dataset to database handler does not work with the configured graph database provider. Cannot add support for multi-user access control mode. Please use a supported graph dataset to database handler or set the environment variables ENABLE_BACKEND_ACCESS_CONTROL to false to switch off multi-user access control mode.\n"
+            f"Selected graph database provider: {graph_db_config.graph_database_provider}\n"
+            f"Selected graph dataset to database handler: {graph_handler}\n"
+            f"Supported dataset to database handlers: {list(supported_dataset_database_handlers.keys())}\n"
+        )
+
+    if (
+        supported_dataset_database_handlers[vector_handler]["handler_provider"]
+        != vector_db_config.vector_db_provider
+    ):
+        raise EnvironmentError(
+            "The selected vector dataset to database handler does not work with the configured vector database provider. Cannot add support for multi-user access control mode. Please use a supported vector dataset to database handler or set the environment variables ENABLE_BACKEND_ACCESS_CONTROL to false to switch off multi-user access control mode.\n"
+            f"Selected vector database provider: {vector_db_config.vector_db_provider}\n"
+            f"Selected vector dataset to database handler: {vector_handler}\n"
+            f"Supported dataset to database handlers: {list(supported_dataset_database_handlers.keys())}\n"
+        )
+
+    return True
+
+
+def backend_access_control_enabled():
+    backend_access_control = os.environ.get("ENABLE_BACKEND_ACCESS_CONTROL", None)
+    if backend_access_control is None:
+        # If backend access control is not defined in environment variables,
+        # enable it by default if graph and vector DBs can support it, otherwise disable it
+        return multi_user_support_possible()
+    elif backend_access_control.lower() == "true":
+        # If enabled, ensure that the current graph and vector DBs can support it
+        return multi_user_support_possible()
+    return False
+
+
 async def set_database_global_context_variables(dataset: Union[str, UUID], user_id: UUID):
     """
     If backend access control is enabled this function will ensure all datasets have their own databases,
@@ -38,16 +102,17 @@ async def set_database_global_context_variables(dataset: Union[str, UUID], user_
 
     """
 
-    base_config = get_base_config()
-
-    if not os.getenv("ENABLE_BACKEND_ACCESS_CONTROL", "false").lower() == "true":
+    if not backend_access_control_enabled():
         return
 
     user = await get_user(user_id)
 
     # To ensure permissions are enforced properly all datasets will have their own databases
     dataset_database = await get_or_create_dataset_database(dataset, user)
+    # Ensure that all connection info is resolved properly
+    dataset_database = await resolve_dataset_database_connection_info(dataset_database)
 
+    base_config = get_base_config()
     data_root_directory = os.path.join(
         base_config.data_root_directory, str(user.tenant_id or user.id)
     )
@@ -56,19 +121,31 @@ async def set_database_global_context_variables(dataset: Union[str, UUID], user_
     )
 
     # Set vector and graph database configuration based on dataset database information
+    # TODO: Add better handling of vector and graph config accross Cognee.
+    #  LRU_CACHE takes into account order of inputs, if order of inputs is changed it will be registered as a new DB adapter
     vector_config = {
-        "vector_db_url": os.path.join(
-            databases_directory_path, dataset_database.vector_database_name
-        ),
-        "vector_db_key": "",
-        "vector_db_provider": "lancedb",
+        "vector_db_provider": dataset_database.vector_database_provider,
+        "vector_db_url": dataset_database.vector_database_url,
+        "vector_db_key": dataset_database.vector_database_key,
+        "vector_db_name": dataset_database.vector_database_name,
     }
 
     graph_config = {
-        "graph_database_provider": "kuzu",
+        "graph_database_provider": dataset_database.graph_database_provider,
+        "graph_database_url": dataset_database.graph_database_url,
+        "graph_database_name": dataset_database.graph_database_name,
+        "graph_database_key": dataset_database.graph_database_key,
         "graph_file_path": os.path.join(
             databases_directory_path, dataset_database.graph_database_name
         ),
+        "graph_database_username": dataset_database.graph_database_connection_info.get(
+            "graph_database_username", ""
+        ),
+        "graph_database_password": dataset_database.graph_database_connection_info.get(
+            "graph_database_password", ""
+        ),
+        "graph_dataset_database_handler": "",
+        "graph_database_port": "",
     }
 
