bisheng-langchain 0.3.5.dev1__py3-none-any.whl → 0.3.7__py3-none-any.whl

bisheng_langchain/memory/__init__.py

@@ -0,0 +1,3 @@
+from .redis import ConversationRedisMemory
+
+__all__ = ['ConversationRedisMemory']

bisheng_langchain/memory/redis.py

@@ -0,0 +1,104 @@
+import json
+from typing import Any, Dict, List, Optional
+
+import redis
+from langchain.memory.chat_memory import BaseChatMemory
+from langchain_core.messages import (AIMessage, BaseMessage, HumanMessage, get_buffer_string,
+                                     message_to_dict, messages_from_dict)
+from langchain_core.pydantic_v1 import root_validator
+from pydantic import Field
+
+
+class ConversationRedisMemory(BaseChatMemory):
+    """Using redis for storing conversation memory."""
+    redis_client: redis.Redis = Field(default=None, exclude=True)
+    human_prefix: str = 'Human'
+    ai_prefix: str = 'AI'
+    session_id: str = 'session'
+    memory_key: str = 'history'  #: :meta private:
+    redis_url: str
+    redis_prefix: str = 'redis_buffer_'
+    ttl: Optional[int] = None
+
+    @root_validator()
+    def validate_environment(cls, values: Dict) -> Dict:
+        redis_url = values.get('redis_url')
+        if not redis_url:
+            raise ValueError('Redis URL must be set')
+        pool = redis.ConnectionPool.from_url(redis_url, max_connections=1)
+        values['redis_client'] = redis.StrictRedis(connection_pool=pool)
+        return values
+
+    @property
+    def buffer(self) -> Any:
+        """String buffer of memory."""
+        return self.buffer_as_messages if self.return_messages else self.buffer_as_str
+
+    async def abuffer(self) -> Any:
+        """String buffer of memory."""
+        return (await self.abuffer_as_messages()
+                if self.return_messages else await self.abuffer_as_str())
+
+    def _buffer_as_str(self, messages: List[BaseMessage]) -> str:
+        return get_buffer_string(
+            messages,
+            human_prefix=self.human_prefix,
+            ai_prefix=self.ai_prefix,
+        )
+
+    @property
+    def buffer_as_str(self) -> str:
+        """Exposes the buffer as a string in case return_messages is True."""
+        messages = self.buffer_as_messages
+        return self._buffer_as_str(messages)
+
+        # return self._buffer_as_str(self.chat_memory.messages)
+
+    async def abuffer_as_str(self) -> str:
+        """Exposes the buffer as a string in case return_messages is True."""
+        # messages = await self.chat_memory.aget_messages()
+        messages = self.buffer_as_messages
+        return self._buffer_as_str(messages)
+
+    @property
+    def buffer_as_messages(self) -> List[BaseMessage]:
+        """Exposes the buffer as a list of messages in case return_messages is False."""
+        # return self.chat_memory.messages
+        redis_value = self.redis_client.lrange(self.redis_prefix + self.session_id, 0, -1)
+        items = [json.loads(m.decode('utf-8')) for m in redis_value[::-1]]
+        messages = messages_from_dict(items)
+        return messages
+
+    async def abuffer_as_messages(self) -> List[BaseMessage]:
+        """Exposes the buffer as a list of messages in case return_messages is False."""
+        self.buffer_as_messages
+
+    @property
+    def memory_variables(self) -> List[str]:
+        """Will always return list of memory variables.
+
+        :meta private:
+        """
+        return [self.memory_key]
+
+    def load_memory_variables(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
+        """Return history buffer."""
+        return {self.memory_key: self.buffer}
+
+    async def aload_memory_variables(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
+        """Return key-value pairs given the text input to the chain."""
+        buffer = await self.abuffer()
+        return {self.memory_key: buffer}
+
+    def save_context(self, inputs: Dict[str, Any], outputs: Dict[str, str]) -> None:
+        """Save context from this conversation to buffer."""
+        input_str, output_str = self._get_input_output(inputs, outputs)
+
+        input_message_str = json.dumps(message_to_dict(HumanMessage(content=input_str)),
+                                       ensure_ascii=False)
+        output_message_str = json.dumps(message_to_dict(AIMessage(content=output_str)),
+                                        ensure_ascii=False)
+        self.redis_client.lpush(self.redis_prefix + self.session_id, input_message_str)
+        self.redis_client.lpush(self.redis_prefix + self.session_id, output_message_str)
+        if self.ttl:
+            self.redis_client.expire(self.redis_prefix + self.session_id, self.ttl)

