langchain_core API Reference

langchain_core.agents

Classes

agents.AgentAction	A full description of an action for an ActionAgent to execute.
agents.AgentActionMessageLog	Override init to support instantiation by position for backward compat.
agents.AgentFinish	The final return value of an ActionAgent.
agents.AgentStep	The result of running an AgentAction.

Functions

langchain_core.beta

Classes

beta.runnables.context.Context()
beta.runnables.context.ContextGet	Create a new model by parsing and validating input data from keyword arguments.
beta.runnables.context.ContextSet	Create a new model by parsing and validating input data from keyword arguments.
beta.runnables.context.PrefixContext([prefix])

Functions

beta.runnables.context.aconfig_with_context(...)
beta.runnables.context.config_with_context(...)

langchain_core.caches

Classes

caches.BaseCache()

Base interface for cache.

langchain_core.callbacks

Classes

callbacks.base.AsyncCallbackHandler()	Async callback handler that handles callbacks from LangChain.
callbacks.base.BaseCallbackHandler()	Base callback handler that handles callbacks from LangChain.
callbacks.base.BaseCallbackManager(handlers)	Base callback manager that handles callbacks from LangChain.
callbacks.base.CallbackManagerMixin()	Mixin for callback manager.
callbacks.base.ChainManagerMixin()	Mixin for chain callbacks.
callbacks.base.LLMManagerMixin()	Mixin for LLM callbacks.
callbacks.base.RetrieverManagerMixin()	Mixin for Retriever callbacks.
callbacks.base.RunManagerMixin()	Mixin for run manager.
callbacks.base.ToolManagerMixin()	Mixin for tool callbacks.
callbacks.manager.AsyncCallbackManager(handlers)	Async callback manager that handles callbacks from LangChain.
callbacks.manager.AsyncCallbackManagerForChainGroup(...)	Async callback manager for the chain group.
callbacks.manager.AsyncCallbackManagerForChainRun(*, ...)	Async callback manager for chain run.
callbacks.manager.AsyncCallbackManagerForLLMRun(*, ...)	Async callback manager for LLM run.
callbacks.manager.AsyncCallbackManagerForRetrieverRun(*, ...)	Async callback manager for retriever run.
callbacks.manager.AsyncCallbackManagerForToolRun(*, ...)	Async callback manager for tool run.
callbacks.manager.AsyncParentRunManager(*, ...)	Async Parent Run Manager.
callbacks.manager.AsyncRunManager(*, run_id, ...)	Async Run Manager.
callbacks.manager.BaseRunManager(*, run_id, ...)	Base class for run manager (a bound callback manager).
callbacks.manager.CallbackManager(handlers)	Callback manager that handles callbacks from LangChain.
callbacks.manager.CallbackManagerForChainGroup(...)	Callback manager for the chain group.
callbacks.manager.CallbackManagerForChainRun(*, ...)	Callback manager for chain run.
callbacks.manager.CallbackManagerForLLMRun(*, ...)	Callback manager for LLM run.
callbacks.manager.CallbackManagerForRetrieverRun(*, ...)	Callback manager for retriever run.
callbacks.manager.CallbackManagerForToolRun(*, ...)	Callback manager for tool run.
callbacks.manager.ParentRunManager(*, ...[, ...])	Sync Parent Run Manager.
callbacks.manager.RunManager(*, run_id, ...)	Sync Run Manager.
callbacks.stdout.StdOutCallbackHandler([color])	Callback Handler that prints to std out.
callbacks.streaming_stdout.StreamingStdOutCallbackHandler()	Callback handler for streaming.

Functions

callbacks.manager.ahandle_event(handlers, ...)	Generic event handler for AsyncCallbackManager.
callbacks.manager.atrace_as_chain_group(...)	Get an async callback manager for a chain group in a context manager.
callbacks.manager.handle_event(handlers, ...)	Generic event handler for CallbackManager.
callbacks.manager.trace_as_chain_group(...)	Get a callback manager for a chain group in a context manager.

langchain_core.chat_history

Classes

chat_history.BaseChatMessageHistory()

Abstract base class for storing chat message history.

langchain_core.chat_sessions

Classes

chat_sessions.ChatSession

Chat Session represents a single conversation, channel, or other group of messages.

langchain_core.documents

Classes

documents.base.Document	Class for storing a piece of text and associated metadata.
documents.transformers.BaseDocumentTransformer()	Abstract base class for document transformation systems.

langchain_core.embeddings

Classes

embeddings.Embeddings()

Interface for embedding models.

langchain_core.example_selectors

Logic for selecting examples to include in prompts.