     storage_config = {

cognee/eval_framework/Dockerfile

@@ -0,0 +1,29 @@
+FROM python:3.11-slim
+
+# Set environment variables
+ENV PIP_NO_CACHE_DIR=true
+ENV PATH="${PATH}:/root/.poetry/bin"
+ENV PYTHONPATH=/app
+ENV SKIP_MIGRATIONS=true
+
+# System dependencies
+RUN apt-get update && apt-get install -y \
+    gcc \
+    libpq-dev \
+    git \
+    curl \
+    build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+COPY pyproject.toml poetry.lock README.md /app/
+
+RUN pip install poetry
+
+RUN poetry config virtualenvs.create false
+
+RUN poetry install --extras distributed --extras evals --extras deepeval --no-root
+
+COPY cognee/ /app/cognee
+COPY distributed/ /app/distributed

cognee/eval_framework/answer_generation/answer_generation_executor.py

@@ -35,6 +35,16 @@ class AnswerGeneratorExecutor:
             retrieval_context = await retriever.get_context(query_text)
             search_results = await retriever.get_completion(query_text, retrieval_context)
 
+            ############
+            #:TODO This is a quick fix until we don't structure retriever results properly but lets not leave it like this...this is needed now due to the changed combined retriever structure..
+            if isinstance(retrieval_context, list):
+                retrieval_context = await retriever.convert_retrieved_objects_to_context(
+                    triplets=retrieval_context
+                )
+
+            if isinstance(search_results, str):
+                search_results = [search_results]
+            #############
             answer = {
                 "question": query_text,
                 "answer": search_results[0],

cognee/eval_framework/answer_generation/run_question_answering_module.py

@@ -35,7 +35,7 @@ async def create_and_insert_answers_table(questions_payload):
 
 
 async def run_question_answering(
-    params: dict, system_prompt="answer_simple_question.txt", top_k: Optional[int] = None
+    params: dict, system_prompt="answer_simple_question_benchmark.txt", top_k: Optional[int] = None
 ) -> List[dict]:
     if params.get("answering_questions"):
         logger.info("Question answering started...")

cognee/eval_framework/corpus_builder/task_getters/get_cascade_graph_tasks.py

@@ -8,7 +8,6 @@ from cognee.modules.users.models import User
 from cognee.shared.data_models import KnowledgeGraph
 from cognee.shared.utils import send_telemetry
 from cognee.tasks.documents import (
-    check_permissions_on_dataset,
     classify_documents,
     extract_chunks_from_documents,
 )
@@ -31,7 +30,6 @@ async def get_cascade_graph_tasks(
         cognee_config = get_cognify_config()
         default_tasks = [
             Task(classify_documents),
-            Task(check_permissions_on_dataset, user=user, permissions=["write"]),
             Task(
                 extract_chunks_from_documents, max_chunk_tokens=get_max_chunk_tokens()
             ),  # Extract text chunks based on the document type.

cognee/eval_framework/corpus_builder/task_getters/get_default_tasks_by_indices.py

@@ -30,8 +30,8 @@ async def get_no_summary_tasks(
     ontology_file_path=None,
 ) -> List[Task]:
     """Returns default tasks without summarization tasks."""
-    # Get base tasks (0=classify, 1=check_permissions, 2=extract_chunks)
-    base_tasks = await get_default_tasks_by_indices([0, 1, 2], chunk_size, chunker)
+    # Get base tasks (0=classify, 1=extract_chunks)
+    base_tasks = await get_default_tasks_by_indices([0, 1], chunk_size, chunker)
 
     ontology_adapter = RDFLibOntologyResolver(ontology_file=ontology_file_path)
 
@@ -51,8 +51,8 @@ async def get_just_chunks_tasks(
     chunk_size: int = None, chunker=TextChunker, user=None
 ) -> List[Task]:
     """Returns default tasks with only chunk extraction and data points addition."""
-    # Get base tasks (0=classify, 1=check_permissions, 2=extract_chunks)
-    base_tasks = await get_default_tasks_by_indices([0, 1, 2], chunk_size, chunker)
+    # Get base tasks (0=classify, 1=extract_chunks)
+    base_tasks = await get_default_tasks_by_indices([0, 1], chunk_size, chunker)
 
     add_data_points_task = Task(add_data_points, task_config={"batch_size": 10})
 

cognee/eval_framework/eval_config.py

@@ -14,7 +14,7 @@ class EvalConfig(BaseSettings):
 
