graphiti-core 0.3.16__tar.gz → 0.3.18__tar.gz

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: graphiti-core
-Version: 0.3.16
+Version: 0.3.18
 Summary: A temporal graph building library
 License: Apache-2.0
 Author: Paul Paliychuk
@@ -14,10 +14,9 @@ Classifier: Programming Language :: Python :: 3.12
 Requires-Dist: diskcache (>=5.6.3,<6.0.0)
 Requires-Dist: neo4j (>=5.23.0,<6.0.0)
 Requires-Dist: numpy (>=1.0.0)
-Requires-Dist: openai (>=1.50.2,<2.0.0)
+Requires-Dist: openai (>=1.52.2,<2.0.0)
 Requires-Dist: pydantic (>=2.8.2,<3.0.0)
 Requires-Dist: tenacity (<9.0.0)
-Requires-Dist: voyageai (>=0.2.3,<0.3.0)
 Description-Content-Type: text/markdown
 
 <div align="center">

graphiti_core-0.3.18/graphiti_core/cross_encoder/bge_reranker_client.py

@@ -0,0 +1,45 @@
+"""
+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 asyncio
+from typing import List, Tuple
+
+from sentence_transformers import CrossEncoder
+
+from graphiti_core.cross_encoder.client import CrossEncoderClient
+
+
+class BGERerankerClient(CrossEncoderClient):
+    def __init__(self):
+        self.model = CrossEncoder('BAAI/bge-reranker-v2-m3')
+
+    async def rank(self, query: str, passages: List[str]) -> List[Tuple[str, float]]:
+        if not passages:
+            return []
+
+        input_pairs = [[query, passage] for passage in passages]
+
+        # Run the synchronous predict method in an executor
+        loop = asyncio.get_running_loop()
+        scores = await loop.run_in_executor(None, self.model.predict, input_pairs)
+
+        ranked_passages = sorted(
+            [(passage, float(score)) for passage, score in zip(passages, scores)],
+            key=lambda x: x[1],
+            reverse=True,
+        )
+
+        return ranked_passages

graphiti_core-0.3.18/graphiti_core/cross_encoder/client.py

@@ -0,0 +1,41 @@
+"""
+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 abc import ABC, abstractmethod
+from typing import List, Tuple
+
+
+class CrossEncoderClient(ABC):
+    """
+    CrossEncoderClient is an abstract base class that defines the interface
+    for cross-encoder models used for ranking passages based on their relevance to a query.
+    It allows for different implementations of cross-encoder models to be used interchangeably.
+    """
+
+    @abstractmethod
+    async def rank(self, query: str, passages: List[str]) -> List[Tuple[str, float]]:
+        """
+        Rank the given passages based on their relevance to the query.
+
+        Args:
+            query (str): The query string.
+            passages (List[str]): A list of passages to rank.
+
+        Returns:
+            List[Tuple[str, float]]: A list of tuples containing the passage and its score,
+                                     sorted in descending order of relevance.
+        """
+        pass

