graphiti-core 0.11.3__tar.gz → 0.11.4__tar.gz

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: graphiti-core
-Version: 0.11.3
+Version: 0.11.4
 Summary: A temporal graph building library
 License: Apache-2.0
 Author: Paul Paliychuk

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/graphiti.py

@@ -396,7 +396,7 @@ class Graphiti:
                 episode.content = ''
 
             await add_nodes_and_edges_bulk(
-                self.driver, [episode], episodic_edges, nodes, entity_edges
+                self.driver, [episode], episodic_edges, hydrated_nodes, entity_edges
             )
 
             # Update any communities

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/anthropic_client.py

@@ -28,7 +28,7 @@ from pydantic import BaseModel, ValidationError
 
 from ..prompts.models import Message
 from .client import LLMClient
-from .config import DEFAULT_MAX_TOKENS, LLMConfig
+from .config import DEFAULT_MAX_TOKENS, LLMConfig, ModelSize
 from .errors import RateLimitError, RefusalError
 
 logger = logging.getLogger(__name__)
@@ -173,6 +173,7 @@ class AnthropicClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         """
         Generate a response from the Anthropic LLM using tool-based approach for all requests.
@@ -263,6 +264,7 @@ class AnthropicClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         """
         Generate a response from the LLM.
@@ -289,7 +291,9 @@ class AnthropicClient(LLMClient):
 
         while retry_count <= max_retries:
             try:
-                response = await self._generate_response(messages, response_model, max_tokens)
+                response = await self._generate_response(
+                    messages, response_model, max_tokens, model_size
+                )
 
                 # If we have a response_model, attempt to validate the response
                 if response_model is not None:

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/client.py

@@ -26,7 +26,7 @@ from pydantic import BaseModel
 from tenacity import retry, retry_if_exception, stop_after_attempt, wait_random_exponential
 
 from ..prompts.models import Message
-from .config import DEFAULT_MAX_TOKENS, LLMConfig
+from .config import DEFAULT_MAX_TOKENS, LLMConfig, ModelSize
 from .errors import RateLimitError
 
 DEFAULT_TEMPERATURE = 0
@@ -55,6 +55,7 @@ class LLMClient(ABC):
 
         self.config = config
         self.model = config.model
+        self.small_model = config.small_model
         self.temperature = config.temperature
         self.max_tokens = config.max_tokens
         self.cache_enabled = cache
@@ -102,9 +103,10 @@ class LLMClient(ABC):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         try:
-            return await self._generate_response(messages, response_model, max_tokens)
+            return await self._generate_response(messages, response_model, max_tokens, model_size)
         except (httpx.HTTPStatusError, RateLimitError) as e:
             raise e
 
@@ -114,6 +116,7 @@ class LLMClient(ABC):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         pass
 
@@ -128,6 +131,7 @@ class LLMClient(ABC):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         if max_tokens is None:
             max_tokens = self.max_tokens
@@ -154,7 +158,9 @@ class LLMClient(ABC):
         for message in messages:
             message.content = self._clean_input(message.content)
 
-        response = await self._generate_response_with_retry(messages, response_model, max_tokens)
+        response = await self._generate_response_with_retry(
+            messages, response_model, max_tokens, model_size
+        )
 
         if self.cache_enabled and self.cache_dir is not None:
             cache_key = self._get_cache_key(messages)

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/config.py

@@ -14,10 +14,17 @@ See the License for the specific language governing permissions and
 limitations under the License.
 """
 
+from enum import Enum
+
 DEFAULT_MAX_TOKENS = 8192
 DEFAULT_TEMPERATURE = 0
 
 
+class ModelSize(Enum):
+    small = 'small'
+    medium = 'medium'
+
+
 class LLMConfig:
     """
     Configuration class for the Language Learning Model (LLM).
@@ -34,6 +41,7 @@ class LLMConfig:
         base_url: str | None = None,
         temperature: float = DEFAULT_TEMPERATURE,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        small_model: str | None = None,
     ):
         """
         Initialize the LLMConfig with the provided parameters.
@@ -43,15 +51,18 @@ class LLMConfig:
                                                 This is required for making authorized requests.
 
                 model (str, optional): The specific LLM model to use for generating responses.
-                                                                Defaults to "gpt-4.1-mini", which appears to be a custom model name.
-                                                                Common values might include "gpt-3.5-turbo" or "gpt-4".
+                                                                Defaults to "gpt-4.1-mini".
 
