graphiti/graphiti_core/utils/maintenance/node_operations.py

"""
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 logging
from collections.abc import Awaitable, Callable
from time import time
from typing import Any

from pydantic import BaseModel

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, NodeResolutions
from graphiti_core.prompts.extract_nodes import (
    EntitySummary,
    ExtractedEntities,
    ExtractedEntity,
    MissedEntities,
)
from graphiti_core.search.search import search
from graphiti_core.search.search_config import SearchResults
from graphiti_core.search.search_config_recipes import NODE_HYBRID_SEARCH_RRF
from graphiti_core.search.search_filters import SearchFilters
from graphiti_core.utils.datetime_utils import utc_now
from graphiti_core.utils.maintenance.dedup_helpers import (
    DedupCandidateIndexes,
    DedupResolutionState,
    _build_candidate_indexes,
    _resolve_with_similarity,
)
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__)

NodeSummaryFilter = Callable[[EntityNode], Awaitable[bool]]


async def extract_nodes_reflexion(
    llm_client: LLMClient,
    episode: EpisodicNode,
    previous_episodes: list[EpisodicNode],
    node_names: list[str],
    group_id: str | None = None,
) -> list[str]:
    # Prepare context for LLM
    context = {
        'episode_content': episode.content,
        'previous_episodes': [ep.content for ep in previous_episodes],
        'extracted_entities': node_names,
    }

    llm_response = await llm_client.generate_response(
        prompt_library.extract_nodes.reflexion(context),
        MissedEntities,
        group_id=group_id,
    )
    missed_entities = llm_response.get('missed_entities', [])

    return missed_entities


async def extract_nodes(
    clients: GraphitiClients,
    episode: EpisodicNode,
    previous_episodes: list[EpisodicNode],
    entity_types: dict[str, type[BaseModel]] | None = None,
    excluded_entity_types: list[str] | None = None,
) -> list[EntityNode]:
    start = time()
    llm_client = clients.llm_client
    llm_response = {}
    custom_prompt = ''
    entities_missed = True
    reflexion_iterations = 0

    entity_types_context = [
        {
            'entity_type_id': 0,
            'entity_type_name': 'Entity',
            'entity_type_description': 'Default entity classification. Use this entity type if the entity is not one of the other listed types.',
        }
    ]

    entity_types_context += (
        [
            {
                'entity_type_id': i + 1,
                'entity_type_name': type_name,
                'entity_type_description': type_model.__doc__,
            }
            for i, (type_name, type_model) in enumerate(entity_types.items())
        ]
        if entity_types is not None
        else []
    )

    context = {
        'episode_content': episode.content,
        'episode_timestamp': episode.valid_at.isoformat(),
        'previous_episodes': [ep.content for ep in previous_episodes],
        'custom_prompt': custom_prompt,
        'entity_types': entity_types_context,
        'source_description': episode.source_description,
    }

    while entities_missed and reflexion_iterations <= MAX_REFLEXION_ITERATIONS:
        if episode.source == EpisodeType.message:
            llm_response = await llm_client.generate_response(
                prompt_library.extract_nodes.extract_message(context),
                response_model=ExtractedEntities,
                group_id=episode.group_id,
            )
        elif episode.source == EpisodeType.text:
            llm_response = await llm_client.generate_response(
                prompt_library.extract_nodes.extract_text(context),
                response_model=ExtractedEntities,
                group_id=episode.group_id,
            )
        elif episode.source == EpisodeType.json:
            llm_response = await llm_client.generate_response(
                prompt_library.extract_nodes.extract_json(context),
                response_model=ExtractedEntities,
                group_id=episode.group_id,
            )

        response_object = ExtractedEntities(**llm_response)

        extracted_entities: list[ExtractedEntity] = response_object.extracted_entities

        reflexion_iterations += 1
        if reflexion_iterations < MAX_REFLEXION_ITERATIONS:
            missing_entities = await extract_nodes_reflexion(
                llm_client,
                episode,
                previous_episodes,
                [entity.name for entity in extracted_entities],
                episode.group_id,
            )

            entities_missed = len(missing_entities) != 0

            custom_prompt = 'Make sure that the following entities are extracted: '
            for entity in missing_entities:
                custom_prompt += f'\n{entity},'

    filtered_extracted_entities = [entity for entity in extracted_entities if entity.name.strip()]
    end = time()
    logger.debug(f'Extracted new nodes: {filtered_extracted_entities} in {(end - start) * 1000} ms')
    # Convert the extracted data into EntityNode objects
    extracted_nodes = []
    for extracted_entity in filtered_extracted_entities:
        type_id = extracted_entity.entity_type_id
        if 0 <= type_id < len(entity_types_context):
            entity_type_name = entity_types_context[extracted_entity.entity_type_id].get(
                'entity_type_name'
            )
        else:
            entity_type_name = 'Entity'