     # Question answering params
     answering_questions: bool = True
-    qa_engine: str = "cognee_completion"  # Options: 'cognee_completion' or 'cognee_graph_completion' or 'cognee_graph_completion_cot' or 'cognee_graph_completion_context_extension'
+    qa_engine: str = "cognee_graph_completion"  # Options: 'cognee_completion' or 'cognee_graph_completion' or 'cognee_graph_completion_cot' or 'cognee_graph_completion_context_extension'
 
     # Evaluation params
     evaluating_answers: bool = True
@@ -25,7 +25,7 @@ class EvalConfig(BaseSettings):
         "EM",
         "f1",
     ]  # Use only 'correctness' for DirectLLM
-    deepeval_model: str = "gpt-5-mini"
+    deepeval_model: str = "gpt-4o-mini"
 
     # Metrics params
     calculate_metrics: bool = True

cognee/eval_framework/modal_run_eval.py

@@ -2,7 +2,6 @@ import modal
 import os
 import asyncio
 import datetime
-import hashlib
 import json
 from cognee.shared.logging_utils import get_logger
 from cognee.eval_framework.eval_config import EvalConfig
@@ -10,6 +9,9 @@ from cognee.eval_framework.corpus_builder.run_corpus_builder import run_corpus_b
 from cognee.eval_framework.answer_generation.run_question_answering_module import (
     run_question_answering,
 )
+import pathlib
+from os import path
+from modal import Image
 from cognee.eval_framework.evaluation.run_evaluation_module import run_evaluation
 from cognee.eval_framework.metrics_dashboard import create_dashboard
 
@@ -38,22 +40,19 @@ def read_and_combine_metrics(eval_params: dict) -> dict:
 
 app = modal.App("modal-run-eval")
 
-image = (
-    modal.Image.from_dockerfile(path="Dockerfile_modal", force_build=False)
-    .copy_local_file("pyproject.toml", "pyproject.toml")
-    .copy_local_file("poetry.lock", "poetry.lock")
-    .env(
-        {
-            "ENV": os.getenv("ENV"),
-            "LLM_API_KEY": os.getenv("LLM_API_KEY"),
-            "OPENAI_API_KEY": os.getenv("OPENAI_API_KEY"),
-        }
-    )
-    .pip_install("protobuf", "h2", "deepeval", "gdown", "plotly")
-)
+image = Image.from_dockerfile(
+    path=pathlib.Path(path.join(path.dirname(__file__), "Dockerfile")).resolve(),
+    force_build=False,
+).add_local_python_source("cognee")
 
 
-@app.function(image=image, concurrency_limit=10, timeout=86400, volumes={"/data": vol})
+@app.function(
+    image=image,
+    max_containers=10,
+    timeout=86400,
+    volumes={"/data": vol},
+    secrets=[modal.Secret.from_name("eval_secrets")],
+)
 async def modal_run_eval(eval_params=None):
     """Runs evaluation pipeline and returns combined metrics results."""
     if eval_params is None:
@@ -105,18 +104,7 @@ async def main():
     configs = [
         EvalConfig(
             task_getter_type="Default",
-            number_of_samples_in_corpus=10,
-            benchmark="HotPotQA",
-            qa_engine="cognee_graph_completion",
-            building_corpus_from_scratch=True,
-            answering_questions=True,
-            evaluating_answers=True,
-            calculate_metrics=True,
-            dashboard=True,
-        ),
-        EvalConfig(
-            task_getter_type="Default",
-            number_of_samples_in_corpus=10,
+            number_of_samples_in_corpus=25,
             benchmark="TwoWikiMultiHop",
             qa_engine="cognee_graph_completion",
             building_corpus_from_scratch=True,
@@ -127,7 +115,7 @@ async def main():
         ),
         EvalConfig(
             task_getter_type="Default",
-            number_of_samples_in_corpus=10,
+            number_of_samples_in_corpus=25,
             benchmark="Musique",
             qa_engine="cognee_graph_completion",
             building_corpus_from_scratch=True,

cognee/infrastructure/databases/cache/config.py

@@ -1,6 +1,6 @@
 from pydantic_settings import BaseSettings, SettingsConfigDict
 from functools import lru_cache
-from typing import Optional
+from typing import Optional, Literal
 
 
 class CacheConfig(BaseSettings):
@@ -15,6 +15,7 @@ class CacheConfig(BaseSettings):
     - agentic_lock_timeout: Maximum time (in seconds) to wait for the lock release.
     """
 
