* Fix datetime comparison errors by normalizing to UTC
Applied ensure_utc() to all datetime comparisons in edge_operations.py to prevent TypeError when comparing timezone-naive and timezone-aware datetimes. Removed redundant tzinfo checks since ensure_utc() handles both None and naive datetimes.
Fixed comparisons at:
- Lines 419, 423: resolve_edge_contradictions function
- Line 430: resolve_edge_contradictions function
- Line 627: resolve_extracted_edge function (removed redundant tzinfo checks)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Update uv.lock
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix sorting with mixed timezone-aware/naive datetimes
Normalize datetime to UTC in sort key to prevent TypeError when comparing mixed timezone-aware and timezone-naive datetimes during sorting.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
- Created examples/opentelemetry/ with working stdout tracing example
- Uses Kuzu in-memory database for zero-setup requirement
- Demonstrates ingestion and search with distributed tracing
- Updated OTEL_TRACING.md with simplified documentation and Kuzu example
- Uses local editable graphiti-core install for development
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
Changes to `to_prompt_json()` helper to default to minified JSON (no indentation) instead of 2-space indentation. This reduces token consumption in LLM prompts while maintaining all necessary information.
- Changed default `indent` parameter from `2` to `None` in `prompt_helpers.py`
- Updated all prompt modules to remove explicit `indent=2` arguments
- Minor code formatting fixes in LLM clients
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
* Add OpenTelemetry distributed tracing support
- Add tracer abstraction with no-op and OpenTelemetry implementations
- Instrument add_episode and add_episode_bulk with tracing spans
- Instrument LLM client with cache-aware tracing
- Add configurable span name prefix support
- Refactor add_episode methods to improve code quality
- Add OTEL_TRACING.md documentation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix linting errors in tracing implementation
- Remove unused episodes_by_uuid variable
- Fix tracer type annotations for context manager support
- Replace isinstance tuple with union syntax
- Use contextlib.suppress for exception handling
- Fix import ordering and use AbstractContextManager
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Address PR review feedback on tracing implementation
Critical fixes:
- Remove flawed error span creation in graphiti.py that created orphaned spans
- Restructure LLM client tracing to create span once at start, eliminating code duplication
- Initialize LLM client tracer to NoOpTracer by default to fix type checking
Enhancements:
- Add comprehensive span attributes to add_episode: reference_time, entity/edge type counts, previous episodes count, invalidated edge count, community count
- Optimize isinstance check for better performance
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Add prompt name tracking to OpenTelemetry tracing spans
Add prompt_name parameter to all LLM client generate_response() methods
and set it as a span attribute in the llm.generate span. This enables
better observability by identifying which prompt template was used for
each LLM call.
Changes:
- Add prompt_name parameter to LLMClient.generate_response() base method
- Add prompt_name parameter and tracing to OpenAIBaseClient,
AnthropicClient, GeminiClient, and OpenAIGenericClient
- Update all 14 LLM call sites across maintenance operations to include
prompt_name:
- edge_operations.py: 4 calls
- node_operations.py: 6 calls (note: 7 listed but only 6 unique)
- temporal_operations.py: 2 calls
- community_operations.py: 2 calls
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix exception handling in add_episode to record errors in OpenTelemetry span
Moved try-except block inside the OpenTelemetry span context and added
proper error recording with span.set_status() and span.record_exception().
This ensures exceptions are captured in the distributed trace, matching
the pattern used in add_episode_bulk.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Refactor prompt structure: move MESSAGES after instructions
Reordered prompt structure in extract_nodes.py to place MESSAGES section
after instructions/guidelines in both extract_attributes and extract_summary
functions for improved prompt clarity.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Add sentence-aware text truncator for entity summaries
- Created truncate_at_sentence() utility function that truncates text at
sentence boundaries while respecting max character limits
- Added MAX_SUMMARY_CHARS constant (250 chars) for entity summaries
- Applied truncator to entity summaries in prompts (extract_nodes.py)
- Applied truncator to LLM-generated summaries (node_operations.py)
- Added comprehensive test suite for truncation logic
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Clean up formatting in extract_attributes prompt
- Remove extra blank lines
- Fix indentation of MESSAGES tag
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Bump version to 0.22.0pre3
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Refactor summary prompts to use character limit and prevent meta-commentary
- Changed summary length constraint from "8 sentences" to "250 characters" for more predictable output
- Created reusable summary_instructions snippet in snippets.py with clear BAD/GOOD examples
- Added explicit instruction to output only factual content without meta-commentary
- Applied consistent formatting across extract_nodes.py and summarize_nodes.py
- Bumped version to 0.22.0pre2
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Add copyright header to snippets.py
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Enforce shorter summaries with 8 sentence limit
Replace 250-word limit with 8 sentence limit for node summaries to improve conciseness. Also update prompt system message for summarize_context to better reflect its dual purpose of generating summaries and attributes.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Update graphiti_core/prompts/summarize_nodes.py
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* Bump version to 0.22.0pre1
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Update graphiti_core/prompts/summarize_nodes.py
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* Refactor node extraction for better maintainability
- Extract helper functions from extract_attributes_from_node to improve code organization
- Add _extract_entity_attributes, _extract_entity_summary, and _build_episode_context helpers
- Apply consistent formatting (double quotes per ruff configuration)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Apply consistent single quote style throughout node_operations
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* cleanup
* cleanup
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Bump version to 0.22.0pre0
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Add group_id parameter to get_extraction_language_instruction
Enable consumers to provide group-specific language extraction
instructions by passing group_id through the call chain.
