graphiti-core 0.22.0rc1__tar.gz → 0.22.0rc3__tar.gz

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: graphiti-core
-Version: 0.22.0rc1
+Version: 0.22.0rc3
 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-0.22.0rc3}/graphiti_core/prompts/extract_nodes.py

@@ -18,49 +18,48 @@ from typing import Any, Protocol, TypedDict
 
 from pydantic import BaseModel, Field
 
+from graphiti_core.utils.text_utils import MAX_SUMMARY_CHARS
+
 from .models import Message, PromptFunction, PromptVersion
 from .prompt_helpers import to_prompt_json
+from .snippets import summary_instructions
 
 
 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",
+        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 8 sentences.",
+        description=f'Summary containing the important information about the entity. Under {MAX_SUMMARY_CHARS} characters.',
     )
 
 
@@ -128,8 +127,8 @@ 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),
     ]
 
 
@@ -161,8 +160,8 @@ 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),
     ]
 
 
@@ -192,8 +191,8 @@ 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),
     ]
 
 
@@ -216,8 +215,8 @@ 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),
     ]
 
 
@@ -248,32 +247,31 @@ 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>
-        {to_prompt_json(context['previous_episodes'], indent=2)}
-        {to_prompt_json(context['episode_content'], indent=2)}
-        </MESSAGES>
-
-        Given the above MESSAGES and the following ENTITY, update any of its attributes based on the information provided
+        Given the MESSAGES and the following ENTITY, update any of its attributes based on the information provided
         in MESSAGES. Use the provided attribute descriptions to better understand how each attribute should be determined.
 
         Guidelines:
         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.
+
+        <MESSAGES>
+        {to_prompt_json(context['previous_episodes'], indent=2)}
+        {to_prompt_json(context['episode_content'], indent=2)}
+        </MESSAGES>
         
         <ENTITY>
         {context['node']}
@@ -286,27 +284,22 @@ 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"""
+        Given the MESSAGES and the ENTITY, update the summary that combines relevant information about the entity
+        from the messages and relevant information from the existing summary.
+        
+        {summary_instructions}
 
         <MESSAGES>
         {to_prompt_json(context['previous_episodes'], indent=2)}
         {to_prompt_json(context['episode_content'], indent=2)}
         </MESSAGES>
 
-        Given the above MESSAGES and the following ENTITY, update the summary that combines relevant information about the entity
-        from the messages and relevant information from the existing summary.
-        
-        Guidelines:
-        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. 
-        4. Keep the summary concise and to the point. SUMMARIES MUST BE LESS THAN 8 SENTENCES.
-
         <ENTITY>
         {context['node']}
         </ENTITY>
@@ -316,11 +309,11 @@ 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,
 }

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/graphiti_core/prompts/prompt_helpers.py

@@ -1,3 +1,19 @@
+"""
+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 json
 from typing import Any
 

graphiti_core-0.22.0rc3/graphiti_core/prompts/snippets.py

@@ -0,0 +1,29 @@
+"""
+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.
+"""
+
+summary_instructions = """Guidelines:
+        1. Output only factual content. Never explain what you're doing, why, or mention limitations/constraints. 
+        2. Only use the provided messages, entity, and entity context to set attribute values.
+        3. Keep the summary concise and to the point. STATE FACTS DIRECTLY IN UNDER 250 CHARACTERS.
+
+        Example summaries:
+        BAD: "This is the only activity in the context. The user listened to this song. No other details were provided to include in this summary."
+        GOOD: "User played 'Blue Monday' by New Order (electronic genre) on 2024-12-03 at 14:22 UTC."
+        BAD: "Based on the messages provided, the user attended a meeting. This summary focuses on that event as it was the main topic discussed."
+        GOOD: "User attended Q3 planning meeting with sales team on March 15."
+        BAD: "The context shows John ordered pizza. Due to length constraints, other details are omitted from this summary."
+        GOOD: "John ordered pepperoni pizza from Mario's at 7:30 PM, delivered to office."
+        """

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/graphiti_core/prompts/summarize_nodes.py

@@ -20,19 +20,18 @@ from pydantic import BaseModel, Field
 
 from .models import Message, PromptFunction, PromptVersion
 from .prompt_helpers import to_prompt_json
+from .snippets import summary_instructions
 
 
 class Summary(BaseModel):
     summary: str = Field(
         ...,
-        description="Summary containing the important information about the entity. Under 8 sentences",
+        description='Summary containing the important information about the entity. Under 250 characters',
     )
 
 
 class SummaryDescription(BaseModel):
-    description: str = Field(
-        ..., description="One sentence description of the provided summary"
-    )
+    description: str = Field(..., description='One sentence description of the provided summary')
 
 
 class Prompt(Protocol):
@@ -50,15 +49,15 @@ class Versions(TypedDict):
 def summarize_pair(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
-            role="system",
-            content="You are a helpful assistant that combines summaries.",
+            role='system',
+            content='You are a helpful assistant that combines summaries.',
         ),
         Message(
-            role="user",
+            role='user',
             content=f"""
         Synthesize the information from the following two summaries into a single succinct summary.
         