Classes

example_selectors.base.BaseExampleSelector()	Interface for selecting examples to include in prompts.
example_selectors.length_based.LengthBasedExampleSelector	Select examples based on length.
example_selectors.semantic_similarity.MaxMarginalRelevanceExampleSelector	ExampleSelector that selects examples based on Max Marginal Relevance.
example_selectors.semantic_similarity.SemanticSimilarityExampleSelector	Example selector that selects examples based on SemanticSimilarity.

Functions

example_selectors.semantic_similarity.sorted_values(values)

Return a list of values in dict sorted by key.

langchain_core.exceptions

Classes

exceptions.LangChainException	General LangChain exception.
exceptions.OutputParserException(error[, ...])	Exception that output parsers should raise to signify a parsing error.
exceptions.TracerException	Base class for exceptions in tracers module.

langchain_core.language_models

Classes

language_models.base.BaseLanguageModel	Abstract base class for interfacing with language models.
language_models.chat_models.BaseChatModel	Base class for Chat models.
language_models.chat_models.SimpleChatModel	Simple Chat Model.
language_models.llms.BaseLLM	Base LLM abstract interface.
language_models.llms.LLM	Base LLM abstract class.

Functions

language_models.chat_models.agenerate_from_stream(stream)
language_models.chat_models.generate_from_stream(stream)
language_models.llms.create_base_retry_decorator(...)	Create a retry decorator for a given LLM and provided list of error types.
language_models.llms.get_prompts(params, prompts)	Get prompts that are already cached.
language_models.llms.update_cache(...)	Update the cache and get the LLM output.

langchain_core.load

Serialization and deserialization.

Classes

load.load.Reviver([secrets_map, ...])	Reviver for JSON objects.
load.serializable.BaseSerialized	Base class for serialized objects.
load.serializable.Serializable	Serializable base class.
load.serializable.SerializedConstructor	Serialized constructor.
load.serializable.SerializedNotImplemented	Serialized not implemented.
load.serializable.SerializedSecret	Serialized secret.

Functions

load.dump.default(obj)	Return a default value for a Serializable object or a SerializedNotImplemented object.
load.dump.dumpd(obj)	Return a json dict representation of an object.
load.dump.dumps(obj, *[, pretty])	Return a json string representation of an object.
load.load.load(obj, *[, secrets_map, ...])	Revive a LangChain class from a JSON object.
load.load.loads(text, *[, secrets_map, ...])	Revive a LangChain class from a JSON string.
load.serializable.to_json_not_implemented(obj)	Serialize a "not implemented" object.
load.serializable.try_neq_default(value, ...)

langchain_core.memory

Classes

memory.BaseMemory

Abstract base class for memory in Chains.

langchain_core.messages

Classes

messages.ai.AIMessage	A Message from an AI.
messages.ai.AIMessageChunk	A Message chunk from an AI.
messages.base.BaseMessage	The base abstract Message class.
messages.base.BaseMessageChunk	A Message chunk, which can be concatenated with other Message chunks.
messages.chat.ChatMessage	A Message that can be assigned an arbitrary speaker (i.e.
messages.chat.ChatMessageChunk	A Chat Message chunk.
messages.function.FunctionMessage	A Message for passing the result of executing a function back to a model.
messages.function.FunctionMessageChunk	A Function Message chunk.
messages.human.HumanMessage	A Message from a human.
messages.human.HumanMessageChunk	A Human Message chunk.
messages.system.SystemMessage	A Message for priming AI behavior, usually passed in as the first of a sequence of input messages.
messages.system.SystemMessageChunk	A System Message chunk.
messages.tool.ToolMessage	A Message for passing the result of executing a tool back to a model.
messages.tool.ToolMessageChunk	A Tool Message chunk.

Functions

messages.base.merge_content(first_content, ...)
messages.base.message_to_dict(message)
messages.base.messages_to_dict(messages)	Convert a sequence of Messages to a list of dictionaries.

langchain_core.output_parsers

Classes

output_parsers.base.BaseGenerationOutputParser	Base class to parse the output of an LLM call.
output_parsers.base.BaseLLMOutputParser()	Abstract base class for parsing the outputs of a model.
output_parsers.base.BaseOutputParser	Base class to parse the output of an LLM call.
output_parsers.list.CommaSeparatedListOutputParser	Parse the output of an LLM call to a comma-separated list.
output_parsers.list.ListOutputParser	Parse the output of an LLM call to a list.
output_parsers.list.MarkdownListOutputParser	Parse a markdown list.
output_parsers.list.NumberedListOutputParser	Parse a numbered list.
output_parsers.string.StrOutputParser	OutputParser that parses LLMResult into the top likely string.
output_parsers.transform.BaseCumulativeTransformOutputParser	Base class for an output parser that can handle streaming input.
output_parsers.transform.BaseTransformOutputParser	Base class for an output parser that can handle streaming input.