Changes:
- Add optional group_id parameter to get_extraction_language_instruction()
- Add group_id parameter to all LLMClient.generate_response() methods
- Pass group_id through to language instruction function
- Maintain backward compatibility with default None value
Users can now customize extraction per group:
```python
def custom_instruction(group_id: str | None = None) -> str:
if group_id == 'spanish-users':
return '\n\nExtract in Spanish.'
return '\n\nExtract in original language.'
client.get_extraction_language_instruction = custom_instruction
```
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Pass group_id to generate_response in extraction operations
Thread group_id parameter through all extraction-related generate_response()
calls where it's naturally available (via episode.group_id or node.group_id).
This enables consumers to override get_extraction_language_instruction() with
group-specific language preferences.
Changes:
- edge_operations.py: Pass group_id in extract_edges()
- node_operations.py: Pass episode.group_id in extract_nodes() and
node.group_id in extract_attributes_from_node()
- node_operations.py: Add group_id parameter to extract_nodes_reflexion()
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix type inconsistency in extract_nodes_reflexion parameter
Change group_id parameter from str = '' to str | None = None to match
the pattern used throughout the codebase and align with the optional
nature of group_id in generate_response().
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Remove ensure_ascii parameter and uv.lock file
* Reset uv.lock to main branch version
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: Improve edge extraction entity ID validation
Fixes invalid entity ID references in edge extraction that caused warnings like:
"WARNING: source or target node not filled WILL_FIND. source_node_uuid: 23 and target_node_uuid: 3"
Changes:
- Format ENTITIES list as proper JSON in prompt for better LLM parsing
- Clarify field descriptions to reference entity id from ENTITIES list
- Add explicit entity ID validation as #1 extraction rule with examples
- Improve error logging (removed PII, added entity count and valid range)
These changes follow patterns from extract_nodes.py and dedupe_nodes.py where
entity referencing works reliably.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* wip
* fix: Align fact field naming and add description
- Change extraction rule to reference 'fact' instead of 'fact_text'
- Add descriptive text for fact field in Edge model
* fix: Remove ensure_ascii parameter from to_prompt_json call
Align with other to_prompt_json calls that don't use ensure_ascii
* fix: Use validated target_node_idx variable consistently
Line 190 was using raw edge_data.target_entity_id instead of the
validated target_node_idx variable, creating inconsistency with line 189
* fix: Improve edge extraction validation checks
- Add explicit check for empty nodes list
- Use more explicit 0 <= idx comparison instead of -1 < idx
- Prevents nonsensical error message when no entities provided
* chore: Restore uv.lock from main branch
Previously deleted in commit 7e4464b, now restored to match main branch state
* Update uv.lock
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Optimize edge deduplication prompt for caching and clarity
- Restructure prompt to place invariant instructions at top and dynamic context at bottom for better LLM caching
- Change 'id' to 'idx' in edge context lists to avoid confusion with other identifiers
- Remove 'fact_type_id' from edge types context as LLM only needs fact_type_name
- Remove dynamic range values from prompt instructions (e.g., "range 0-N")
- Add debug logging before LLM call to track input sizes
- Add validation logging after LLM response to catch invalid idx values
- Clarify that duplicate_facts uses EXISTING FACTS idx and contradicted_facts uses INVALIDATION CANDIDATES idx
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Address terminology consistency and edge case logging
- Update Pydantic field descriptions to use 'idx' instead of 'ids' for consistency
- Fix debug logging to handle empty list edge case (avoid 'idx 0--1' display)
Note on review feedback:
- Validation is intentionally non-redundant: warnings provide visibility, list comprehensions ensure robustness
- WARNING level is appropriate for LLM output issues (not system errors)
- Existing test coverage is sufficient for this defensive logging addition
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Remove ensure_ascii configuration parameter
- Changed to_prompt_json default from ensure_ascii=True to False
- Removed ensure_ascii parameter from Graphiti.__init__ and GraphitiClients
- Removed ensure_ascii from all function signatures and context dictionaries
- Removed ensure_ascii from all test files