graphiti_core-0.3.18/graphiti_core/cross_encoder/openai_reranker_client.py

@@ -0,0 +1,113 @@
+"""
+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 asyncio
+import logging
+from typing import Any
+
+import openai
+from openai import AsyncOpenAI
+from pydantic import BaseModel
+
+from ..llm_client import LLMConfig, RateLimitError
+from ..prompts import Message
+from .client import CrossEncoderClient
+
+logger = logging.getLogger(__name__)
+
+DEFAULT_MODEL = 'gpt-4o-mini'
+
+
+class BooleanClassifier(BaseModel):
+    isTrue: bool
+
+
+class OpenAIRerankerClient(CrossEncoderClient):
+    def __init__(self, config: LLMConfig | None = None):
+        """
+        Initialize the OpenAIClient with the provided configuration, cache setting, and client.
+
+        Args:
+            config (LLMConfig | None): The configuration for the LLM client, including API key, model, base URL, temperature, and max tokens.
+            cache (bool): Whether to use caching for responses. Defaults to False.
+            client (Any | None): An optional async client instance to use. If not provided, a new AsyncOpenAI client is created.
+
+        """
+        if config is None:
+            config = LLMConfig()
+
+        self.config = config
+        self.client = AsyncOpenAI(api_key=config.api_key, base_url=config.base_url)
+
+    async def rank(self, query: str, passages: list[str]) -> list[tuple[str, float]]:
+        openai_messages_list: Any = [
+            [
+                Message(
+                    role='system',
+                    content='You are an expert tasked with determining whether the passage is relevant to the query',
+                ),
+                Message(
+                    role='user',
+                    content=f"""
+                           Respond with "True" if PASSAGE is relevant to QUERY and "False" otherwise. 
+                           <PASSAGE>
+                           {query}
+                           </PASSAGE>
+                           {passage}
+                           <QUERY>
+                           </QUERY>
+                           """,
+                ),
+            ]
+            for passage in passages
+        ]
+        try:
+            responses = await asyncio.gather(
+                *[
+                    self.client.chat.completions.create(
+                        model=DEFAULT_MODEL,
+                        messages=openai_messages,
+                        temperature=0,
+                        max_tokens=1,
+                        logit_bias={'6432': 1, '7983': 1},
+                        logprobs=True,
+                        top_logprobs=2,
+                    )
+                    for openai_messages in openai_messages_list
+                ]
+            )
+
+            responses_top_logprobs = [
+                response.choices[0].logprobs.content[0].top_logprobs
+                if response.choices[0].logprobs is not None
+                and response.choices[0].logprobs.content is not None
+                else []
+                for response in responses
+            ]
+            scores: list[float] = []
+            for top_logprobs in responses_top_logprobs:
+                for logprob in top_logprobs:
+                    if bool(logprob.token):
+                        scores.append(logprob.logprob)
+
+            results = [(passage, score) for passage, score in zip(passages, scores)]
+            results.sort(reverse=True, key=lambda x: x[1])
+            return results
+        except openai.RateLimitError as e:
+            raise RateLimitError from e
+        except Exception as e:
+            logger.error(f'Error in generating LLM response: {e}')
+            raise

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/embedder/openai.py

@@ -42,7 +42,9 @@ class OpenAIEmbedder(EmbedderClient):
         self.client = AsyncOpenAI(api_key=config.api_key, base_url=config.base_url)
 
     async def create(
-        self, input: str | List[str] | Iterable[int] | Iterable[Iterable[int]]
+        self, input_data: str | List[str] | Iterable[int] | Iterable[Iterable[int]]
     ) -> list[float]:
-        result = await self.client.embeddings.create(input=input, model=self.config.embedding_model)
+        result = await self.client.embeddings.create(
+            input=input_data, model=self.config.embedding_model
+        )
         return result.data[0].embedding[: self.config.embedding_dim]

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/embedder/voyage.py

@@ -41,7 +41,18 @@ class VoyageAIEmbedder(EmbedderClient):
         self.client = voyageai.AsyncClient(api_key=config.api_key)
 
     async def create(
-        self, input: str | List[str] | Iterable[int] | Iterable[Iterable[int]]
+        self, input_data: str | List[str] | Iterable[int] | Iterable[Iterable[int]]
     ) -> list[float]:
-        result = await self.client.embed(input, model=self.config.embedding_model)
+        if isinstance(input_data, str):
+            input_list = [input_data]
+        elif isinstance(input_data, List):
+            input_list = [str(i) for i in input_data if i]
+        else:
+            input_list = [str(i) for i in input_data if i is not None]
+
+        input_list = [i for i in input_list if i]
+        if len(input_list) == 0:
+            return []
+
+        result = await self.client.embed(input_list, model=self.config.embedding_model)
         return result.embeddings[0][: self.config.embedding_dim]

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/graphiti.py

@@ -23,8 +23,11 @@ from dotenv import load_dotenv
 from neo4j import AsyncGraphDatabase
 from pydantic import BaseModel
 