langchain_core.outputs

Classes

outputs.chat_generation.ChatGeneration	A single chat generation output.
outputs.chat_generation.ChatGenerationChunk	A ChatGeneration chunk, which can be concatenated with other
outputs.chat_result.ChatResult	Class that contains all results for a single chat model call.
outputs.generation.Generation	A single text generation output.
outputs.generation.GenerationChunk	A Generation chunk, which can be concatenated with other Generation chunks.
outputs.llm_result.LLMResult	Class that contains all results for a batched LLM call.
outputs.run_info.RunInfo	Class that contains metadata for a single execution of a Chain or model.

langchain_core.prompt_values

Classes

prompt_values.ChatPromptValue	Chat prompt value.
prompt_values.ChatPromptValueConcrete	Chat prompt value which explicitly lists out the message types it accepts.
prompt_values.PromptValue	Base abstract class for inputs to any language model.
prompt_values.StringPromptValue	String prompt value.

langchain_core.prompts

Prompt is the input to the model.

Prompt is often constructed from multiple components. Prompt classes and functions make constructing

and working with prompts easy.

Class hierarchy:

BasePromptTemplate --> PipelinePromptTemplate
                       StringPromptTemplate --> PromptTemplate
                                                FewShotPromptTemplate
                                                FewShotPromptWithTemplates
                       BaseChatPromptTemplate --> AutoGPTPrompt
                                                  ChatPromptTemplate --> AgentScratchPadChatPromptTemplate



BaseMessagePromptTemplate --> MessagesPlaceholder
                              BaseStringMessagePromptTemplate --> ChatMessagePromptTemplate
                                                                  HumanMessagePromptTemplate
                                                                  AIMessagePromptTemplate
                                                                  SystemMessagePromptTemplate

Classes

prompts.base.BasePromptTemplate	Base class for all prompt templates, returning a prompt.
prompts.chat.AIMessagePromptTemplate	AI message prompt template.
prompts.chat.BaseChatPromptTemplate	Base class for chat prompt templates.
prompts.chat.BaseMessagePromptTemplate	Base class for message prompt templates.
prompts.chat.BaseStringMessagePromptTemplate	Base class for message prompt templates that use a string prompt template.
prompts.chat.ChatMessagePromptTemplate	Chat message prompt template.
prompts.chat.ChatPromptTemplate	A prompt template for chat models.
prompts.chat.HumanMessagePromptTemplate	Human message prompt template.
prompts.chat.MessagesPlaceholder	Prompt template that assumes variable is already list of messages.
prompts.chat.SystemMessagePromptTemplate	System message prompt template.
prompts.few_shot.FewShotChatMessagePromptTemplate	Chat prompt template that supports few-shot examples.
prompts.few_shot.FewShotPromptTemplate	Prompt template that contains few shot examples.
prompts.few_shot_with_templates.FewShotPromptWithTemplates	Prompt template that contains few shot examples.
prompts.pipeline.PipelinePromptTemplate	A prompt template for composing multiple prompt templates together.
prompts.prompt.PromptTemplate	A prompt template for a language model.
prompts.string.StringPromptTemplate	String prompt that exposes the format method, returning a prompt.

Functions

prompts.base.format_document(doc, prompt)	Format a document into a string based on a prompt template.
prompts.loading.load_prompt(path)	Unified method for loading a prompt from LangChainHub or local fs.
prompts.loading.load_prompt_from_config(config)	Load prompt from Config Dict.
prompts.string.check_valid_template(...)	Check that template string is valid.
prompts.string.get_template_variables(...)	Get the variables from the template.
prompts.string.jinja2_formatter(template, ...)	Format a template using jinja2.
prompts.string.validate_jinja2(template, ...)	Validate that the input variables are valid for the template.

langchain_core.retrievers

Classes

retrievers.BaseRetriever

Abstract base class for a Document retrieval system.

langchain_core.runnables

LangChain Runnable and the LangChain Expression Language (LCEL).

The LangChain Expression Language (LCEL) offers a declarative method to build production-grade programs that harness the power of LLMs.

Programs created using LCEL and LangChain Runnables inherently support synchronous, asynchronous, batch, and streaming operations.

