cognee 0.5.0.dev1__tar.gz → 0.5.1.dev0__tar.gz

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/.env.template

@@ -91,6 +91,15 @@ DB_NAME=cognee_db
 #DB_USERNAME=cognee
 #DB_PASSWORD=cognee
 
+# -- Advanced: Custom database connection arguments (optional) ---------------
+# Pass additional connection parameters as JSON. Useful for SSL, timeouts, etc.
+# Examples:
+#   For PostgreSQL with SSL:
+#     DATABASE_CONNECT_ARGS='{"sslmode": "require", "connect_timeout": 10}'
+#   For SQLite with custom timeout:
+#     DATABASE_CONNECT_ARGS='{"timeout": 60}'
+#DATABASE_CONNECT_ARGS='{}'
+
 ################################################################################
 # 🕸️ Graph Database settings
 ################################################################################

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cognee
-Version: 0.5.0.dev1
+Version: 0.5.1.dev0
 Summary: Cognee - is a library for enriching LLM context with a semantic layer for better understanding and reasoning.
 Project-URL: Homepage, https://www.cognee.ai
 Project-URL: Repository, https://github.com/topoteretes/cognee

cognee-0.5.1.dev0/alembic/versions/a1b2c3d4e5f6_add_label_column_to_data.py

@@ -0,0 +1,38 @@
+"""Add label column to data table
+
+Revision ID: a1b2c3d4e5f6
+Revises: 211ab850ef3d
+Create Date: 2025-11-17 17:54:32.123456
+
+"""
+
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+# revision identifiers, used by Alembic.
+revision: str = "a1b2c3d4e5f6"
+down_revision: Union[str, None] = "46a6ce2bd2b2"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def _get_column(inspector, table, name, schema=None):
+    for col in inspector.get_columns(table, schema=schema):
+        if col["name"] == name:
+            return col
+    return None
+
+
+def upgrade() -> None:
+    conn = op.get_bind()
+    insp = sa.inspect(conn)
+
+    label_column = _get_column(insp, "data", "label")
+    if not label_column:
+        op.add_column("data", sa.Column("label", sa.String(), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("data", "label")

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/api/v1/add/add.py

@@ -10,13 +10,14 @@ from cognee.modules.pipelines.layers.reset_dataset_pipeline_run_status import (
 )
 from cognee.modules.engine.operations.setup import setup
 from cognee.tasks.ingestion import ingest_data, resolve_data_directories
+from cognee.tasks.ingestion.data_item import DataItem
 from cognee.shared.logging_utils import get_logger
 
 logger = get_logger()
 
 
 async def add(
-    data: Union[BinaryIO, list[BinaryIO], str, list[str]],
+    data: Union[BinaryIO, list[BinaryIO], str, list[str], DataItem, list[DataItem]],
     dataset_name: str = "main_dataset",
     user: User = None,
     node_set: Optional[List[str]] = None,

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/api/v1/datasets/routers/get_datasets_router.py

@@ -44,6 +44,7 @@ class DatasetDTO(OutDTO):
 class DataDTO(OutDTO):
     id: UUID
     name: str
+    label: Optional[str] = None
     created_at: datetime
     updated_at: Optional[datetime] = None
     extension: str

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/api/v1/memify/routers/get_memify_router.py

@@ -90,6 +90,7 @@ def get_memify_router() -> APIRouter:
                 dataset=payload.dataset_id if payload.dataset_id else payload.dataset_name,
                 node_name=payload.node_name,
                 user=user,
+                run_in_background=payload.run_in_background,
             )
 
             if isinstance(memify_run, PipelineRunErrored):

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/databases/relational/config.py

@@ -1,4 +1,5 @@
 import os
+import json
 import pydantic
 from typing import Union
 from functools import lru_cache
@@ -19,6 +20,7 @@ class RelationalConfig(BaseSettings):
     db_username: Union[str, None] = None  # "cognee"
     db_password: Union[str, None] = None  # "cognee"
     db_provider: str = "sqlite"
+    database_connect_args: Union[str, None] = None
 
     model_config = SettingsConfigDict(env_file=".env", extra="allow")
 
@@ -30,6 +32,17 @@ class RelationalConfig(BaseSettings):
             databases_directory_path = os.path.join(base_config.system_root_directory, "databases")
             self.db_path = databases_directory_path
 