+from graphiti_core.cross_encoder.client import CrossEncoderClient
+from graphiti_core.cross_encoder.openai_reranker_client import OpenAIRerankerClient
 from graphiti_core.edges import EntityEdge, EpisodicEdge
 from graphiti_core.embedder import EmbedderClient, OpenAIEmbedder
+from graphiti_core.helpers import DEFAULT_DATABASE
 from graphiti_core.llm_client import LLMClient, OpenAIClient
 from graphiti_core.nodes import CommunityNode, EntityNode, EpisodeType, EpisodicNode
 from graphiti_core.search.search import SearchConfig, search
@@ -92,6 +95,7 @@ class Graphiti:
         password: str,
         llm_client: LLMClient | None = None,
         embedder: EmbedderClient | None = None,
+        cross_encoder: CrossEncoderClient | None = None,
         store_raw_episode_content: bool = True,
     ):
         """
@@ -131,7 +135,7 @@ class Graphiti:
         Graphiti if you're using the default OpenAIClient.
         """
         self.driver = AsyncGraphDatabase.driver(uri, auth=(user, password))
-        self.database = 'neo4j'
+        self.database = DEFAULT_DATABASE
         self.store_raw_episode_content = store_raw_episode_content
         if llm_client:
             self.llm_client = llm_client
@@ -141,6 +145,10 @@ class Graphiti:
             self.embedder = embedder
         else:
             self.embedder = OpenAIEmbedder()
+        if cross_encoder:
+            self.cross_encoder = cross_encoder
+        else:
+            self.cross_encoder = OpenAIRerankerClient()
 
     async def close(self):
         """
@@ -648,6 +656,7 @@ class Graphiti:
             await search(
                 self.driver,
                 self.embedder,
+                self.cross_encoder,
                 query,
                 group_ids,
                 search_config,
@@ -663,8 +672,18 @@ class Graphiti:
         config: SearchConfig,
         group_ids: list[str] | None = None,
         center_node_uuid: str | None = None,
+        bfs_origin_node_uuids: list[str] | None = None,
     ) -> SearchResults:
-        return await search(self.driver, self.embedder, query, group_ids, config, center_node_uuid)
+        return await search(
+            self.driver,
+            self.embedder,
+            self.cross_encoder,
+            query,
+            group_ids,
+            config,
+            center_node_uuid,
+            bfs_origin_node_uuids,
+        )
 
     async def get_nodes_by_query(
         self,
@@ -716,7 +735,13 @@ class Graphiti:
 
         nodes = (
             await search(
-                self.driver, self.embedder, query, group_ids, search_config, center_node_uuid
+                self.driver,
+                self.embedder,
+                self.cross_encoder,
+                query,
+                group_ids,
+                search_config,
+                center_node_uuid,
             )
         ).nodes
         return nodes

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/search/search.py

@@ -21,6 +21,7 @@ from time import time
 
 from neo4j import AsyncDriver
 
+from graphiti_core.cross_encoder.client import CrossEncoderClient
 from graphiti_core.edges import EntityEdge
 from graphiti_core.embedder import EmbedderClient
 from graphiti_core.errors import SearchRerankerError
@@ -39,10 +40,12 @@ from graphiti_core.search.search_config import (
 from graphiti_core.search.search_utils import (
     community_fulltext_search,
     community_similarity_search,
+    edge_bfs_search,
     edge_fulltext_search,
     edge_similarity_search,
     episode_mentions_reranker,
     maximal_marginal_relevance,
+    node_bfs_search,
     node_distance_reranker,
     node_fulltext_search,
     node_similarity_search,
@@ -55,40 +58,49 @@ logger = logging.getLogger(__name__)
 async def search(
     driver: AsyncDriver,
     embedder: EmbedderClient,
+    cross_encoder: CrossEncoderClient,
     query: str,
     group_ids: list[str] | None,
     config: SearchConfig,
     center_node_uuid: str | None = None,
+    bfs_origin_node_uuids: list[str] | None = None,
 ) -> SearchResults:
     start = time()
-    query = query.replace('\n', ' ')
+    query_vector = await embedder.create(input=[query.replace('\n', ' ')])
+
     # if group_ids is empty, set it to None
     group_ids = group_ids if group_ids else None
     edges, nodes, communities = await asyncio.gather(
         edge_search(
             driver,
-            embedder,
+            cross_encoder,
             query,
+            query_vector,
             group_ids,
             config.edge_config,
             center_node_uuid,
+            bfs_origin_node_uuids,
             config.limit,
         ),
         node_search(
             driver,
-            embedder,
+            cross_encoder,
             query,
+            query_vector,
             group_ids,
             config.node_config,
             center_node_uuid,
+            bfs_origin_node_uuids,
             config.limit,
         ),
         community_search(
             driver,
-            embedder,
+            cross_encoder,
             query,
+            query_vector,
             group_ids,
             config.community_config,
+            bfs_origin_node_uuids,
             config.limit,
         ),
     )
@@ -99,27 +111,27 @@ async def search(
         communities=communities,
     )
 