-        IMPORTANT: Keep the summary concise and to the point. SUMMARIES MUST BE LESS THAN 8 SENTENCES.
+        IMPORTANT: Keep the summary concise and to the point. SUMMARIES MUST BE LESS THAN 250 CHARACTERS.
 
         Summaries:
         {to_prompt_json(context['node_summaries'], indent=2)}
@@ -70,29 +69,25 @@ def summarize_pair(context: dict[str, Any]) -> list[Message]:
 def summarize_context(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
-            role="system",
-            content="You are a helpful assistant that generates a summary and attributes from provided text.",
+            role='system',
+            content='You are a helpful assistant that generates a summary and attributes from provided text.',
         ),
         Message(
-            role="user",
+            role='user',
             content=f"""
-            
-        <MESSAGES>
-        {to_prompt_json(context['previous_episodes'], indent=2)}
-        {to_prompt_json(context['episode_content'], indent=2)}
-        </MESSAGES>
-        
-        Given the above MESSAGES and the following ENTITY name, create a summary for the ENTITY. Your summary must only use
+        Given the MESSAGES and the ENTITY name, create a summary for the ENTITY. Your summary must only use
         information from the provided MESSAGES. Your summary should also only contain information relevant to the
         provided ENTITY. 
         
         In addition, extract any values for the provided entity properties based on their descriptions.
         If the value of the entity property cannot be found in the current context, set the value of the property to the Python value None.
         
-        Guidelines:
-        1. Do not hallucinate entity property values if they cannot be found in the current context.
-        2. Only use the provided messages, entity, and entity context to set attribute values.
-        3. Keep the summary concise and to the point. SUMMARIES MUST BE LESS THAN 8 SENTENCES.
+        {summary_instructions}
+
+        <MESSAGES>
+        {to_prompt_json(context['previous_episodes'], indent=2)}
+        {to_prompt_json(context['episode_content'], indent=2)}
+        </MESSAGES>
         
         <ENTITY>
         {context['node_name']}
@@ -113,14 +108,14 @@ def summarize_context(context: dict[str, Any]) -> list[Message]:
 def summary_description(context: dict[str, Any]) -> list[Message]:
     return [
         Message(
-            role="system",
-            content="You are a helpful assistant that describes provided contents in a single sentence.",
+            role='system',
+            content='You are a helpful assistant that describes provided contents in a single sentence.',
         ),
         Message(
-            role="user",
+            role='user',
             content=f"""
         Create a short one sentence description of the summary that explains what kind of information is summarized.
-        Summaries must be under 8 sentences.
+        Summaries must be under 250 characters.
 
         Summary:
         {to_prompt_json(context['summary'], indent=2)}
@@ -130,7 +125,7 @@ def summary_description(context: dict[str, Any]) -> list[Message]:
 
 
 versions: Versions = {
-    "summarize_pair": summarize_pair,
-    "summarize_context": summarize_context,
-    "summary_description": summary_description,
+    'summarize_pair': summarize_pair,
+    'summarize_context': summarize_context,
+    'summary_description': summary_description,
 }

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/graphiti_core/utils/maintenance/node_operations.py

@@ -53,6 +53,7 @@ from graphiti_core.utils.maintenance.dedup_helpers import (
 from graphiti_core.utils.maintenance.edge_operations import (
     filter_existing_duplicate_of_edges,
 )
+from graphiti_core.utils.text_utils import MAX_SUMMARY_CHARS, truncate_at_sentence
 
 logger = logging.getLogger(__name__)
 
@@ -547,7 +548,7 @@ async def _extract_entity_summary(
     summary_context = _build_episode_context(
         node_data={
             'name': node.name,
-            'summary': node.summary,
+            'summary': truncate_at_sentence(node.summary, MAX_SUMMARY_CHARS),
             'entity_types': node.labels,
             'attributes': node.attributes,
         },
@@ -562,7 +563,7 @@ async def _extract_entity_summary(
         group_id=node.group_id,
     )
 
-    node.summary = summary_response.get('summary', '')
+    node.summary = truncate_at_sentence(summary_response.get('summary', ''), MAX_SUMMARY_CHARS)
 
 
 def _build_episode_context(

graphiti_core-0.22.0rc3/graphiti_core/utils/text_utils.py

@@ -0,0 +1,53 @@
+"""
+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 re
+
+# Maximum length for entity/node summaries
+MAX_SUMMARY_CHARS = 250
+
+
+def truncate_at_sentence(text: str, max_chars: int) -> str:
+    """
+    Truncate text at or about max_chars while respecting sentence boundaries.
+
+    Attempts to truncate at the last complete sentence before max_chars.
+    If no sentence boundary is found before max_chars, truncates at max_chars.
+
+    Args:
+        text: The text to truncate
+        max_chars: Maximum number of characters
+
+    Returns:
+        Truncated text
+    """
+    if not text or len(text) <= max_chars:
+        return text
+
+    # Find all sentence boundaries (., !, ?) up to max_chars
+    truncated = text[:max_chars]
+
+    # Look for sentence boundaries: period, exclamation, or question mark followed by space or end
+    sentence_pattern = r'[.!?](?:\s|$)'
+    matches = list(re.finditer(sentence_pattern, truncated))
+
+    if matches:
+        # Truncate at the last sentence boundary found
+        last_match = matches[-1]
+        return text[: last_match.end()].rstrip()
+
+    # No sentence boundary found, truncate at max_chars
+    return truncated.rstrip()