- All JSON serialization now preserves Unicode characters by default
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* format
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: Improve deduplication ID validation and logging
- Add comprehensive logging to verify IDs sent to LLM (sent vs received)
- Enhance prompt with explicit ID bounds (0 through N-1)
- Add validation warnings for missing and extra IDs from LLM responses
- Improve error message clarity for invalid dedupe IDs
- Log actual IDs sent to LLM to confirm no index leakage
This helps diagnose cases where the LLM returns IDs outside the valid
range (e.g., ID 19 when only 0-18 were sent).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Remove redundant logging parameter
Address reviewer comment about redundant third parameter in debug log statement.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Address reviewer comments on list slicing and prompt clarity
- Fix list slicing bug: change <= to < to avoid gap when exactly 20 elements
(previously would skip element 10 when showing 21 elements)
- Consolidate redundant prompt phrasing while maintaining clarity
(reduced from 3 sentences to 2, keeping essential constraints)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Remove redundant prompt text to reduce token usage
Consolidate 'using these exact IDs (0 through N-1)' with following sentence
to eliminate repetition. Changes:
- 'using these exact IDs (0 through {N-1}). Do not skip IDs or use IDs outside this range'
- 'with IDs 0 through {N-1}. Do not skip or add IDs'
Saves ~15 tokens per deduplication call.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
- Consolidate issue-triage.yml and issue-deduplication.yml into single workflow with sequential jobs
- Create daily_issue_maintenance.yml with three jobs:
- find-legacy-duplicates: Manual job to scan all open issues for duplicates
- check-stale-issues: Daily job to request confirmation on issues >60 days old
- close-unconfirmed-issues: Daily job to close issues without confirmation after 14 days
- Update triage to use gh CLI tools with database-specific labels (neo4j, falkordb, neptune)
- Separate deduplication into dedicated job using MCP GitHub tools
- Add "duplicate" label to both real-time and batch deduplication workflows
- Update claude-code-review.yml to use latest Sonnet model
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
Add NodeSummaryFilter callback parameter to extract_attributes_from_nodes
and extract_attributes_from_node functions, allowing consumers to
selectively skip summary regeneration for specific nodes.
This enables downstream applications to implement custom logic for
throttling or filtering which nodes should have summaries regenerated,
reducing unnecessary LLM calls and token costs.
Key changes:
- Add NodeSummaryFilter type alias: Callable[[EntityNode], Awaitable[bool]]
- Update extract_attributes_from_nodes with optional should_summarize_node parameter
- Update extract_attributes_from_node with conditional summary generation logic
- Add 5 comprehensive test cases covering callback functionality
- Maintain full backwards compatibility (default None = all summaries generated)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
chore: Update Claude review prompt to focus on critical feedback only
Added instruction to eliminate positive feedback from code reviews, reducing noise and focusing on actionable improvements.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
* fix: Fix typo in JSON entity extraction prompt
Change "an entities" to "any entities" in guideline 1 of the extract_json prompt.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Update graphiti_core/prompts/extract_nodes.py
Co-authored-by: Daniel Chalef <131175+danielchalef@users.noreply.github.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Daniel Chalef <131175+danielchalef@users.noreply.github.com>
* chore: Update edge extraction prompt to paraphrase instead of quote
- Changed instruction 5 to request paraphrasing rather than verbatim quoting
- Updated string quotes to use double quotes for consistency
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* chore: Format edge_operations.py and update lock file
- Minor formatting fix in edge_operations.py list comprehension
- Update uv.lock with version bump to 0.21.0rc8
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* fix: Prevent duplicate edge facts within same episode
This fixes three related bugs that allowed verbatim duplicate edge facts:
1. Fixed LLM deduplication: Changed related_edges_context to use integer
indices instead of UUIDs, matching the EdgeDuplicate model expectations.
2. Fixed batch deduplication: Removed episode skip in dedupe_edges_bulk
that prevented comparing edges from the same episode. Added self-comparison
guard to prevent edge from comparing against itself.