-    end = time()
+    latency = (time() - start) * 1000
 
-    logger.info(f'search returned context for query {query} in {(end - start) * 1000} ms')
+    logger.debug(f'search returned context for query {query} in {latency} ms')
 
     return results
 
 
 async def edge_search(
     driver: AsyncDriver,
-    embedder: EmbedderClient,
+    cross_encoder: CrossEncoderClient,
     query: str,
+    query_vector: list[float],
     group_ids: list[str] | None,
     config: EdgeSearchConfig | None,
     center_node_uuid: str | None = None,
+    bfs_origin_node_uuids: list[str] | None = None,
     limit=DEFAULT_SEARCH_LIMIT,
 ) -> list[EntityEdge]:
     if config is None:
         return []
 
-    query_vector = await embedder.create(input=[query])
-
     search_results: list[list[EntityEdge]] = list(
         await asyncio.gather(
             *[
@@ -127,6 +139,7 @@ async def edge_search(
                 edge_similarity_search(
                     driver, query_vector, None, None, group_ids, 2 * limit, config.sim_min_score
                 ),
+                edge_bfs_search(driver, bfs_origin_node_uuids, config.bfs_max_depth, 2 * limit),
             ]
         )
     )
@@ -147,6 +160,15 @@ async def edge_search(
         reranked_uuids = maximal_marginal_relevance(
             query_vector, search_result_uuids_and_vectors, config.mmr_lambda
         )
+    elif config.reranker == EdgeReranker.cross_encoder:
+        search_result_uuids = [[edge.uuid for edge in result] for result in search_results]
+
+        rrf_result_uuids = rrf(search_result_uuids)
+        rrf_edges = [edge_uuid_map[uuid] for uuid in rrf_result_uuids][:limit]
+
+        fact_to_uuid_map = {edge.fact: edge.uuid for edge in rrf_edges}
+        reranked_facts = await cross_encoder.rank(query, list(fact_to_uuid_map.keys()))
+        reranked_uuids = [fact_to_uuid_map[fact] for fact, _ in reranked_facts]
     elif config.reranker == EdgeReranker.node_distance:
         if center_node_uuid is None:
             raise SearchRerankerError('No center node provided for Node Distance reranker')
@@ -177,18 +199,18 @@ async def edge_search(
 
 async def node_search(
     driver: AsyncDriver,
-    embedder: EmbedderClient,
+    cross_encoder: CrossEncoderClient,
     query: str,
+    query_vector: list[float],
     group_ids: list[str] | None,
     config: NodeSearchConfig | None,
     center_node_uuid: str | None = None,
+    bfs_origin_node_uuids: list[str] | None = None,
     limit=DEFAULT_SEARCH_LIMIT,
 ) -> list[EntityNode]:
     if config is None:
         return []
 
