Cross encoder reranker in search query (#202)

* cross encoder reranker

* update reranker

* add openai reranker

* format

* mypy

* update

* updates

* MyPy typing

* bump version

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/graphiti.py

@@ -23,8 +23,11 @@
 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 @@ def __init__(
         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 @@ def __init__(
         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 @@ def __init__(
             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 @@ async def search(
             await search(
                 self.driver,
                 self.embedder,
+                self.cross_encoder,
                 query,
                 group_ids,
                 search_config,
@@ -663,8 +672,18 @@ async def _search(
         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 @@ async def get_nodes_by_query(
 
         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/search/search.py

@@ -21,6 +21,7 @@
 
 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
@@ -39,6 +40,7 @@
 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,
@@ -55,10 +57,12 @@
 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_vector = await embedder.create(input=[query.replace('\n', ' ')])
@@ -68,28 +72,34 @@ async def search(
     edges, nodes, communities = await asyncio.gather(
         edge_search(
             driver,
+            cross_encoder,
             query,
             query_vector,
             group_ids,
             config.edge_config,
             center_node_uuid,
+            bfs_origin_node_uuids,
             config.limit,
         ),
         node_search(
             driver,
+            cross_encoder,
             query,
             query_vector,
             group_ids,
             config.node_config,
             center_node_uuid,
+            bfs_origin_node_uuids,
             config.limit,
         ),
         community_search(
             driver,
+            cross_encoder,
             query,
             query_vector,
             group_ids,
             config.community_config,
+            bfs_origin_node_uuids,
             config.limit,
         ),
     )
@@ -109,11 +119,13 @@ async def search(
 
 async def edge_search(
     driver: AsyncDriver,
+    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:
@@ -126,6 +138,7 @@ 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),
             ]
         )
     )
@@ -146,6 +159,10 @@ 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:
+        fact_to_uuid_map = {edge.fact: edge.uuid for result in search_results for edge in result}
+        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')
@@ -176,11 +193,13 @@ async def edge_search(
 
 async def node_search(
     driver: AsyncDriver,
+    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:
@@ -212,6 +231,12 @@ 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:
+        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]
     elif config.reranker == NodeReranker.episode_mentions:
         reranked_uuids = await episode_mentions_reranker(driver, search_result_uuids)
     elif config.reranker == NodeReranker.node_distance:
@@ -228,10 +253,12 @@ async def node_search(
 
 async def community_search(
     driver: AsyncDriver,
+    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:
@@ -268,6 +295,12 @@ 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]