graphiti-core 0.21.0rc13__tar.gz → 0.22.0rc1__tar.gz

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/.github/workflows/issue-triage.yml

@@ -98,6 +98,7 @@ jobs:
       - name: Check for duplicate issues
         uses: anthropics/claude-code-action@v1
         with:
+          allowed_non_write_users: "*"
           prompt: |
             Analyze this new issue and check if it's a duplicate of existing issues in the repository.
 

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: graphiti-core
-Version: 0.21.0rc13
+Version: 0.22.0rc1
 Summary: A temporal graph building library
 Project-URL: Homepage, https://help.getzep.com/graphiti/graphiti/overview
 Project-URL: Repository, https://github.com/getzep/graphiti

graphiti_core-0.22.0rc1/graphiti_core/driver/driver.py

@@ -0,0 +1,117 @@
+"""
+Copyright 2024, Zep Software, Inc.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+"""
+
+import copy
+import logging
+import os
+from abc import ABC, abstractmethod
+from collections.abc import Coroutine
+from enum import Enum
+from typing import Any
+
+from dotenv import load_dotenv
+
+logger = logging.getLogger(__name__)
+
+DEFAULT_SIZE = 10
+
+load_dotenv()
+
+ENTITY_INDEX_NAME = os.environ.get('ENTITY_INDEX_NAME', 'entities')
+EPISODE_INDEX_NAME = os.environ.get('EPISODE_INDEX_NAME', 'episodes')
+COMMUNITY_INDEX_NAME = os.environ.get('COMMUNITY_INDEX_NAME', 'communities')
+ENTITY_EDGE_INDEX_NAME = os.environ.get('ENTITY_EDGE_INDEX_NAME', 'entity_edges')
+
+
+class GraphProvider(Enum):
+    NEO4J = 'neo4j'
+    FALKORDB = 'falkordb'
+    KUZU = 'kuzu'
+    NEPTUNE = 'neptune'
+
+
+class GraphDriverSession(ABC):
+    provider: GraphProvider
+
+    async def __aenter__(self):
+        return self
+
+    @abstractmethod
+    async def __aexit__(self, exc_type, exc, tb):
+        # No cleanup needed for Falkor, but method must exist
+        pass
+
+    @abstractmethod
+    async def run(self, query: str, **kwargs: Any) -> Any:
+        raise NotImplementedError()
+
+    @abstractmethod
+    async def close(self):
+        raise NotImplementedError()
+
+    @abstractmethod
+    async def execute_write(self, func, *args, **kwargs):
+        raise NotImplementedError()
+
+
+class GraphDriver(ABC):
+    provider: GraphProvider
+    fulltext_syntax: str = (
+        ''  # Neo4j (default) syntax does not require a prefix for fulltext queries
+    )
+    _database: str
+    aoss_client: Any  # type: ignore
+
+    @abstractmethod
+    def execute_query(self, cypher_query_: str, **kwargs: Any) -> Coroutine:
+        raise NotImplementedError()
+
+    @abstractmethod
+    def session(self, database: str | None = None) -> GraphDriverSession:
+        raise NotImplementedError()
+
+    @abstractmethod
+    def close(self):
+        raise NotImplementedError()
+
+    @abstractmethod
+    def delete_all_indexes(self) -> Coroutine:
+        raise NotImplementedError()
+
+    def with_database(self, database: str) -> 'GraphDriver':
+        """
+        Returns a shallow copy of this driver with a different default database.
+        Reuses the same connection (e.g. FalkorDB, Neo4j).
+        """
+        cloned = copy.copy(self)
+        cloned._database = database
+
+        return cloned
+
+    def build_fulltext_query(
+        self, query: str, group_ids: list[str] | None = None, max_query_length: int = 128
+    ) -> str:
+        """
+        Specific fulltext query builder for database providers.
+        Only implemented by providers that need custom fulltext query building.
+        """
+        raise NotImplementedError(f'build_fulltext_query not implemented for {self.provider}')
+
+    async def save_to_aoss(self, name: str, data: list[dict]) -> int:
+        return 0
+
+    async def clear_aoss_indices(self):
+        return 1