                 base_url (str, optional): The base URL of the LLM API service.
                                                                         Defaults to "https://api.openai.com", which is OpenAI's standard API endpoint.
                                                                         This can be changed if using a different provider or a custom endpoint.
+
+                small_model (str, optional): The specific LLM model to use for generating responses of simpler prompts.
+                                                                Defaults to "gpt-4.1-nano".
         """
         self.base_url = base_url
         self.api_key = api_key
         self.model = model
+        self.small_model = small_model
         self.temperature = temperature
         self.max_tokens = max_tokens

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/gemini_client.py

@@ -24,7 +24,7 @@ from pydantic import BaseModel
 
 from ..prompts.models import Message
 from .client import LLMClient
-from .config import DEFAULT_MAX_TOKENS, LLMConfig
+from .config import DEFAULT_MAX_TOKENS, LLMConfig, ModelSize
 from .errors import RateLimitError
 
 logger = logging.getLogger(__name__)
@@ -82,6 +82,7 @@ class GeminiClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         """
         Generate a response from the Gemini language model.
@@ -167,6 +168,7 @@ class GeminiClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         """
         Generate a response from the Gemini language model.
@@ -185,5 +187,8 @@ class GeminiClient(LLMClient):
 
         # Call the internal _generate_response method
         return await self._generate_response(
-            messages=messages, response_model=response_model, max_tokens=max_tokens
+            messages=messages,
+            response_model=response_model,
+            max_tokens=max_tokens,
+            model_size=model_size,
         )

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/groq_client.py

@@ -25,7 +25,7 @@ from pydantic import BaseModel
 
 from ..prompts.models import Message
 from .client import LLMClient
-from .config import LLMConfig
+from .config import LLMConfig, ModelSize
 from .errors import RateLimitError
 
 logger = logging.getLogger(__name__)
@@ -49,6 +49,7 @@ class GroqClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         msgs: list[ChatCompletionMessageParam] = []
         for m in messages:

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/openai_client.py

@@ -25,12 +25,13 @@ from pydantic import BaseModel
 
 from ..prompts.models import Message
 from .client import MULTILINGUAL_EXTRACTION_RESPONSES, LLMClient
-from .config import DEFAULT_MAX_TOKENS, LLMConfig
+from .config import DEFAULT_MAX_TOKENS, LLMConfig, ModelSize
 from .errors import RateLimitError, RefusalError
 
 logger = logging.getLogger(__name__)
 
 DEFAULT_MODEL = 'gpt-4.1-mini'
+DEFAULT_SMALL_MODEL = 'gpt-4.1-nano'
 
 
 class OpenAIClient(LLMClient):
@@ -94,6 +95,7 @@ class OpenAIClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         openai_messages: list[ChatCompletionMessageParam] = []
         for m in messages:
@@ -103,8 +105,13 @@ class OpenAIClient(LLMClient):
             elif m.role == 'system':
                 openai_messages.append({'role': 'system', 'content': m.content})
         try:
+            if model_size == ModelSize.small:
+                model = self.small_model or DEFAULT_SMALL_MODEL
+            else:
+                model = self.model or DEFAULT_MODEL
+
             response = await self.client.beta.chat.completions.parse(
-                model=self.model or DEFAULT_MODEL,
+                model=model,
                 messages=openai_messages,
                 temperature=self.temperature,
                 max_tokens=max_tokens or self.max_tokens,
@@ -132,6 +139,7 @@ class OpenAIClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         if max_tokens is None:
             max_tokens = self.max_tokens
@@ -144,7 +152,9 @@ class OpenAIClient(LLMClient):
 
         while retry_count <= self.MAX_RETRIES:
             try:
-                response = await self._generate_response(messages, response_model, max_tokens)
+                response = await self._generate_response(
+                    messages, response_model, max_tokens, model_size
+                )
                 return response
             except (RateLimitError, RefusalError):
                 # These errors should not trigger retries

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/llm_client/openai_generic_client.py

@@ -26,7 +26,7 @@ from pydantic import BaseModel
 
 from ..prompts.models import Message
 from .client import MULTILINGUAL_EXTRACTION_RESPONSES, LLMClient
-from .config import DEFAULT_MAX_TOKENS, LLMConfig
+from .config import DEFAULT_MAX_TOKENS, LLMConfig, ModelSize
 from .errors import RateLimitError, RefusalError
 
 logger = logging.getLogger(__name__)
@@ -89,6 +89,7 @@ class OpenAIGenericClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int = DEFAULT_MAX_TOKENS,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         openai_messages: list[ChatCompletionMessageParam] = []
         for m in messages:
@@ -118,6 +119,7 @@ class OpenAIGenericClient(LLMClient):
         messages: list[Message],
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
+        model_size: ModelSize = ModelSize.medium,
     ) -> dict[str, typing.Any]:
         if max_tokens is None:
             max_tokens = self.max_tokens
@@ -139,7 +141,7 @@ class OpenAIGenericClient(LLMClient):
         while retry_count <= self.MAX_RETRIES:
             try:
                 response = await self._generate_response(
-                    messages, response_model, max_tokens=max_tokens
+                    messages, response_model, max_tokens=max_tokens, model_size=model_size
                 )
                 return response
             except (RateLimitError, RefusalError):