+        # Parse database_connect_args if provided as JSON string
+        if self.database_connect_args and isinstance(self.database_connect_args, str):
+            try:
+                parsed_args = json.loads(self.database_connect_args)
+                if isinstance(parsed_args, dict):
+                    self.database_connect_args = parsed_args
+                else:
+                    self.database_connect_args = {}
+            except json.JSONDecodeError:
+                self.database_connect_args = {}
+
         return self
 
     def to_dict(self) -> dict:
@@ -40,7 +53,8 @@ class RelationalConfig(BaseSettings):
         --------
 
             - dict: A dictionary containing database configuration settings including db_path,
-              db_name, db_host, db_port, db_username, db_password, and db_provider.
+              db_name, db_host, db_port, db_username, db_password, db_provider, and
+              database_connect_args.
         """
         return {
             "db_path": self.db_path,
@@ -50,6 +64,7 @@ class RelationalConfig(BaseSettings):
             "db_username": self.db_username,
             "db_password": self.db_password,
             "db_provider": self.db_provider,
+            "database_connect_args": self.database_connect_args,
         }
 
 

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/databases/relational/create_relational_engine.py

@@ -1,3 +1,4 @@
+from sqlalchemy import URL
 from .sqlalchemy.SqlAlchemyAdapter import SQLAlchemyAdapter
 from functools import lru_cache
 
@@ -11,6 +12,7 @@ def create_relational_engine(
     db_username: str,
     db_password: str,
     db_provider: str,
+    database_connect_args: dict = None,
 ):
     """
     Create a relational database engine based on the specified parameters.
@@ -29,6 +31,7 @@ def create_relational_engine(
         - db_password (str): The password for database authentication, required for
           PostgreSQL.
         - db_provider (str): The type of database provider (e.g., 'sqlite' or 'postgres').
+        - database_connect_args (dict, optional): Database driver connection arguments.
 
     Returns:
     --------
@@ -43,12 +46,19 @@ def create_relational_engine(
             # Test if asyncpg is available
             import asyncpg
 
-            connection_string = (
-                f"postgresql+asyncpg://{db_username}:{db_password}@{db_host}:{db_port}/{db_name}"
+            # Handle special characters in username and password like # or @
+            connection_string = URL.create(
+                "postgresql+asyncpg",
+                username=db_username,
+                password=db_password,
+                host=db_host,
+                port=int(db_port),
+                database=db_name,
             )
+
         except ImportError:
             raise ImportError(
                 "PostgreSQL dependencies are not installed. Please install with 'pip install cognee\"[postgres]\"' or 'pip install cognee\"[postgres-binary]\"' to use PostgreSQL functionality."
             )
 
-    return SQLAlchemyAdapter(connection_string)
+    return SQLAlchemyAdapter(connection_string, connect_args=database_connect_args)

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/databases/relational/sqlalchemy/SqlAlchemyAdapter.py

@@ -29,10 +29,31 @@ class SQLAlchemyAdapter:
     functions.
     """
 
-    def __init__(self, connection_string: str):
+    def __init__(self, connection_string: str, connect_args: dict = None):
+        """
+        Initialize the SQLAlchemy adapter with connection settings.
+
+        Parameters:
+        -----------
+            connection_string (str): The database connection string (e.g., 'sqlite:///path/to/db'
+                or 'postgresql://user:pass@host:port/db').
+            connect_args (dict, optional): Database driver connection arguments.
+                Configuration is loaded from RelationalConfig.database_connect_args, which reads
+                from the DATABASE_CONNECT_ARGS environment variable.
+
+                Examples:
+                    PostgreSQL with SSL:
+                        DATABASE_CONNECT_ARGS='{"sslmode": "require", "connect_timeout": 10}'
+
+                    SQLite with custom timeout:
+                        DATABASE_CONNECT_ARGS='{"timeout": 60}'
+        """
         self.db_path: str = None
         self.db_uri: str = connection_string
 
+        # Use provided connect_args (already parsed from config)
+        final_connect_args = connect_args or {}
+
         if "sqlite" in connection_string:
             [prefix, db_path] = connection_string.split("///")
             self.db_path = db_path
@@ -53,7 +74,7 @@ class SQLAlchemyAdapter:
             self.engine = create_async_engine(
                 connection_string,
                 poolclass=NullPool,
-                connect_args={"timeout": 30},
+                connect_args={**{"timeout": 30}, **final_connect_args},
             )
         else:
             self.engine = create_async_engine(
@@ -63,6 +84,7 @@ class SQLAlchemyAdapter:
                 pool_recycle=280,
                 pool_pre_ping=True,
                 pool_timeout=280,
+                connect_args=final_connect_args,
             )
 
         self.sessionmaker = async_sessionmaker(bind=self.engine, expire_on_commit=False)