-    query_vector = await embedder.create(input=[query])
-
     search_results: list[list[EntityNode]] = list(
         await asyncio.gather(
             *[
@@ -196,6 +218,7 @@ async def node_search(
                 node_similarity_search(
                     driver, query_vector, group_ids, 2 * limit, config.sim_min_score
                 ),
+                node_bfs_search(driver, bfs_origin_node_uuids, config.bfs_max_depth, 2 * limit),
             ]
         )
     )
@@ -215,6 +238,15 @@ async def node_search(
         reranked_uuids = maximal_marginal_relevance(
             query_vector, search_result_uuids_and_vectors, config.mmr_lambda
         )
+    elif config.reranker == NodeReranker.cross_encoder:
+        # use rrf as a preliminary reranker
+        rrf_result_uuids = rrf(search_result_uuids)
+        rrf_results = [node_uuid_map[uuid] for uuid in rrf_result_uuids][:limit]
+
+        summary_to_uuid_map = {node.summary: node.uuid for node in rrf_results}
+
+        reranked_summaries = await cross_encoder.rank(query, list(summary_to_uuid_map.keys()))
+        reranked_uuids = [summary_to_uuid_map[fact] for fact, _ in reranked_summaries]
     elif config.reranker == NodeReranker.episode_mentions:
         reranked_uuids = await episode_mentions_reranker(driver, search_result_uuids)
     elif config.reranker == NodeReranker.node_distance:
@@ -231,17 +263,17 @@ async def node_search(
 
 async def community_search(
     driver: AsyncDriver,
-    embedder: EmbedderClient,
+    cross_encoder: CrossEncoderClient,
     query: str,
+    query_vector: list[float],
     group_ids: list[str] | None,
     config: CommunitySearchConfig | None,
+    bfs_origin_node_uuids: list[str] | None = None,
     limit=DEFAULT_SEARCH_LIMIT,
 ) -> list[CommunityNode]:
     if config is None:
         return []
 
-    query_vector = await embedder.create(input=[query])
-
     search_results: list[list[CommunityNode]] = list(
         await asyncio.gather(
             *[
@@ -273,6 +305,12 @@ async def community_search(
         reranked_uuids = maximal_marginal_relevance(
             query_vector, search_result_uuids_and_vectors, config.mmr_lambda
         )
+    elif config.reranker == CommunityReranker.cross_encoder:
+        summary_to_uuid_map = {
+            node.summary: node.uuid for result in search_results for node in result
+        }
+        reranked_summaries = await cross_encoder.rank(query, list(summary_to_uuid_map.keys()))
+        reranked_uuids = [summary_to_uuid_map[fact] for fact, _ in reranked_summaries]
 
     reranked_communities = [community_uuid_map[uuid] for uuid in reranked_uuids]
 

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/search/search_config.py

@@ -20,7 +20,11 @@ from pydantic import BaseModel, Field
 
 from graphiti_core.edges import EntityEdge
 from graphiti_core.nodes import CommunityNode, EntityNode
-from graphiti_core.search.search_utils import DEFAULT_MIN_SCORE, DEFAULT_MMR_LAMBDA
+from graphiti_core.search.search_utils import (
+    DEFAULT_MIN_SCORE,
+    DEFAULT_MMR_LAMBDA,
+    MAX_SEARCH_DEPTH,
+)
 
 DEFAULT_SEARCH_LIMIT = 10
 
@@ -28,11 +32,13 @@ DEFAULT_SEARCH_LIMIT = 10
 class EdgeSearchMethod(Enum):
     cosine_similarity = 'cosine_similarity'
     bm25 = 'bm25'
+    bfs = 'breadth_first_search'
 
 
 class NodeSearchMethod(Enum):
     cosine_similarity = 'cosine_similarity'
     bm25 = 'bm25'
+    bfs = 'breadth_first_search'
 
 
 class CommunitySearchMethod(Enum):
@@ -45,6 +51,7 @@ class EdgeReranker(Enum):
     node_distance = 'node_distance'
     episode_mentions = 'episode_mentions'
     mmr = 'mmr'
+    cross_encoder = 'cross_encoder'
 
 
 class NodeReranker(Enum):
@@ -52,11 +59,13 @@ class NodeReranker(Enum):
     node_distance = 'node_distance'
     episode_mentions = 'episode_mentions'
     mmr = 'mmr'
+    cross_encoder = 'cross_encoder'
 