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "graphiti-core"
 description = "A temporal graph building library"
-version = "0.22.0pre1"
+version = "0.22.0pre3"
 authors = [
     { name = "Paul Paliychuk", email = "paul@getzep.com" },
     { name = "Preston Rasmussen", email = "preston@getzep.com" },

graphiti_core-0.22.0rc3/tests/test_text_utils.py

@@ -0,0 +1,106 @@
+"""
+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.
+"""
+
+from graphiti_core.utils.text_utils import MAX_SUMMARY_CHARS, truncate_at_sentence
+
+
+def test_truncate_at_sentence_short_text():
+    """Test that short text is returned unchanged."""
+    text = 'This is a short sentence.'
+    result = truncate_at_sentence(text, 100)
+    assert result == text
+
+
+def test_truncate_at_sentence_empty():
+    """Test that empty text is handled correctly."""
+    assert truncate_at_sentence('', 100) == ''
+    assert truncate_at_sentence(None, 100) is None
+
+
+def test_truncate_at_sentence_exact_length():
+    """Test text at exactly max_chars."""
+    text = 'A' * 100
+    result = truncate_at_sentence(text, 100)
+    assert result == text
+
+
+def test_truncate_at_sentence_with_period():
+    """Test truncation at sentence boundary with period."""
+    text = 'First sentence. Second sentence. Third sentence. Fourth sentence.'
+    result = truncate_at_sentence(text, 40)
+    assert result == 'First sentence. Second sentence.'
+    assert len(result) <= 40
+
+
+def test_truncate_at_sentence_with_question():
+    """Test truncation at sentence boundary with question mark."""
+    text = 'What is this? This is a test. More text here.'
+    result = truncate_at_sentence(text, 30)
+    assert result == 'What is this? This is a test.'
+    assert len(result) <= 32
+
+
+def test_truncate_at_sentence_with_exclamation():
+    """Test truncation at sentence boundary with exclamation mark."""
+    text = 'Hello world! This is exciting. And more text.'
+    result = truncate_at_sentence(text, 30)
+    assert result == 'Hello world! This is exciting.'
+    assert len(result) <= 32
+
+
+def test_truncate_at_sentence_no_boundary():
+    """Test truncation when no sentence boundary exists before max_chars."""
+    text = 'This is a very long sentence without any punctuation marks near the beginning'
+    result = truncate_at_sentence(text, 30)
+    assert len(result) <= 30
+    assert result.startswith('This is a very long sentence')
+
+
+def test_truncate_at_sentence_multiple_periods():
+    """Test with multiple sentence endings."""
+    text = 'A. B. C. D. E. F. G. H.'
+    result = truncate_at_sentence(text, 10)
+    assert result == 'A. B. C.'
+    assert len(result) <= 10
+
+
+def test_truncate_at_sentence_strips_trailing_whitespace():
+    """Test that trailing whitespace is stripped."""
+    text = 'First sentence.   Second sentence.'
+    result = truncate_at_sentence(text, 20)
+    assert result == 'First sentence.'
+    assert not result.endswith(' ')
+
+
+def test_max_summary_chars_constant():
+    """Test that MAX_SUMMARY_CHARS is set to expected value."""
+    assert MAX_SUMMARY_CHARS == 250
+
+
+def test_truncate_at_sentence_realistic_summary():
+    """Test with a realistic entity summary."""
+    text = (
+        'John is a software engineer who works at a tech company in San Francisco. '
+        'He has been programming for over 10 years and specializes in Python and distributed systems. '
+        'John enjoys hiking on weekends and is learning to play guitar. '
+        'He graduated from MIT with a degree in computer science.'
+    )
+    result = truncate_at_sentence(text, MAX_SUMMARY_CHARS)
+    assert len(result) <= MAX_SUMMARY_CHARS
+    # Should keep complete sentences
+    assert result.endswith('.')
+    # Should include at least the first sentence
+    assert 'John is a software engineer' in result

{graphiti_core-0.22.0rc1 → graphiti_core-0.22.0rc3}/uv.lock

@@ -783,7 +783,7 @@ wheels = [
 
 [[package]]
 name = "graphiti-core"
-version = "0.22.0rc1"
+version = "0.22.0rc3"
 source = { editable = "." }
 dependencies = [
     { name = "diskcache" },