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/databases/vector/create_vector_engine.py

@@ -1,3 +1,5 @@
+from sqlalchemy import URL
+
 from .supported_databases import supported_databases
 from .embeddings import get_embedding_engine
 from cognee.infrastructure.databases.graph.config import get_graph_context_config
@@ -66,8 +68,13 @@ def create_vector_engine(
         if not (db_host and db_port and db_name and db_username and db_password):
             raise EnvironmentError("Missing requred pgvector credentials!")
 
-        connection_string: str = (
-            f"postgresql+asyncpg://{db_username}:{db_password}@{db_host}:{db_port}/{db_name}"
+        connection_string = URL.create(
+            "postgresql+asyncpg",
+            username=db_username,
+            password=db_password,
+            host=db_host,
+            port=int(db_port),
+            database=db_name,
         )
 
         try:

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/llm/LLMGateway.py

@@ -37,19 +37,6 @@ class LLMGateway:
                 **kwargs,
             )
 
-    @staticmethod
-    def create_structured_output(
-        text_input: str, system_prompt: str, response_model: Type[BaseModel]
-    ) -> BaseModel:
-        from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.get_llm_client import (
-            get_llm_client,
-        )
-
-        llm_client = get_llm_client()
-        return llm_client.create_structured_output(
-            text_input=text_input, system_prompt=system_prompt, response_model=response_model
-        )
-
     @staticmethod
     def create_transcript(input) -> Coroutine:
         from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.get_llm_client import (

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/anthropic/adapter.py

@@ -3,7 +3,9 @@ from typing import Type
 from pydantic import BaseModel
 import litellm
 import instructor
+import anthropic
 from cognee.shared.logging_utils import get_logger
+from cognee.modules.observability.get_observe import get_observe
 from tenacity import (
     retry,
     stop_after_delay,
@@ -12,38 +14,41 @@ from tenacity import (
     before_sleep_log,
 )
 
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
-    LLMInterface,
+from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.generic_llm_api.adapter import (
+    GenericAPIAdapter,
 )
 from cognee.shared.rate_limiting import llm_rate_limiter_context_manager
 from cognee.infrastructure.llm.config import get_llm_config
 
 logger = get_logger()
+observe = get_observe()
 
 
-class AnthropicAdapter(LLMInterface):
+class AnthropicAdapter(GenericAPIAdapter):
     """
     Adapter for interfacing with the Anthropic API, enabling structured output generation
     and prompt display.
     """
 
-    name = "Anthropic"
-    model: str
     default_instructor_mode = "anthropic_tools"
 
-    def __init__(self, max_completion_tokens: int, model: str = None, instructor_mode: str = None):
-        import anthropic
-
+    def __init__(
+        self, api_key: str, model: str, max_completion_tokens: int, instructor_mode: str = None
+    ):
+        super().__init__(
+            api_key=api_key,
+            model=model,
+            max_completion_tokens=max_completion_tokens,
+            name="Anthropic",
+        )
         self.instructor_mode = instructor_mode if instructor_mode else self.default_instructor_mode
 
         self.aclient = instructor.patch(
-            create=anthropic.AsyncAnthropic(api_key=get_llm_config().llm_api_key).messages.create,
+            create=anthropic.AsyncAnthropic(api_key=self.api_key).messages.create,
             mode=instructor.Mode(self.instructor_mode),
         )
 
-        self.model = model
-        self.max_completion_tokens = max_completion_tokens
-
+    @observe(as_type="generation")
     @retry(
         stop=stop_after_delay(128),
         wait=wait_exponential_jitter(8, 128),

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/gemini/adapter.py

@@ -1,4 +1,4 @@
-"""Adapter for Generic API LLM provider API"""
+"""Adapter for Gemini API LLM provider"""
 
 import litellm
 import instructor
@@ -8,13 +8,9 @@ from openai import ContentFilterFinishReasonError
 from litellm.exceptions import ContentPolicyViolationError
 from instructor.core import InstructorRetryException
 
-from cognee.infrastructure.llm.exceptions import ContentPolicyFilterError
-from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
-    LLMInterface,
-)
 import logging
 from cognee.shared.rate_limiting import llm_rate_limiter_context_manager
-from cognee.shared.logging_utils import get_logger
+
 from tenacity import (
     retry,
     stop_after_delay,
@@ -23,55 +19,65 @@ from tenacity import (
     before_sleep_log,
 )
 