        # Check if this entity type should be excluded
        if excluded_entity_types and entity_type_name in excluded_entity_types:
            logger.debug(f'Excluding entity "{extracted_entity.name}" of type "{entity_type_name}"')
            continue

        labels: list[str] = list({'Entity', str(entity_type_name)})

        new_node = EntityNode(
            name=extracted_entity.name,
            group_id=episode.group_id,
            labels=labels,
            summary='',
            created_at=utc_now(),
        )
        extracted_nodes.append(new_node)
        logger.debug(f'Created new node: {new_node.name} (UUID: {new_node.uuid})')

    logger.debug(f'Extracted nodes: {[(n.name, n.uuid) for n in extracted_nodes]}')

    return extracted_nodes


async def _collect_candidate_nodes(
    clients: GraphitiClients,
    extracted_nodes: list[EntityNode],
    existing_nodes_override: list[EntityNode] | None,
) -> list[EntityNode]:
    """Search per extracted name and return unique candidates with overrides honored in order."""
    search_results: list[SearchResults] = await semaphore_gather(
        *[
            search(
                clients=clients,
                query=node.name,
                group_ids=[node.group_id],
                search_filter=SearchFilters(),
                config=NODE_HYBRID_SEARCH_RRF,
            )
            for node in extracted_nodes
        ]
    )

    candidate_nodes: list[EntityNode] = [node for result in search_results for node in result.nodes]

    if existing_nodes_override is not None:
        candidate_nodes.extend(existing_nodes_override)

    seen_candidate_uuids: set[str] = set()
    ordered_candidates: list[EntityNode] = []
    for candidate in candidate_nodes:
        if candidate.uuid in seen_candidate_uuids:
            continue
        seen_candidate_uuids.add(candidate.uuid)
        ordered_candidates.append(candidate)

    return ordered_candidates


async def _resolve_with_llm(
    llm_client: LLMClient,
    extracted_nodes: list[EntityNode],
    indexes: DedupCandidateIndexes,
    state: DedupResolutionState,
    episode: EpisodicNode | None,
    previous_episodes: list[EpisodicNode] | None,
    entity_types: dict[str, type[BaseModel]] | None,
) -> None:
    """Escalate unresolved nodes to the dedupe prompt so the LLM can select or reject duplicates.

    The guardrails below defensively ignore malformed or duplicate LLM responses so the
    ingestion workflow remains deterministic even when the model misbehaves.
    """
    if not state.unresolved_indices:
        return

    entity_types_dict: dict[str, type[BaseModel]] = entity_types if entity_types is not None else {}

    llm_extracted_nodes = [extracted_nodes[i] for i in state.unresolved_indices]

    extracted_nodes_context = [
        {
            'id': i,
            'name': node.name,
            'entity_type': node.labels,
            'entity_type_description': entity_types_dict.get(
                next((item for item in node.labels if item != 'Entity'), '')
            ).__doc__
            or 'Default Entity Type',
        }
        for i, node in enumerate(llm_extracted_nodes)
    ]

    sent_ids = [ctx['id'] for ctx in extracted_nodes_context]
    logger.debug(
        'Sending %d entities to LLM for deduplication with IDs 0-%d (actual IDs sent: %s)',
        len(llm_extracted_nodes),
        len(llm_extracted_nodes) - 1,
        sent_ids if len(sent_ids) < 20 else f'{sent_ids[:10]}...{sent_ids[-10:]}',
    )
    if llm_extracted_nodes:
        sample_size = min(3, len(extracted_nodes_context))
        logger.debug(
            'First %d entities: %s',
            sample_size,
            [(ctx['id'], ctx['name']) for ctx in extracted_nodes_context[:sample_size]],
        )
        if len(extracted_nodes_context) > 3:
            logger.debug(
                'Last %d entities: %s',
                sample_size,
                [(ctx['id'], ctx['name']) for ctx in extracted_nodes_context[-sample_size:]],
            )