+    cache_backend: Literal["redis", "fs"] = "fs"
     caching: bool = False
     shared_kuzu_lock: bool = False
     cache_host: str = "localhost"
@@ -28,6 +29,7 @@ class CacheConfig(BaseSettings):
 
     def to_dict(self) -> dict:
         return {
+            "cache_backend": self.cache_backend,
             "caching": self.caching,
             "shared_kuzu_lock": self.shared_kuzu_lock,
             "cache_host": self.cache_host,

cognee/infrastructure/databases/cache/fscache/FsCacheAdapter.py

@@ -0,0 +1,151 @@
+import asyncio
+import json
+import os
+from datetime import datetime
+import time
+import threading
+import diskcache as dc
+
+from cognee.infrastructure.databases.cache.cache_db_interface import CacheDBInterface
+from cognee.infrastructure.databases.exceptions.exceptions import (
+    CacheConnectionError,
+    SharedKuzuLockRequiresRedisError,
+)
+from cognee.infrastructure.files.storage.get_storage_config import get_storage_config
+from cognee.shared.logging_utils import get_logger
+
+logger = get_logger("FSCacheAdapter")
+
+
+class FSCacheAdapter(CacheDBInterface):
+    def __init__(self):
+        default_key = "sessions_db"
+
+        storage_config = get_storage_config()
+        data_root_directory = storage_config["data_root_directory"]
+        cache_directory = os.path.join(data_root_directory, ".cognee_fs_cache", default_key)
+        os.makedirs(cache_directory, exist_ok=True)
+        self.cache = dc.Cache(directory=cache_directory)
+        self.cache.expire()
+
+        logger.debug(f"FSCacheAdapter initialized with cache directory: {cache_directory}")
+
+    def acquire_lock(self):
+        """Lock acquisition is not available for filesystem cache backend."""
+        message = "Shared Kuzu lock requires Redis cache backend."
+        logger.error(message)
+        raise SharedKuzuLockRequiresRedisError()
+
+    def release_lock(self):
+        """Lock release is not available for filesystem cache backend."""
+        message = "Shared Kuzu lock requires Redis cache backend."
+        logger.error(message)
+        raise SharedKuzuLockRequiresRedisError()
+
+    async def add_qa(
+        self,
+        user_id: str,
+        session_id: str,
+        question: str,
+        context: str,
+        answer: str,
+        ttl: int | None = 86400,
+    ):
+        try:
+            session_key = f"agent_sessions:{user_id}:{session_id}"
+
+            qa_entry = {
+                "time": datetime.utcnow().isoformat(),
+                "question": question,
+                "context": context,
+                "answer": answer,
+            }
+
+            existing_value = self.cache.get(session_key)
+            if existing_value is not None:
+                value: list = json.loads(existing_value)
+                value.append(qa_entry)
+            else:
+                value = [qa_entry]
+
+            self.cache.set(session_key, json.dumps(value), expire=ttl)
+        except Exception as e:
+            error_msg = f"Unexpected error while adding Q&A to diskcache: {str(e)}"
+            logger.error(error_msg)
+            raise CacheConnectionError(error_msg) from e
+
+    async def get_latest_qa(self, user_id: str, session_id: str, last_n: int = 5):
+        session_key = f"agent_sessions:{user_id}:{session_id}"
+        value = self.cache.get(session_key)
+        if value is None:
+            return None
+        entries = json.loads(value)
+        return entries[-last_n:] if len(entries) > last_n else entries
+
+    async def get_all_qas(self, user_id: str, session_id: str):
+        session_key = f"agent_sessions:{user_id}:{session_id}"
+        value = self.cache.get(session_key)
+        if value is None:
+            return None
+        return json.loads(value)
+
+    async def close(self):
+        if self.cache is not None:
+            self.cache.expire()
+            self.cache.close()
+
+
+async def main():
+    adapter = FSCacheAdapter()
+    session_id = "demo_session"
+    user_id = "demo_user_id"
+
+    print("\nAdding sample Q/A pairs...")
+    await adapter.add_qa(
+        user_id,
+        session_id,
+        "What is Redis?",
+        "Basic DB context",
+        "Redis is an in-memory data store.",
+    )
+    await adapter.add_qa(
+        user_id,
+        session_id,
+        "Who created Redis?",
+        "Historical context",
+        "Salvatore Sanfilippo (antirez).",
+    )
+
+    print("\nLatest QA:")
+    latest = await adapter.get_latest_qa(user_id, session_id)
+    print(json.dumps(latest, indent=2))
+
+    print("\nLast 2 QAs:")
+    last_two = await adapter.get_latest_qa(user_id, session_id, last_n=2)
+    print(json.dumps(last_two, indent=2))
+
+    session_id = "session_expire_demo"
+
+    await adapter.add_qa(
+        user_id,
+        session_id,
+        "What is Redis?",
+        "Database context",
+        "Redis is an in-memory data store.",
+    )
+
+    await adapter.add_qa(
+        user_id,
+        session_id,
+        "Who created Redis?",
+        "History context",
+        "Salvatore Sanfilippo (antirez).",
+    )
+
+    print(await adapter.get_all_qas(user_id, session_id))
+
+    await adapter.close()
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