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/driver/neo4j_driver.py

@@ -22,28 +22,9 @@ from neo4j import AsyncGraphDatabase, EagerResult
 from typing_extensions import LiteralString
 
 from graphiti_core.driver.driver import GraphDriver, GraphDriverSession, GraphProvider
-from graphiti_core.helpers import semaphore_gather
 
 logger = logging.getLogger(__name__)
 
-try:
-    import boto3
-    from opensearchpy import (
-        AIOHttpConnection,
-        AsyncOpenSearch,
-        AWSV4SignerAuth,
-        Urllib3AWSV4SignerAuth,
-        Urllib3HttpConnection,
-    )
-
-    _HAS_OPENSEARCH = True
-except ImportError:
-    boto3 = None
-    OpenSearch = None
-    Urllib3AWSV4SignerAuth = None
-    Urllib3HttpConnection = None
-    _HAS_OPENSEARCH = False
-
 
 class Neo4jDriver(GraphDriver):
     provider = GraphProvider.NEO4J
@@ -54,11 +35,6 @@ class Neo4jDriver(GraphDriver):
         user: str | None,
         password: str | None,
         database: str = 'neo4j',
-        aoss_host: str | None = None,
-        aoss_port: int | None = None,
-        aws_profile_name: str | None = None,
-        aws_region: str | None = None,
-        aws_service: str | None = None,
     ):
         super().__init__()
         self.client = AsyncGraphDatabase.driver(
@@ -68,24 +44,6 @@ class Neo4jDriver(GraphDriver):
         self._database = database
 
         self.aoss_client = None
-        if aoss_host and aoss_port and boto3 is not None:
-            try:
-                region = aws_region
-                service = aws_service
-                credentials = boto3.Session(profile_name=aws_profile_name).get_credentials()
-                auth = AWSV4SignerAuth(credentials, region or '', service or '')
-
-                self.aoss_client = AsyncOpenSearch(
-                    hosts=[{'host': aoss_host, 'port': aoss_port}],
-                    auth=auth,
-                    use_ssl=True,
-                    verify_certs=True,
-                    connection_class=AIOHttpConnection,
-                    pool_maxsize=20,
-                )  # type: ignore
-            except Exception as e:
-                logger.warning(f'Failed to initialize OpenSearch client: {e}')
-                self.aoss_client = None
 
     async def execute_query(self, cypher_query_: LiteralString, **kwargs: Any) -> EagerResult:
         # Check if database_ is provided in kwargs.
@@ -111,13 +69,6 @@ class Neo4jDriver(GraphDriver):
         return await self.client.close()
 
     def delete_all_indexes(self) -> Coroutine:
-        if self.aoss_client:
-            return semaphore_gather(
-                self.client.execute_query(
-                    'CALL db.indexes() YIELD name DROP INDEX name',
-                ),
-                self.delete_aoss_indices(),
-            )
         return self.client.execute_query(
             'CALL db.indexes() YIELD name DROP INDEX name',
         )

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/driver/neptune_driver.py

@@ -22,21 +22,16 @@ from typing import Any
 
 import boto3
 from langchain_aws.graphs import NeptuneAnalyticsGraph, NeptuneGraph
-from opensearchpy import OpenSearch, Urllib3AWSV4SignerAuth, Urllib3HttpConnection
+from opensearchpy import OpenSearch, Urllib3AWSV4SignerAuth, Urllib3HttpConnection, helpers
 
-from graphiti_core.driver.driver import (
-    DEFAULT_SIZE,
-    GraphDriver,
-    GraphDriverSession,
-    GraphProvider,
-)
+from graphiti_core.driver.driver import GraphDriver, GraphDriverSession, GraphProvider
 