+from cognee.infrastructure.llm.exceptions import ContentPolicyFilterError
+from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.generic_llm_api.adapter import (
+    GenericAPIAdapter,
+)
+from cognee.shared.logging_utils import get_logger
+from cognee.modules.observability.get_observe import get_observe
+
 logger = get_logger()
+observe = get_observe()
 
 
-class GeminiAdapter(LLMInterface):
+class GeminiAdapter(GenericAPIAdapter):
     """
     Adapter for Gemini API LLM provider.
 
     This class initializes the API adapter with necessary credentials and configurations for
     interacting with the gemini LLM models. It provides methods for creating structured outputs
-    based on user input and system prompts.
+    based on user input and system prompts, as well as multimodal processing capabilities.
 
     Public methods:
-    - acreate_structured_output(text_input: str, system_prompt: str, response_model:
-    Type[BaseModel]) -> BaseModel
+    - acreate_structured_output(text_input: str, system_prompt: str, response_model: Type[BaseModel]) -> BaseModel
+    - create_transcript(input) -> BaseModel: Transcribe audio files to text
+    - transcribe_image(input) -> BaseModel: Inherited from GenericAPIAdapter
     """
 
-    name: str
-    model: str
-    api_key: str
     default_instructor_mode = "json_mode"
 
     def __init__(
         self,
-        endpoint,
         api_key: str,
         model: str,
-        api_version: str,
         max_completion_tokens: int,
+        endpoint: str = None,
+        api_version: str = None,
+        transcription_model: str = None,
         instructor_mode: str = None,
         fallback_model: str = None,
         fallback_api_key: str = None,
         fallback_endpoint: str = None,
     ):
-        self.model = model
-        self.api_key = api_key
-        self.endpoint = endpoint
-        self.api_version = api_version
-        self.max_completion_tokens = max_completion_tokens
-
-        self.fallback_model = fallback_model
-        self.fallback_api_key = fallback_api_key
-        self.fallback_endpoint = fallback_endpoint
-
+        super().__init__(
+            api_key=api_key,
+            model=model,
+            max_completion_tokens=max_completion_tokens,
+            name="Gemini",
+            endpoint=endpoint,
+            api_version=api_version,
+            transcription_model=transcription_model,
+            fallback_model=fallback_model,
+            fallback_api_key=fallback_api_key,
+            fallback_endpoint=fallback_endpoint,
+        )
         self.instructor_mode = instructor_mode if instructor_mode else self.default_instructor_mode
 
         self.aclient = instructor.from_litellm(
             litellm.acompletion, mode=instructor.Mode(self.instructor_mode)
         )
 
+    @observe(as_type="generation")
     @retry(
         stop=stop_after_delay(128),
         wait=wait_exponential_jitter(8, 128),

{cognee-0.5.0.dev1 → cognee-0.5.1.dev0}/cognee/infrastructure/llm/structured_output_framework/litellm_instructor/llm/generic_llm_api/adapter.py

@@ -1,8 +1,10 @@
 """Adapter for Generic API LLM provider API"""
 
+import base64
+import mimetypes
 import litellm
 import instructor
-from typing import Type
+from typing import Type, Optional
 from pydantic import BaseModel
 from openai import ContentFilterFinishReasonError
 from litellm.exceptions import ContentPolicyViolationError
@@ -12,6 +14,8 @@ from cognee.infrastructure.llm.exceptions import ContentPolicyFilterError
 from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.llm_interface import (
     LLMInterface,
 )
+from cognee.infrastructure.files.utils.open_data_file import open_data_file
+from cognee.modules.observability.get_observe import get_observe
 import logging
 from cognee.shared.rate_limiting import llm_rate_limiter_context_manager
 from cognee.shared.logging_utils import get_logger
@@ -23,7 +27,12 @@ from tenacity import (
     before_sleep_log,
 )
 
+from cognee.infrastructure.llm.structured_output_framework.litellm_instructor.llm.types import (
+    TranscriptionReturnType,
+)
+
 logger = get_logger()
+observe = get_observe()
 
 
 class GenericAPIAdapter(LLMInterface):
@@ -39,18 +48,19 @@ class GenericAPIAdapter(LLMInterface):
     Type[BaseModel]) -> BaseModel
     """
 