Support for async allows servers hosting LCEL based programs to scale better for higher concurrent loads.

Streaming of intermediate outputs as they’re being generated allows for creating more responsive UX.

This module contains schema and implementation of LangChain Runnables primitives.

Classes

runnables.base.Runnable()	A unit of work that can be invoked, batched, streamed, transformed and composed.
runnables.base.RunnableBinding	Wrap a runnable with additional functionality.
runnables.base.RunnableBindingBase	A runnable that delegates calls to another runnable with a set of kwargs.
runnables.base.RunnableEach	A runnable that delegates calls to another runnable with each element of the input sequence.
runnables.base.RunnableEachBase	A runnable that delegates calls to another runnable with each element of the input sequence.
runnables.base.RunnableGenerator(transform)	A runnable that runs a generator function.
runnables.base.RunnableLambda(func[, afunc])	RunnableLambda converts a python callable into a Runnable.
runnables.base.RunnableMap	alias of RunnableParallel
runnables.base.RunnableParallel	A runnable that runs a mapping of runnables in parallel, and returns a mapping of their outputs.
runnables.base.RunnableSequence	A sequence of runnables, where the output of each is the input of the next.
runnables.base.RunnableSerializable	A Runnable that can be serialized to JSON.
runnables.branch.RunnableBranch	A Runnable that selects which branch to run based on a condition.
runnables.config.EmptyDict	Empty dict type.
runnables.config.RunnableConfig	Configuration for a Runnable.
runnables.configurable.DynamicRunnable	A Serializable Runnable that can be dynamically configured.
runnables.configurable.RunnableConfigurableAlternatives	A Runnable that can be dynamically configured.
runnables.configurable.RunnableConfigurableFields	A Runnable that can be dynamically configured.
runnables.configurable.StrEnum(value[, ...])	A string enum.
runnables.fallbacks.RunnableWithFallbacks	A Runnable that can fallback to other Runnables if it fails.
runnables.history.RunnableWithMessageHistory	A runnable that manages chat message history for another runnable.
runnables.passthrough.RunnableAssign	A runnable that assigns key-value pairs to Dict[str, Any] inputs.
runnables.passthrough.RunnablePassthrough	A runnable to passthrough inputs unchanged or with additional keys.
runnables.retry.RunnableRetry	Retry a Runnable if it fails.
runnables.router.RouterInput	A Router input.
runnables.router.RouterRunnable	A runnable that routes to a set of runnables based on Input['key'].
runnables.utils.AddableDict	Dictionary that can be added to another dictionary.
runnables.utils.ConfigurableField(id[, ...])	A field that can be configured by the user.
runnables.utils.ConfigurableFieldMultiOption(id, ...)	A field that can be configured by the user with multiple default values.
runnables.utils.ConfigurableFieldSingleOption(id, ...)	A field that can be configured by the user with a default value.
runnables.utils.ConfigurableFieldSpec(id, ...)	A field that can be configured by the user.
runnables.utils.GetLambdaSource()	Get the source code of a lambda function.
runnables.utils.IsFunctionArgDict()	Check if the first argument of a function is a dict.
runnables.utils.IsLocalDict(name, keys)	Check if a name is a local dict.
runnables.utils.SupportsAdd(*args, **kwargs)	Protocol for objects that support addition.

Functions

runnables.base.coerce_to_runnable(thing)	Coerce a runnable-like object into a Runnable.
runnables.config.acall_func_with_variable_args(...)	Call function that may optionally accept a run_manager and/or config.
runnables.config.call_func_with_variable_args(...)	Call function that may optionally accept a run_manager and/or config.
runnables.config.ensure_config([config])	Ensure that a config is a dict with all keys present.
runnables.config.get_async_callback_manager_for_config(config)	Get an async callback manager for a config.
runnables.config.get_callback_manager_for_config(config)	Get a callback manager for a config.
runnables.config.get_config_list(config, length)	Get a list of configs from a single config or a list of configs.
runnables.config.get_executor_for_config(config)	Get an executor for a config.
runnables.config.merge_configs(*configs)	Merge multiple configs into one.
runnables.config.patch_config(config, *[, ...])	Patch a config with new values.
runnables.configurable.make_options_spec(...)	Make a ConfigurableFieldSpec for a ConfigurableFieldSingleOption or ConfigurableFieldMultiOption.
runnables.configurable.prefix_config_spec(...)
runnables.passthrough.aidentity(x)	An async identity function
runnables.passthrough.identity(x)	An identity function
runnables.utils.aadd(addables)	Asynchronously add a sequence of addable objects together.
runnables.utils.accepts_config(callable)	Check if a callable accepts a config argument.
runnables.utils.accepts_run_manager(callable)	Check if a callable accepts a run_manager argument.
runnables.utils.add(addables)	Add a sequence of addable objects together.
runnables.utils.gated_coro(semaphore, coro)	Run a coroutine with a semaphore.
runnables.utils.gather_with_concurrency(n, ...)	Gather coroutines with a limit on the number of concurrent coroutines.
runnables.utils.get_function_first_arg_dict_keys(func)	Get the keys of the first argument of a function if it is a dict.
runnables.utils.get_lambda_source(func)	Get the source code of a lambda function.
runnables.utils.get_unique_config_specs(specs)	Get the unique config specs from a sequence of config specs.
runnables.utils.indent_lines_after_first(...)	Indent all lines of text after the first line.