    existing_nodes_context = [
        {
            **{
                'idx': i,
                'name': candidate.name,
                'entity_types': candidate.labels,
            },
            **candidate.attributes,
        }
        for i, candidate in enumerate(indexes.existing_nodes)
    ]

    context = {
        'extracted_nodes': extracted_nodes_context,
        'existing_nodes': existing_nodes_context,
        '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 else []
        ),
    }

    llm_response = await llm_client.generate_response(
        prompt_library.dedupe_nodes.nodes(context),
        response_model=NodeResolutions,
    )

    node_resolutions: list[NodeDuplicate] = NodeResolutions(**llm_response).entity_resolutions

    valid_relative_range = range(len(state.unresolved_indices))
    processed_relative_ids: set[int] = set()

    received_ids = {r.id for r in node_resolutions}
    expected_ids = set(valid_relative_range)
    missing_ids = expected_ids - received_ids
    extra_ids = received_ids - expected_ids

    logger.debug(
        'Received %d resolutions for %d entities',
        len(node_resolutions),
        len(state.unresolved_indices),
    )

    if missing_ids:
        logger.warning('LLM did not return resolutions for IDs: %s', sorted(missing_ids))

    if extra_ids:
        logger.warning(
            'LLM returned invalid IDs outside valid range 0-%d: %s (all returned IDs: %s)',
            len(state.unresolved_indices) - 1,
            sorted(extra_ids),
            sorted(received_ids),
        )

    for resolution in node_resolutions:
        relative_id: int = resolution.id
        duplicate_idx: int = resolution.duplicate_idx

        if relative_id not in valid_relative_range:
            logger.warning(
                'Skipping invalid LLM dedupe id %d (valid range: 0-%d, received %d resolutions)',
                relative_id,
                len(state.unresolved_indices) - 1,
                len(node_resolutions),
            )
            continue

        if relative_id in processed_relative_ids:
            logger.warning('Duplicate LLM dedupe id %s received; ignoring.', relative_id)
            continue
        processed_relative_ids.add(relative_id)

        original_index = state.unresolved_indices[relative_id]
        extracted_node = extracted_nodes[original_index]

        resolved_node: EntityNode
        if duplicate_idx == -1:
            resolved_node = extracted_node
        elif 0 <= duplicate_idx < len(indexes.existing_nodes):
            resolved_node = indexes.existing_nodes[duplicate_idx]
        else:
            logger.warning(
                'Invalid duplicate_idx %s for extracted node %s; treating as no duplicate.',
                duplicate_idx,
                extracted_node.uuid,
            )
            resolved_node = extracted_node

        state.resolved_nodes[original_index] = resolved_node
        state.uuid_map[extracted_node.uuid] = resolved_node.uuid
        if resolved_node.uuid != extracted_node.uuid:
            state.duplicate_pairs.append((extracted_node, resolved_node))


async def resolve_extracted_nodes(
    clients: GraphitiClients,
    extracted_nodes: list[EntityNode],
    episode: EpisodicNode | None = None,
    previous_episodes: list[EpisodicNode] | None = None,
    entity_types: dict[str, type[BaseModel]] | None = None,
    existing_nodes_override: list[EntityNode] | None = None,
) -> tuple[list[EntityNode], dict[str, str], list[tuple[EntityNode, EntityNode]]]:
    """Search for existing nodes, resolve deterministic matches, then escalate holdouts to the LLM dedupe prompt."""
    llm_client = clients.llm_client
    driver = clients.driver
    existing_nodes = await _collect_candidate_nodes(
        clients,
        extracted_nodes,
        existing_nodes_override,
    )

    indexes: DedupCandidateIndexes = _build_candidate_indexes(existing_nodes)

    state = DedupResolutionState(
        resolved_nodes=[None] * len(extracted_nodes),
        uuid_map={},
        unresolved_indices=[],
    )