 
 class CommunityReranker(Enum):
     rrf = 'reciprocal_rank_fusion'
     mmr = 'mmr'
+    cross_encoder = 'cross_encoder'
 
 
 class EdgeSearchConfig(BaseModel):
@@ -64,6 +73,7 @@ class EdgeSearchConfig(BaseModel):
     reranker: EdgeReranker = Field(default=EdgeReranker.rrf)
     sim_min_score: float = Field(default=DEFAULT_MIN_SCORE)
     mmr_lambda: float = Field(default=DEFAULT_MMR_LAMBDA)
+    bfs_max_depth: int = Field(default=MAX_SEARCH_DEPTH)
 
 
 class NodeSearchConfig(BaseModel):
@@ -71,6 +81,7 @@ class NodeSearchConfig(BaseModel):
     reranker: NodeReranker = Field(default=NodeReranker.rrf)
     sim_min_score: float = Field(default=DEFAULT_MIN_SCORE)
     mmr_lambda: float = Field(default=DEFAULT_MMR_LAMBDA)
+    bfs_max_depth: int = Field(default=MAX_SEARCH_DEPTH)
 
 
 class CommunitySearchConfig(BaseModel):
@@ -78,6 +89,7 @@ class CommunitySearchConfig(BaseModel):
     reranker: CommunityReranker = Field(default=CommunityReranker.rrf)
     sim_min_score: float = Field(default=DEFAULT_MIN_SCORE)
     mmr_lambda: float = Field(default=DEFAULT_MMR_LAMBDA)
+    bfs_max_depth: int = Field(default=MAX_SEARCH_DEPTH)
 
 
 class SearchConfig(BaseModel):

{graphiti_core-0.3.16 → graphiti_core-0.3.18}/graphiti_core/search/search_config_recipes.py

@@ -48,14 +48,41 @@ COMBINED_HYBRID_SEARCH_MMR = SearchConfig(
     edge_config=EdgeSearchConfig(
         search_methods=[EdgeSearchMethod.bm25, EdgeSearchMethod.cosine_similarity],
         reranker=EdgeReranker.mmr,
+        mmr_lambda=1,
     ),
     node_config=NodeSearchConfig(
         search_methods=[NodeSearchMethod.bm25, NodeSearchMethod.cosine_similarity],
         reranker=NodeReranker.mmr,
+        mmr_lambda=1,
     ),
     community_config=CommunitySearchConfig(
         search_methods=[CommunitySearchMethod.bm25, CommunitySearchMethod.cosine_similarity],
         reranker=CommunityReranker.mmr,
+        mmr_lambda=1,
+    ),
+)
+
+# Performs a full-text search, similarity search, and bfs with cross_encoder reranking over edges, nodes, and communities
+COMBINED_HYBRID_SEARCH_CROSS_ENCODER = SearchConfig(
+    edge_config=EdgeSearchConfig(
+        search_methods=[
+            EdgeSearchMethod.bm25,
+            EdgeSearchMethod.cosine_similarity,
+            EdgeSearchMethod.bfs,
+        ],
+        reranker=EdgeReranker.cross_encoder,
+    ),
+    node_config=NodeSearchConfig(
+        search_methods=[
+            NodeSearchMethod.bm25,
+            NodeSearchMethod.cosine_similarity,
+            NodeSearchMethod.bfs,
+        ],
+        reranker=NodeReranker.cross_encoder,
+    ),
+    community_config=CommunitySearchConfig(
+        search_methods=[CommunitySearchMethod.bm25, CommunitySearchMethod.cosine_similarity],
+        reranker=CommunityReranker.cross_encoder,
     ),
 )
 
@@ -81,7 +108,6 @@ EDGE_HYBRID_SEARCH_NODE_DISTANCE = SearchConfig(
         search_methods=[EdgeSearchMethod.bm25, EdgeSearchMethod.cosine_similarity],
         reranker=EdgeReranker.node_distance,
     ),
-    limit=30,
 )
 
 # performs a hybrid search over edges with episode mention reranking