-    name: str
-    model: str
-    api_key: str
+    MAX_RETRIES = 5
     default_instructor_mode = "json_mode"
 
     def __init__(
         self,
-        endpoint,
         api_key: str,
         model: str,
-        name: str,
         max_completion_tokens: int,
+        name: str,
+        endpoint: str = None,
+        api_version: str = None,
+        transcription_model: str = None,
+        image_transcribe_model: str = None,
         instructor_mode: str = None,
         fallback_model: str = None,
         fallback_api_key: str = None,
@@ -59,9 +69,11 @@ class GenericAPIAdapter(LLMInterface):
         self.name = name
         self.model = model
         self.api_key = api_key
+        self.api_version = api_version
         self.endpoint = endpoint
         self.max_completion_tokens = max_completion_tokens
-
+        self.transcription_model = transcription_model or model
+        self.image_transcribe_model = image_transcribe_model or model
         self.fallback_model = fallback_model
         self.fallback_api_key = fallback_api_key
         self.fallback_endpoint = fallback_endpoint
@@ -72,6 +84,7 @@ class GenericAPIAdapter(LLMInterface):
             litellm.acompletion, mode=instructor.Mode(self.instructor_mode)
         )
 
+    @observe(as_type="generation")
     @retry(
         stop=stop_after_delay(128),
         wait=wait_exponential_jitter(8, 128),
@@ -173,3 +186,115 @@ class GenericAPIAdapter(LLMInterface):
                     raise ContentPolicyFilterError(
                         f"The provided input contains content that is not aligned with our content policy: {text_input}"
                     ) from error
+
+    @observe(as_type="transcription")
+    @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) -> TranscriptionReturnType:
+        """
+        Generate an audio transcript from a user query.
+
+        This method creates a transcript from the specified audio file, raising a
+        FileNotFoundError if the file does not exist. The audio file is processed and the
+        transcription is retrieved from the API.
+
+        Parameters:
+        -----------
+            - input: The path to the audio file that needs to be transcribed.
+
+        Returns:
+        --------
+            The generated transcription of the audio file.
+        """
+        async with open_data_file(input, mode="rb") as audio_file:
+            encoded_string = base64.b64encode(audio_file.read()).decode("utf-8")
+        mime_type, _ = mimetypes.guess_type(input)
+        if not mime_type or not mime_type.startswith("audio/"):
+            raise ValueError(
+                f"Could not determine MIME type for audio file: {input}. Is the extension correct?"
+            )
+        response = await litellm.acompletion(
+            model=self.transcription_model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": [
+                        {
+                            "type": "file",
+                            "file": {"file_data": f"data:{mime_type};base64,{encoded_string}"},
+                        },
+                        {"type": "text", "text": "Transcribe the following audio precisely."},
+                    ],
+                }
+            ],
+            api_key=self.api_key,
+            api_version=self.api_version,
+            max_completion_tokens=self.max_completion_tokens,
+            api_base=self.endpoint,
+            max_retries=self.MAX_RETRIES,
+        )
+
+        return TranscriptionReturnType(response.choices[0].message.content, response)
+
+    @observe(as_type="transcribe_image")
+    @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.
+
+        This method encodes the image and sends a request to the API to obtain a
+        description of the contents of the image.
+
+        Parameters:
+        -----------
+            - input: The path to the image file that needs to be transcribed.
+
+        Returns:
+        --------
+            - BaseModel: A structured output generated by the model, returned as an instance of
+              BaseModel.
+        """
+        async with open_data_file(input, mode="rb") as image_file:
+            encoded_image = base64.b64encode(image_file.read()).decode("utf-8")
+        mime_type, _ = mimetypes.guess_type(input)
+        if not mime_type or not mime_type.startswith("image/"):
+            raise ValueError(
+                f"Could not determine MIME type for image file: {input}. Is the extension correct?"
+            )
+        response = await litellm.acompletion(
+            model=self.image_transcribe_model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": [
+                        {
+                            "type": "text",
+                            "text": "What's in this image?",
+                        },
+                        {
+                            "type": "image_url",
+                            "image_url": {
+                                "url": f"data:{mime_type};base64,{encoded_image}",
+                            },
+                        },
+                    ],
+                }
+            ],
+            api_key=self.api_key,
+            api_base=self.endpoint,
+            api_version=self.api_version,
+            max_completion_tokens=300,
+            max_retries=self.MAX_RETRIES,
+        )
+        return response