 logger = logging.getLogger(__name__)
+DEFAULT_SIZE = 10
 
-neptune_aoss_indices = [
+aoss_indices = [
     {
         'index_name': 'node_name_and_summary',
-        'alias_name': 'entities',
         'body': {
             'mappings': {
                 'properties': {
@@ -54,7 +49,6 @@ neptune_aoss_indices = [
     },
     {
         'index_name': 'community_name',
-        'alias_name': 'communities',
         'body': {
             'mappings': {
                 'properties': {
@@ -71,7 +65,6 @@ neptune_aoss_indices = [
     },
     {
         'index_name': 'episode_content',
-        'alias_name': 'episodes',
         'body': {
             'mappings': {
                 'properties': {
@@ -95,7 +88,6 @@ neptune_aoss_indices = [
     },
     {
         'index_name': 'edge_name_and_fact',
-        'alias_name': 'facts',
         'body': {
             'mappings': {
                 'properties': {
@@ -228,27 +220,52 @@ class NeptuneDriver(GraphDriver):
     async def _delete_all_data(self) -> Any:
         return await self.execute_query('MATCH (n) DETACH DELETE n')
 
+    def delete_all_indexes(self) -> Coroutine[Any, Any, Any]:
+        return self.delete_all_indexes_impl()
+
+    async def delete_all_indexes_impl(self) -> Coroutine[Any, Any, Any]:
+        # No matter what happens above, always return True
+        return self.delete_aoss_indices()
+
     async def create_aoss_indices(self):
-        for index in neptune_aoss_indices:
+        for index in aoss_indices:
             index_name = index['index_name']
             client = self.aoss_client
-            if not client:
-                raise ValueError(
-                    'You must provide an AOSS endpoint to create an OpenSearch driver.'
-                )
             if not client.indices.exists(index=index_name):
-                await client.indices.create(index=index_name, body=index['body'])
-
-            alias_name = index.get('alias_name', index_name)
-
-            if not client.indices.exists_alias(name=alias_name, index=index_name):
-                await client.indices.put_alias(index=index_name, name=alias_name)
-
+                client.indices.create(index=index_name, body=index['body'])
         # Sleep for 1 minute to let the index creation complete
         await asyncio.sleep(60)
 
-    def delete_all_indexes(self) -> Coroutine[Any, Any, Any]:
-        return self.delete_all_indexes_impl()
+    async def delete_aoss_indices(self):
+        for index in aoss_indices:
+            index_name = index['index_name']
+            client = self.aoss_client
+            if client.indices.exists(index=index_name):
+                client.indices.delete(index=index_name)
+
+    def run_aoss_query(self, name: str, query_text: str, limit: int = 10) -> dict[str, Any]:
+        for index in aoss_indices:
+            if name.lower() == index['index_name']:
+                index['query']['query']['multi_match']['query'] = query_text
+                query = {'size': limit, 'query': index['query']}
+                resp = self.aoss_client.search(body=query['query'], index=index['index_name'])
+                return resp
+        return {}
+
+    def save_to_aoss(self, name: str, data: list[dict]) -> int:
+        for index in aoss_indices:
+            if name.lower() == index['index_name']:
+                to_index = []
+                for d in data:
+                    item = {'_index': name, '_id': d['uuid']}
+                    for p in index['body']['mappings']['properties']:
+                        if p in d:
+                            item[p] = d[p]
+                    to_index.append(item)
+                success, failed = helpers.bulk(self.aoss_client, to_index, stats_only=True)
+                return success
+
+        return 0
 
 
 class NeptuneDriverSession(GraphDriverSession):

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/llm_client/client.py

@@ -33,12 +33,16 @@ DEFAULT_TEMPERATURE = 0
 DEFAULT_CACHE_DIR = './llm_cache'
 
 
-def get_extraction_language_instruction() -> str:
+def get_extraction_language_instruction(group_id: str | None = None) -> str:
     """Returns instruction for language extraction behavior.
 
     Override this function to customize language extraction:
     - Return empty string to disable multilingual instructions
     - Return custom instructions for specific language requirements
+    - Use group_id to provide different instructions per group/partition
+
+    Args:
+        group_id: Optional partition identifier for the graph
 
     Returns:
         str: Language instruction to append to system messages
@@ -142,6 +146,7 @@ class LLMClient(ABC):
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
         model_size: ModelSize = ModelSize.medium,
+        group_id: str | None = None,
     ) -> dict[str, typing.Any]:
         if max_tokens is None:
             max_tokens = self.max_tokens
@@ -155,7 +160,7 @@ class LLMClient(ABC):
             )
 