bisheng_langchain/rag/init_retrievers/keyword_retriever.py

@@ -16,7 +16,7 @@ from langchain.text_splitter import TextSplitter
 
 
 class KeywordRetriever(BaseRetriever):
-    keyword_store: VectorStore
+    keyword_store: ElasticKeywordsSearch
     text_splitter: TextSplitter
     search_type: str = 'similarity'
     search_kwargs: dict = Field(default_factory=dict)

bisheng_langchain/vectorstores/elastic_keywords_search.py

@@ -13,6 +13,7 @@ from langchain.llms.base import BaseLLM
 from langchain.prompts.prompt import PromptTemplate
 from langchain.utils import get_from_dict_or_env
 from langchain.vectorstores.base import VectorStore
+from loguru import logger
 
 if TYPE_CHECKING:
     from elasticsearch import Elasticsearch  # noqa: F401
@@ -326,6 +327,49 @@ class ElasticKeywordsSearch(VectorStore, ABC):
             response = client.search(index=index_name, body={'query': script_query, 'size': size})
         return response
 
-    def delete(self, **kwargs: Any) -> None:
+    def delete_index(self, **kwargs: Any) -> None:
         # TODO: Check if this can be done in bulk
         self.client.indices.delete(index=self.index_name)
+
+    def delete(
+        self,
+        ids: Optional[List[str]] = None,
+        refresh_indices: Optional[bool] = True,
+        **kwargs: Any,
+    ) -> Optional[bool]:
+        """Delete documents from the Elasticsearch index.
+
+        Args:
+            ids: List of ids of documents to delete.
+            refresh_indices: Whether to refresh the index
+                            after deleting documents. Defaults to True.
+        """
+        try:
+            from elasticsearch.helpers import BulkIndexError, bulk
+        except ImportError:
+            raise ImportError('Could not import elasticsearch python package. '
+                              'Please install it with `pip install elasticsearch`.')
+
+        body = []
+
+        if ids is None:
+            raise ValueError('ids must be provided.')
+
+        for _id in ids:
+            body.append({'_op_type': 'delete', '_index': self.index_name, '_id': _id})
+
+        if len(body) > 0:
+            try:
+                bulk(self.client, body, refresh=refresh_indices, ignore_status=404)
+                logger.debug(f'Deleted {len(body)} texts from index')
+
+                return True
+            except BulkIndexError as e:
+                logger.error(f'Error deleting texts: {e}')
+                firstError = e.errors[0].get('index', {}).get('error', {})
+                logger.error(f"First error reason: {firstError.get('reason')}")
+                raise e
+
+        else:
+            logger.debug('No texts to delete from index')
+            return False

{bisheng_langchain-0.3.5.dev1.dist-info → bisheng_langchain-0.3.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: bisheng-langchain
-Version: 0.3.5.dev1
+Version: 0.3.7
 Summary: bisheng langchain modules
 Home-page: https://github.com/dataelement/bisheng
 Author: DataElem

{bisheng_langchain-0.3.5.dev1.dist-info → bisheng_langchain-0.3.7.dist-info}/RECORD