3. Added fast-path deduplication: Added exact string matching before parallel
processing in resolve_extracted_edges to catch within-episode duplicates
early, preventing race conditions where concurrent edges can't see each other.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* test: Add tests for edge deduplication fixes
Added three tests to verify the edge deduplication fixes:
1. test_dedupe_edges_bulk_deduplicates_within_episode: Verifies that
dedupe_edges_bulk now compares edges from the same episode after
removing the `if i == j: continue` check.
2. test_resolve_extracted_edge_uses_integer_indices_for_duplicates:
Validates that the LLM receives integer indices for duplicate
detection and correctly processes returned duplicate_facts.
3. test_resolve_extracted_edges_fast_path_deduplication: Confirms that
the fast-path exact string matching deduplicates identical edges
before parallel processing, preventing race conditions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Remove unused variables flagged by ruff
- Remove unused loop variable 'j' in bulk_utils.py
- Remove unused return value 'edges_by_episode' in test
- Replace unused 'edge_uuid' with '_' in test loop
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Replace MULTILINGUAL_EXTRACTION_RESPONSES constant with configurable
get_extraction_language_instruction() function to improve determinism
and allow customization.
Changes:
- Replace constant with function in client.py
- Update all LLM client implementations to use new function
- Maintain backward compatibility with same default behavior
- Enable users to override function for custom language requirements
Users can now customize extraction behavior by monkey-patching:
```python
import graphiti_core.llm_client.client as client
client.get_extraction_language_instruction = lambda: "Custom instruction"
```
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-authored-by: Claude <noreply@anthropic.com>
Add guideline to extract entities from all JSON properties, not just primary fields like name/user. This ensures comprehensive entity extraction while maintaining the existing exclusion of date properties.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-authored-by: Claude <noreply@anthropic.com>
* chore: Update dependencies and enhance edge resolution logic
- Add new dependencies: boto3, opensearch-py, and langchain-aws to pyproject.toml.
- Modify Graphiti class to handle additional parameters in edge resolution.
- Improve edge type handling in deduplication logic by introducing custom edge type names.
- Enhance tests for edge resolution to cover new scenarios and ensure correct behavior.
This update improves the flexibility and functionality of edge operations while ensuring compatibility with new libraries.
* refactor: Clean up test_edge_operations.py and format response returns
- Remove unnecessary stubs for opensearchpy module.
- Format return values in llm_client.generate_response for consistency.
- Enhance readability by ensuring proper indentation and structure in test cases.
This refactor improves the clarity and maintainability of the test suite for edge operations.
* bump version to 0.30.0pre5 and enhance docstring for resolve_extracted_edge function
- Update version in pyproject.toml to 0.30.0pre5.
- Add detailed docstring to resolve_extracted_edge function in edge_operations.py, clarifying parameters and return values.
This update improves documentation clarity for the edge resolution process.
* fix: Add edge type validation based on node labels
- Add DEFAULT_EDGE_NAME constant for 'RELATES_TO'
- Implement pre-resolution validation to reset invalid edge names
- Add post-resolution validation for LLM-returned fact types
- Rename parameter from edge_types to edge_type_candidates for clarity
- Add comprehensive tests for validation scenarios
This ensures edges conform to edge_type_map constraints and prevents
misclassification when edge types don't match node label pairs.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* chore: Bump version to 0.30.0pre4
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Refactor deduplication logic to enhance node resolution and track duplicate pairs (#929)
* Simplify deduplication process in bulk_utils by reusing canonical nodes.
* Update dedup_helpers to store duplicate pairs during resolution.
* Modify node_operations to append duplicate pairs when resolving nodes.
* Add tests to verify deduplication behavior and ensure correct state updates.
* reveret to concurrent dedup with fanout and then reconcilation
* add performance note for deduplication loop in bulk_utils
* enhance deduplication logic in bulk_utils to handle missing canonical nodes gracefully
* Update graphiti_core/utils/bulk_utils.py
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* refactor deduplication logic in bulk_utils to use directed union-find for canonical UUID resolution
* implement _build_directed_uuid_map for efficient UUID resolution in bulk_utils
* document directed union-find lookup in bulk_utils for clarity
---------
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
* add repository guidelines and project structure documentation
* update neo4j image version and modify test command to disable specific databases
* implement deduplication helpers and integrate with node operations
* refactor string formatting to use single quotes in node operations
* enhance deduplication helpers with UUID indexing and update resolution logic
* implement exact fact matching (#931)