         # Add multilingual extraction instructions
-        messages[0].content += get_extraction_language_instruction()
+        messages[0].content += get_extraction_language_instruction(group_id)
 
         if self.cache_enabled and self.cache_dir is not None:
             cache_key = self._get_cache_key(messages)

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/llm_client/gemini_client.py

@@ -357,6 +357,7 @@ class GeminiClient(LLMClient):
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
         model_size: ModelSize = ModelSize.medium,
+        group_id: str | None = None,
     ) -> dict[str, typing.Any]:
         """
         Generate a response from the Gemini language model with retry logic and error handling.
@@ -367,6 +368,7 @@ class GeminiClient(LLMClient):
             response_model (type[BaseModel] | None): An optional Pydantic model to parse the response into.
             max_tokens (int | None): The maximum number of tokens to generate in the response.
             model_size (ModelSize): The size of the model to use (small or medium).
+            group_id (str | None): Optional partition identifier for the graph.
 
         Returns:
             dict[str, typing.Any]: The response from the language model.
@@ -376,7 +378,7 @@ class GeminiClient(LLMClient):
         last_output = None
 
         # Add multilingual extraction instructions
-        messages[0].content += get_extraction_language_instruction()
+        messages[0].content += get_extraction_language_instruction(group_id)
 
         while retry_count < self.MAX_RETRIES:
             try:

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/llm_client/openai_base_client.py

@@ -175,6 +175,7 @@ class BaseOpenAIClient(LLMClient):
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
         model_size: ModelSize = ModelSize.medium,
+        group_id: str | None = None,
     ) -> dict[str, typing.Any]:
         """Generate a response with retry logic and error handling."""
         if max_tokens is None:
@@ -184,7 +185,7 @@ class BaseOpenAIClient(LLMClient):
         last_error = None
 
         # Add multilingual extraction instructions
-        messages[0].content += get_extraction_language_instruction()
+        messages[0].content += get_extraction_language_instruction(group_id)
 
         while retry_count <= self.MAX_RETRIES:
             try:

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/llm_client/openai_generic_client.py

@@ -120,6 +120,7 @@ class OpenAIGenericClient(LLMClient):
         response_model: type[BaseModel] | None = None,
         max_tokens: int | None = None,
         model_size: ModelSize = ModelSize.medium,
+        group_id: str | None = None,
     ) -> dict[str, typing.Any]:
         if max_tokens is None:
             max_tokens = self.max_tokens
@@ -136,7 +137,7 @@ class OpenAIGenericClient(LLMClient):
             )
 
         # Add multilingual extraction instructions
-        messages[0].content += get_extraction_language_instruction()
+        messages[0].content += get_extraction_language_instruction(group_id)
 
         while retry_count <= self.MAX_RETRIES:
             try:

{graphiti_core-0.21.0rc13 → graphiti_core-0.22.0rc1}/graphiti_core/prompts/extract_nodes.py

@@ -23,39 +23,44 @@ from .prompt_helpers import to_prompt_json
 
 
 class ExtractedEntity(BaseModel):
-    name: str = Field(..., description='Name of the extracted entity')
+    name: str = Field(..., description="Name of the extracted entity")
     entity_type_id: int = Field(
-        description='ID of the classified entity type. '
-        'Must be one of the provided entity_type_id integers.',
+        description="ID of the classified entity type. "
+        "Must be one of the provided entity_type_id integers.",
     )
 
 
 class ExtractedEntities(BaseModel):
-    extracted_entities: list[ExtractedEntity] = Field(..., description='List of extracted entities')
+    extracted_entities: list[ExtractedEntity] = Field(
+        ..., description="List of extracted entities"
+    )
 
 
 class MissedEntities(BaseModel):
-    missed_entities: list[str] = Field(..., description="Names of entities that weren't extracted")
+    missed_entities: list[str] = Field(
+        ..., description="Names of entities that weren't extracted"
+    )
 
 
 class EntityClassificationTriple(BaseModel):
-    uuid: str = Field(description='UUID of the entity')
-    name: str = Field(description='Name of the entity')
+    uuid: str = Field(description="UUID of the entity")
+    name: str = Field(description="Name of the entity")
     entity_type: str | None = Field(
-        default=None, description='Type of the entity. Must be one of the provided types or None'
+        default=None,
+        description="Type of the entity. Must be one of the provided types or None",
     )
 
 
 class EntityClassification(BaseModel):
     entity_classifications: list[EntityClassificationTriple] = Field(
-        ..., description='List of entities classification triples.'
+        ..., description="List of entities classification triples."
     )
 
 
 class EntitySummary(BaseModel):
     summary: str = Field(
         ...,
-        description='Summary containing the important information about the entity. Under 250 words',
+        description="Summary containing the important information about the entity. Under 8 sentences.",
     )
 