cognee/infrastructure/databases/cache/get_cache_engine.py

@@ -1,9 +1,11 @@
 """Factory to get the appropriate cache coordination engine (e.g., Redis)."""
 
 from functools import lru_cache
+import os
 from typing import Optional
 from cognee.infrastructure.databases.cache.config import get_cache_config
 from cognee.infrastructure.databases.cache.cache_db_interface import CacheDBInterface
+from cognee.infrastructure.databases.cache.fscache.FsCacheAdapter import FSCacheAdapter
 
 config = get_cache_config()
 
@@ -33,20 +35,28 @@ def create_cache_engine(
 
     Returns:
     --------
-    - CacheDBInterface: An instance of the appropriate cache adapter. :TODO: Now we support only Redis. later if we add more here we can split the logic
+    - CacheDBInterface: An instance of the appropriate cache adapter.
     """
     if config.caching:
         from cognee.infrastructure.databases.cache.redis.RedisAdapter import RedisAdapter
 
-        return RedisAdapter(
-            host=cache_host,
-            port=cache_port,
-            username=cache_username,
-            password=cache_password,
-            lock_name=lock_key,
-            timeout=agentic_lock_expire,
-            blocking_timeout=agentic_lock_timeout,
-        )
+        if config.cache_backend == "redis":
+            return RedisAdapter(
+                host=cache_host,
+                port=cache_port,
+                username=cache_username,
+                password=cache_password,
+                lock_name=lock_key,
+                timeout=agentic_lock_expire,
+                blocking_timeout=agentic_lock_timeout,
+            )
+        elif config.cache_backend == "fs":
+            return FSCacheAdapter()
+        else:
+            raise ValueError(
+                f"Unsupported cache backend: '{config.cache_backend}'. "
+                f"Supported backends are: 'redis', 'fs'"
+            )
     else:
         return None
 

cognee/infrastructure/databases/dataset_database_handler/__init__.py

@@ -0,0 +1,3 @@
+from .dataset_database_handler_interface import DatasetDatabaseHandlerInterface
+from .supported_dataset_database_handlers import supported_dataset_database_handlers
+from .use_dataset_database_handler import use_dataset_database_handler

cognee/infrastructure/databases/dataset_database_handler/dataset_database_handler_interface.py