@@ -108,6 +108,8 @@ bisheng_langchain/gpts/tools/get_current_time/tool.py,sha256=3uvk7Yu07qhZy1sBrFM
 bisheng_langchain/input_output/__init__.py,sha256=sW_GB7MlrHYsqY1Meb_LeimQqNsMz1gH-00Tqb2BUyM,153
 bisheng_langchain/input_output/input.py,sha256=I5YDmgbvvj1o2lO9wi8LE37wM0wP5jkhUREU32YrZMQ,1094
 bisheng_langchain/input_output/output.py,sha256=6U-az6-Cwz665C2YmcH3SYctWVjPFjmW8s70CA_qphk,11585
+bisheng_langchain/memory/__init__.py,sha256=TNqe5l5BqUv4wh3_UH28fYPWQXGLBUYn6QJHsr7vanI,82
+bisheng_langchain/memory/redis.py,sha256=paz72ic5BfLXY6lj2cEbCxrTb8KVMnKMZmG9q7uh_9s,4291
 bisheng_langchain/rag/__init__.py,sha256=Rm_cDxOJINt0H4bOeUo3JctPxaI6xKKXZcS-R_wkoGs,198
 bisheng_langchain/rag/bisheng_rag_chain.py,sha256=2GMDUPJaW-D7tpOQ9qPt2vGZwmcXBS0UrcibO7J2S1g,5999
 bisheng_langchain/rag/bisheng_rag_pipeline.py,sha256=neoBK3TtuQ07_WeuJCzYlvtsDQNepUa_68NT8VCgytw,13749
@@ -126,7 +128,7 @@ bisheng_langchain/rag/config/baseline_s2b_mix.yaml,sha256=rkPfzU2-mvjRrZ0zMHaQsn
 bisheng_langchain/rag/config/baseline_v2.yaml,sha256=RP-DwIRIS_ZK8ixbXi2Z28rKqHD56pWmr2o2WWIwq3Y,2382
 bisheng_langchain/rag/init_retrievers/__init__.py,sha256=qpLLAuqZPtumTlJj17Ie5AbDDmiUiDxYefg_pumqu-c,218
 bisheng_langchain/rag/init_retrievers/baseline_vector_retriever.py,sha256=oRKZZpxlLQAtsubIcAXeXpf1a9h6Pt6uOtNTLeD2jps,2362
-bisheng_langchain/rag/init_retrievers/keyword_retriever.py,sha256=Da4Q5BrfN0GckJaeAgPYMlzQAp9ll7ZGGyvs7OdCQ5c,2513
+bisheng_langchain/rag/init_retrievers/keyword_retriever.py,sha256=NRT0fBx6HFR7j9IbRl_NBuqF7hnL-9v5GCqHpgnrfPQ,2523
 bisheng_langchain/rag/init_retrievers/mix_retriever.py,sha256=Whxq4kjNPLsxnHcVo60usdFFwLTCD-1jO38q08LXkVQ,4653
 bisheng_langchain/rag/init_retrievers/smaller_chunks_retriever.py,sha256=RQ7QLEOOhBrkw-EimXVJqIGa96D-KkNDik2h9hzg9fU,3805
 bisheng_langchain/rag/prompts/__init__.py,sha256=IUCq9gzqGQN_6IDk0D_F5t3mOUI_KbmSzYnnXoX4VKE,223
@@ -150,10 +152,10 @@ bisheng_langchain/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG
 bisheng_langchain/utils/azure_dalle_image_generator.py,sha256=96-_nO4hDSwyPE4rSYop5SgJ-U9CE2un4bTdW0E5RGU,6582
 bisheng_langchain/utils/requests.py,sha256=vWGKyNTxApVeaVdKxqACfIT1Q8wMy-jC3kUv2Ce9Mzc,8688
 bisheng_langchain/vectorstores/__init__.py,sha256=zCZgDe7LyQ0iDkfcm5UJ5NxwKQSRHnqrsjx700Fy11M,213
-bisheng_langchain/vectorstores/elastic_keywords_search.py,sha256=Pm1rS50GJ0HWbjBsFDgs28SVuVbjGSRPOor6yJlnE7w,13347
+bisheng_langchain/vectorstores/elastic_keywords_search.py,sha256=inZarhahRaesrvLqyeRCMQvHGAASY53opEVA0_o8S14,14901
 bisheng_langchain/vectorstores/milvus.py,sha256=xh7NokraKg_Xc9ofz0RVfJ_I36ftnprLJtV-1NfaeyQ,37162
 bisheng_langchain/vectorstores/retriever.py,sha256=hj4nAAl352EV_ANnU2OHJn7omCH3nBK82ydo14KqMH4,4353
-bisheng_langchain-0.3.5.dev1.dist-info/METADATA,sha256=Q20qBElwEheYunRPAoIvCRj8jH4RrXId03MA-SA6JnE,2476
-bisheng_langchain-0.3.5.dev1.dist-info/WHEEL,sha256=eOLhNAGa2EW3wWl_TU484h7q1UNgy0JXjjoqKoxAAQc,92
-bisheng_langchain-0.3.5.dev1.dist-info/top_level.txt,sha256=Z6pPNyCo4ihyr9iqGQbH8sJiC4dAUwA_mAyGRQB5_Fs,18
-bisheng_langchain-0.3.5.dev1.dist-info/RECORD,,
+bisheng_langchain-0.3.7.dist-info/METADATA,sha256=QmKT4P-W7klb8-YIRFq1Kqh8uHfq0454b9sOMgATjy4,2471
+bisheng_langchain-0.3.7.dist-info/WHEEL,sha256=bFJAMchF8aTQGUgMZzHJyDDMPTO3ToJ7x23SLJa1SVo,92
+bisheng_langchain-0.3.7.dist-info/top_level.txt,sha256=Z6pPNyCo4ihyr9iqGQbH8sJiC4dAUwA_mAyGRQB5_Fs,18
+bisheng_langchain-0.3.7.dist-info/RECORD,,

{bisheng_langchain-0.3.5.dev1.dist-info → bisheng_langchain-0.3.7.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: bdist_wheel (0.44.0)
+Generator: bdist_wheel (0.45.0)
 Root-Is-Purelib: true
 Tag: py3-none-any