 
@@ -123,8 +128,8 @@ reference entities. Only extract distinct entities from the CURRENT MESSAGE. Don
 {context['custom_prompt']}
 """
     return [
-        Message(role='system', content=sys_prompt),
-        Message(role='user', content=user_prompt),
+        Message(role="system", content=sys_prompt),
+        Message(role="user", content=user_prompt),
     ]
 
 
@@ -156,8 +161,8 @@ Guidelines:
 3. Do NOT extract any properties that contain dates
 """
     return [
-        Message(role='system', content=sys_prompt),
-        Message(role='user', content=user_prompt),
+        Message(role="system", content=sys_prompt),
+        Message(role="user", content=user_prompt),
     ]
 
 
@@ -187,8 +192,8 @@ Guidelines:
 4. Be as explicit as possible in your node names, using full names and avoiding abbreviations.
 """
     return [
-        Message(role='system', content=sys_prompt),
-        Message(role='user', content=user_prompt),
+        Message(role="system", content=sys_prompt),
+        Message(role="user", content=user_prompt),
     ]
 
 
@@ -211,8 +216,8 @@ Given the above previous messages, current message, and list of extracted entiti
 extracted.
 """
     return [
-        Message(role='system', content=sys_prompt),
-        Message(role='user', content=user_prompt),
+        Message(role="system", content=sys_prompt),
+        Message(role="user", content=user_prompt),
     ]
 
 
@@ -243,19 +248,19 @@ def classify_nodes(context: dict[str, Any]) -> list[Message]:
     3. If none of the provided entity types accurately classify an extracted node, the type should be set to None
 """
     return [
-        Message(role='system', content=sys_prompt),
-        Message(role='user', content=user_prompt),
+        Message(role="system", content=sys_prompt),
+        Message(role="user", content=user_prompt),
     ]
 
 
 def extract_attributes(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
-            role='system',
-            content='You are a helpful assistant that extracts entity properties from the provided text.',
+            role="system",
+            content="You are a helpful assistant that extracts entity properties from the provided text.",
         ),
         Message(
-            role='user',
+            role="user",
             content=f"""
 
         <MESSAGES>
@@ -281,11 +286,11 @@ def extract_attributes(context: dict[str, Any]) -> list[Message]:
 def extract_summary(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
-            role='system',
-            content='You are a helpful assistant that extracts entity summaries from the provided text.',
+            role="system",
+            content="You are a helpful assistant that extracts entity summaries from the provided text.",
         ),
         Message(
-            role='user',
+            role="user",
             content=f"""
 
         <MESSAGES>
@@ -300,7 +305,7 @@ def extract_summary(context: dict[str, Any]) -> list[Message]:
         1. Do not hallucinate entity summary information 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 250 words.
+        4. Keep the summary concise and to the point. SUMMARIES MUST BE LESS THAN 8 SENTENCES.
 
         <ENTITY>
         {context['node']}
@@ -311,11 +316,11 @@ def extract_summary(context: dict[str, Any]) -> list[Message]:
 
 
 versions: Versions = {
-    'extract_message': extract_message,
-    'extract_json': extract_json,
-    'extract_text': extract_text,
-    'reflexion': reflexion,
-    'extract_summary': extract_summary,
-    'classify_nodes': classify_nodes,
-    'extract_attributes': extract_attributes,
+    "extract_message": extract_message,
+    "extract_json": extract_json,
+    "extract_text": extract_text,
+    "reflexion": reflexion,
+    "extract_summary": extract_summary,
+    "classify_nodes": classify_nodes,
+    "extract_attributes": extract_attributes,
 }