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/prompts/dedupe_nodes.py

@@ -27,6 +27,7 @@ class NodeDuplicate(BaseModel):
         ...,
         description='id of the duplicate node. If no duplicate nodes are found, default to -1.',
     )
+    name: str = Field(..., description='Name of the entity.')
 
 
 class Prompt(Protocol):
@@ -43,7 +44,7 @@ def node(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
             role='system',
-            content='You are a helpful assistant that de-duplicates nodes from node lists.',
+            content='You are a helpful assistant that de-duplicates entities from entity lists.',
         ),
         Message(
             role='user',
@@ -54,25 +55,33 @@ def node(context: dict[str, Any]) -> list[Message]:
         <CURRENT MESSAGE>
         {context['episode_content']}
         </CURRENT MESSAGE>
+        <NEW ENTITY>
+        {json.dumps(context['extracted_node'], indent=2)}
+        </NEW ENTITY>
+        <ENTITY TYPE DESCRIPTION>
+        {json.dumps(context['entity_type_description'], indent=2)}
+        </ENTITY TYPE DESCRIPTION>
 
-        <EXISTING NODES>
+        <EXISTING ENTITIES>
         {json.dumps(context['existing_nodes'], indent=2)}
-        </EXISTING NODES>
+        </EXISTING ENTITIES>
+        
+        Given the above EXISTING ENTITIES and their attributes, MESSAGE, and PREVIOUS MESSAGES; Determine if the NEW ENTITY extracted from the conversation
+        is a duplicate entity of one of the EXISTING ENTITIES.
         
-        Given the above EXISTING NODES and their attributes, MESSAGE, and PREVIOUS MESSAGES; Determine if the NEW NODE extracted from the conversation
-        is a duplicate entity of one of the EXISTING NODES.
+        The ENTITY TYPE DESCRIPTION gives more insight into what the entity type means for the NEW ENTITY.
 
-        <NEW NODE>
-        {json.dumps(context['extracted_node'], indent=2)}
-        </NEW NODE>
         Task:
-        If the NEW NODE is a duplicate of any node in EXISTING NODES, set duplicate_node_id to the
-        id of the EXISTING NODE that is the duplicate. If the NEW NODE is not a duplicate of any of the EXISTING NODES,
-        duplicate_node_id should be set to -1.
+        If the NEW ENTITY represents a duplicate entity of any entity in EXISTING ENTITIES, set duplicate_entity_id to the
+        id of the EXISTING ENTITY that is the duplicate. If the NEW ENTITY is not a duplicate of any of the EXISTING ENTITIES,
+        duplicate_entity_id should be set to -1.
+        
+        Also return the most complete name for the entity.
 
         Guidelines:
-        1. Use the name, summary, and attributes of nodes to determine if the entities are duplicates, 
-            duplicate nodes may have different names
+        1. Entities with the same name should be considered duplicates
+        2. Duplicate entities may refer to the same real-world entity even if names differ. Use context clues from the MESSAGES
+            to determine if the NEW ENTITY represents a duplicate entity of one of the EXISTING ENTITIES.
         """,
         ),
     ]

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/prompts/extract_nodes.py

@@ -256,7 +256,7 @@ def extract_attributes(context: dict[str, Any]) -> list[Message]:
         1. Do not hallucinate entity property values if they cannot be found in the current context.
         2. Only use the provided MESSAGES and ENTITY to set attribute values.
         3. The summary attribute represents a summary of the ENTITY, and should be updated with new information about the Entity from the MESSAGES. 
-            Summaries must be no longer than 200 words.
+            Summaries must be no longer than 500 words.
         