@@ -0,0 +1,80 @@
+from typing import Optional
+from uuid import UUID
+from abc import ABC, abstractmethod
+
+from cognee.modules.users.models.User import User
+from cognee.modules.users.models.DatasetDatabase import DatasetDatabase
+
+
+class DatasetDatabaseHandlerInterface(ABC):
+    @classmethod
+    @abstractmethod
+    async def create_dataset(cls, dataset_id: Optional[UUID], user: Optional[User]) -> dict:
+        """
+        Return a dictionary with database connection/resolution info for a graph or vector database for the given dataset.
+        Function can auto handle deploying of the actual database if needed, but is not necessary.
+        Only providing connection info is sufficient, this info will be mapped when trying to connect to the provided dataset in the future.
+        Needed for Cognee multi-tenant/multi-user and backend access control support.
+
+        Dictionary returned from this function will be used to create a DatasetDatabase row in the relational database.
+        From which internal mapping of dataset -> database connection info will be done.
+
+        The returned dictionary is stored verbatim in the relational database and is later passed to
+        resolve_dataset_connection_info() at connection time. For safe credential handling, prefer
+        returning only references to secrets or role identifiers, not plaintext credentials.
+
+        Each dataset needs to map to a unique graph or vector database when backend access control is enabled to facilitate a separation of concern for data.
+
+        Args:
+            dataset_id: UUID of the dataset if needed by the database creation logic
+            user: User object if needed by the database creation logic
+        Returns:
+            dict: Connection info for the created graph or vector database instance.
+        """
+        pass
+
+    @classmethod
+    async def resolve_dataset_connection_info(
+        cls, dataset_database: DatasetDatabase
+    ) -> DatasetDatabase:
+        """
+        Resolve runtime connection details for a dataset’s backing graph/vector database.
+        Function is intended to be overwritten to implement custom logic for resolving connection info.
+
+        This method is invoked right before the application opens a connection for a given dataset.
+        It receives the DatasetDatabase row that was persisted when create_dataset() ran and must
+        return a modified instance of DatasetDatabase with concrete connection parameters that the client/driver can use.
+        Do not update these new DatasetDatabase values in the relational database to avoid storing secure credentials.
+
+        In case of separate graph and vector database handlers, each handler should implement its own logic for resolving
+        connection info and only change parameters related to its appropriate database, the resolution function will then
+        be called one after another with the updated DatasetDatabase value from the previous function as the input.
+
+        Typical behavior:
+        - If the DatasetDatabase row already contains raw connection fields (e.g., host/port/db/user/password
+        or api_url/api_key), return them as-is.
+        - If the row stores only references (e.g., secret IDs, vault paths, cloud resource ARNs/IDs, IAM
+        roles, SSO tokens), resolve those references by calling the appropriate secret manager or provider
+        API to obtain short-lived credentials and assemble the final connection DatasetDatabase object.
+        - Do not persist any resolved or decrypted secrets back to the relational database. Return them only
+        to the caller.
+
+        Args:
+            dataset_database: DatasetDatabase row from the relational database
+        Returns:
+            DatasetDatabase: Updated instance with resolved connection info
+        """
+        return dataset_database
+
+    @classmethod
+    @abstractmethod
+    async def delete_dataset(cls, dataset_database: DatasetDatabase) -> None:
+        """
+        Delete the graph or vector database for the given dataset.
+        Function should auto handle deleting of the actual database or send a request to the proper service to delete/mark the database as not needed for the given dataset.
+        Needed for maintaining a database for Cognee multi-tenant/multi-user and backend access control.
+
+        Args:
+            dataset_database: DatasetDatabase row containing connection/resolution info for the graph or vector database to delete.
+        """
+        pass

cognee/infrastructure/databases/dataset_database_handler/supported_dataset_database_handlers.py

@@ -0,0 +1,18 @@
+from cognee.infrastructure.databases.graph.neo4j_driver.Neo4jAuraDevDatasetDatabaseHandler import (
+    Neo4jAuraDevDatasetDatabaseHandler,
+)
+from cognee.infrastructure.databases.vector.lancedb.LanceDBDatasetDatabaseHandler import (
+    LanceDBDatasetDatabaseHandler,
+)
+from cognee.infrastructure.databases.graph.kuzu.KuzuDatasetDatabaseHandler import (
+    KuzuDatasetDatabaseHandler,
+)
+
+supported_dataset_database_handlers = {
+    "neo4j_aura_dev": {
+        "handler_instance": Neo4jAuraDevDatasetDatabaseHandler,
+        "handler_provider": "neo4j",
+    },
+    "lancedb": {"handler_instance": LanceDBDatasetDatabaseHandler, "handler_provider": "lancedb"},
+    "kuzu": {"handler_instance": KuzuDatasetDatabaseHandler, "handler_provider": "kuzu"},
+}