    _resolve_with_similarity(extracted_nodes, indexes, state)

    await _resolve_with_llm(
        llm_client,
        extracted_nodes,
        indexes,
        state,
        episode,
        previous_episodes,
        entity_types,
    )

    for idx, node in enumerate(extracted_nodes):
        if state.resolved_nodes[idx] is None:
            state.resolved_nodes[idx] = node
            state.uuid_map[node.uuid] = node.uuid

    logger.debug(
        'Resolved nodes: %s',
        [(node.name, node.uuid) for node in state.resolved_nodes if node is not None],
    )

    new_node_duplicates: list[
        tuple[EntityNode, EntityNode]
    ] = await filter_existing_duplicate_of_edges(driver, state.duplicate_pairs)

    return (
        [node for node in state.resolved_nodes if node is not None],
        state.uuid_map,
        new_node_duplicates,
    )


async def extract_attributes_from_nodes(
    clients: GraphitiClients,
    nodes: list[EntityNode],
    episode: EpisodicNode | None = None,
    previous_episodes: list[EpisodicNode] | None = None,
    entity_types: dict[str, type[BaseModel]] | None = None,
    should_summarize_node: NodeSummaryFilter | None = None,
) -> list[EntityNode]:
    llm_client = clients.llm_client
    embedder = clients.embedder
    updated_nodes: list[EntityNode] = await semaphore_gather(
        *[
            extract_attributes_from_node(
                llm_client,
                node,
                episode,
                previous_episodes,
                (
                    entity_types.get(next((item for item in node.labels if item != 'Entity'), ''))
                    if entity_types is not None
                    else None
                ),
                should_summarize_node,
            )
            for node in nodes
        ]
    )

    await create_entity_node_embeddings(embedder, updated_nodes)

    return updated_nodes


async def extract_attributes_from_node(
    llm_client: LLMClient,
    node: EntityNode,
    episode: EpisodicNode | None = None,
    previous_episodes: list[EpisodicNode] | None = None,
    entity_type: type[BaseModel] | None = None,
    should_summarize_node: NodeSummaryFilter | None = None,
) -> EntityNode:
    # Extract attributes if entity type is defined and has attributes
    llm_response = await _extract_entity_attributes(
        llm_client, node, episode, previous_episodes, entity_type
    )

    # Extract summary if needed
    await _extract_entity_summary(
        llm_client, node, episode, previous_episodes, should_summarize_node
    )

    node.attributes.update(llm_response)

    return node


async def _extract_entity_attributes(
    llm_client: LLMClient,
    node: EntityNode,
    episode: EpisodicNode | None,
    previous_episodes: list[EpisodicNode] | None,
    entity_type: type[BaseModel] | None,
) -> dict[str, Any]:
    if entity_type is None or len(entity_type.model_fields) == 0:
        return {}

    attributes_context = _build_episode_context(
        # should not include summary
        node_data={
            'name': node.name,
            'entity_types': node.labels,
            'attributes': node.attributes,
        },
        episode=episode,
        previous_episodes=previous_episodes,
    )

    llm_response = await llm_client.generate_response(
        prompt_library.extract_nodes.extract_attributes(attributes_context),
        response_model=entity_type,
        model_size=ModelSize.small,
        group_id=node.group_id,
    )

    # validate response
    entity_type(**llm_response)

    return llm_response


async def _extract_entity_summary(
    llm_client: LLMClient,
    node: EntityNode,
    episode: EpisodicNode | None,
    previous_episodes: list[EpisodicNode] | None,
    should_summarize_node: NodeSummaryFilter | None,
) -> None:
    if should_summarize_node is not None and not await should_summarize_node(node):
        return

    summary_context = _build_episode_context(
        node_data={
            'name': node.name,
            'summary': truncate_at_sentence(node.summary, MAX_SUMMARY_CHARS),
            'entity_types': node.labels,
            'attributes': node.attributes,
        },
        episode=episode,
        previous_episodes=previous_episodes,
    )

    summary_response = await llm_client.generate_response(
        prompt_library.extract_nodes.extract_summary(summary_context),
        response_model=EntitySummary,
        model_size=ModelSize.small,
        group_id=node.group_id,
    )

    node.summary = truncate_at_sentence(summary_response.get('summary', ''), MAX_SUMMARY_CHARS)


def _build_episode_context(
    node_data: dict[str, Any],
    episode: EpisodicNode | None,
    previous_episodes: list[EpisodicNode] | None,
) -> dict[str, Any]:
    return {
        'node': node_data,
        '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 else []
        ),
    }