         <ENTITY>
         {context['node']}

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/utils/maintenance/edge_operations.py

@@ -27,6 +27,7 @@ from graphiti_core.edges import (
 from graphiti_core.graphiti_types import GraphitiClients
 from graphiti_core.helpers import MAX_REFLEXION_ITERATIONS, semaphore_gather
 from graphiti_core.llm_client import LLMClient
+from graphiti_core.llm_client.config import ModelSize
 from graphiti_core.nodes import CommunityNode, EntityNode, EpisodicNode
 from graphiti_core.prompts import prompt_library
 from graphiti_core.prompts.dedupe_edges import EdgeDuplicate, UniqueFacts
@@ -377,7 +378,9 @@ async def dedupe_extracted_edge(
     }
 
     llm_response = await llm_client.generate_response(
-        prompt_library.dedupe_edges.edge(context), response_model=EdgeDuplicate
+        prompt_library.dedupe_edges.edge(context),
+        response_model=EdgeDuplicate,
+        model_size=ModelSize.small,
     )
 
     duplicate_fact_id: int = llm_response.get('duplicate_fact_id', -1)

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/utils/maintenance/node_operations.py

@@ -25,6 +25,7 @@ from pydantic import BaseModel, Field
 from graphiti_core.graphiti_types import GraphitiClients
 from graphiti_core.helpers import MAX_REFLEXION_ITERATIONS, semaphore_gather
 from graphiti_core.llm_client import LLMClient
+from graphiti_core.llm_client.config import ModelSize
 from graphiti_core.nodes import EntityNode, EpisodeType, EpisodicNode, create_entity_node_embeddings
 from graphiti_core.prompts import prompt_library
 from graphiti_core.prompts.dedupe_nodes import NodeDuplicate
@@ -281,7 +282,6 @@ async def resolve_extracted_node(
                 'id': i,
                 'name': node.name,
                 'entity_types': node.labels,
-                'summary': node.summary,
             },
             **node.attributes,
         }
@@ -291,14 +291,14 @@ async def resolve_extracted_node(
     extracted_node_context = {
         'name': extracted_node.name,
         'entity_type': entity_type.__name__ if entity_type is not None else 'Entity',  # type: ignore
-        'entity_type_description': entity_type.__doc__
-        if entity_type is not None
-        else 'Default Entity Type',
     }
 
     context = {
         'existing_nodes': existing_nodes_context,
         'extracted_node': extracted_node_context,
+        'entity_type_description': entity_type.__doc__
+        if entity_type is not None
+        else 'Default Entity Type',
         'episode_content': episode.content if episode is not None else '',
         'previous_episodes': [ep.content for ep in previous_episodes]
         if previous_episodes is not None
@@ -306,7 +306,9 @@ async def resolve_extracted_node(
     }
 
     llm_response = await llm_client.generate_response(
-        prompt_library.dedupe_nodes.node(context), response_model=NodeDuplicate
+        prompt_library.dedupe_nodes.node(context),
+        response_model=NodeDuplicate,
+        model_size=ModelSize.small,
     )
 
     duplicate_id: int = llm_response.get('duplicate_node_id', -1)
@@ -315,6 +317,8 @@ async def resolve_extracted_node(
         existing_nodes[duplicate_id] if 0 <= duplicate_id < len(existing_nodes) else extracted_node
     )
 
+    node.name = llm_response.get('name', '')
+
     end = time()
     logger.debug(
         f'Resolved node: {extracted_node.name} is {node.name}, in {(end - start) * 1000} ms'
@@ -371,13 +375,9 @@ async def extract_attributes_from_node(
         'summary': (
             str,
             Field(
-                description='Summary containing the important information about the entity. Under 200 words',
+                description='Summary containing the important information about the entity. Under 500 words',
             ),
-        ),
-        'name': (
-            str,
-            Field(description='Name of the ENTITY'),
-        ),
+        )
     }
 
     if entity_type is not None:

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/graphiti_core/utils/maintenance/temporal_operations.py

@@ -20,6 +20,7 @@ from time import time
 
 from graphiti_core.edges import EntityEdge
 from graphiti_core.llm_client import LLMClient
+from graphiti_core.llm_client.config import ModelSize
 from graphiti_core.nodes import EpisodicNode
 from graphiti_core.prompts import prompt_library
 from graphiti_core.prompts.extract_edge_dates import EdgeDates
@@ -81,7 +82,9 @@ async def get_edge_contradictions(
     context = {'new_edge': new_edge_context, 'existing_edges': existing_edge_context}
 
     llm_response = await llm_client.generate_response(
-        prompt_library.invalidate_edges.v2(context), response_model=InvalidatedEdges
+        prompt_library.invalidate_edges.v2(context),
+        response_model=InvalidatedEdges,
+        model_size=ModelSize.small,
     )
 
     contradicted_facts: list[int] = llm_response.get('contradicted_facts', [])

{graphiti_core-0.11.3 → graphiti_core-0.11.4}/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "graphiti-core"
 description = "A temporal graph building library"
-version = "0.11.3"
+version = "0.11.4"
 authors = [
     { "name" = "Paul Paliychuk", "email" = "paul@getzep.com" },
     { "name" = "Preston Rasmussen", "email" = "preston@getzep.com" },