langchain_core.stores

Classes

stores.BaseStore()

Abstract interface for a key-value store.

langchain_core.tools

Base implementation for tools or skills.

Classes

tools.BaseTool	Interface LangChain tools must implement.
tools.SchemaAnnotationError	Raised when 'args_schema' is missing or has an incorrect type annotation.
tools.StructuredTool	Tool that can operate on any number of inputs.
tools.Tool	Tool that takes in function or coroutine directly.
tools.ToolException	An optional exception that tool throws when execution error occurs.

Functions

tools.create_schema_from_function(...)	Create a pydantic schema from a function's signature.
tools.tool(*args[, return_direct, ...])	Make tools out of functions, can be used with or without arguments.

langchain_core.tracers

Classes

tracers.base.BaseTracer(**kwargs)	Base interface for tracers.
tracers.evaluation.EvaluatorCallbackHandler(...)	A tracer that runs a run evaluator whenever a run is persisted.
tracers.langchain.LangChainTracer([...])	An implementation of the SharedTracer that POSTS to the langchain endpoint.
tracers.langchain_v1.LangChainTracerV1(**kwargs)	An implementation of the SharedTracer that POSTS to the langchain endpoint.
tracers.log_stream.LogEntry	A single entry in the run log.
tracers.log_stream.LogStreamCallbackHandler(*)	A tracer that streams run logs to a stream.
tracers.log_stream.RunLog(*ops, state)	A run log.
tracers.log_stream.RunLogPatch(*ops)	A patch to the run log.
tracers.log_stream.RunState	State of the run.
tracers.root_listeners.RootListenersTracer(*, ...)
tracers.run_collector.RunCollectorCallbackHandler([...])	A tracer that collects all nested runs in a list.
tracers.schemas.BaseRun	Base class for Run.
tracers.schemas.ChainRun	Class for ChainRun.
tracers.schemas.LLMRun	Class for LLMRun.
tracers.schemas.Run	Run schema for the V2 API in the Tracer.
tracers.schemas.ToolRun	Class for ToolRun.
tracers.schemas.TracerSession	TracerSessionV1 schema for the V2 API.
tracers.schemas.TracerSessionBase	Base class for TracerSession.
tracers.schemas.TracerSessionV1	TracerSessionV1 schema.
tracers.schemas.TracerSessionV1Base	Base class for TracerSessionV1.
tracers.schemas.TracerSessionV1Create	Create class for TracerSessionV1.
tracers.stdout.ConsoleCallbackHandler(**kwargs)	Tracer that prints to the console.
tracers.stdout.FunctionCallbackHandler(...)	Tracer that calls a function with a single str parameter.

Functions

tracers.context.collect_runs()	Collect all run traces in context.
tracers.context.register_configure_hook(...)
tracers.context.tracing_enabled([session_name])	Get the Deprecated LangChainTracer in a context manager.
tracers.context.tracing_v2_enabled([...])	Instruct LangChain to log all runs in context to LangSmith.
tracers.evaluation.wait_for_all_evaluators()	Wait for all tracers to finish.
tracers.langchain.get_client()	Get the client.
tracers.langchain.log_error_once(method, ...)	Log an error once.
tracers.langchain.wait_for_all_tracers()	Wait for all tracers to finish.
tracers.langchain_v1.get_headers()	Get the headers for the LangChain API.
tracers.schemas.RunTypeEnum()	RunTypeEnum.
tracers.stdout.elapsed(run)	Get the elapsed time of a run.
tracers.stdout.try_json_stringify(obj, fallback)	Try to stringify an object to JSON.

langchain_core.utils

Utility functions for LangChain.

These functions do not depend on any other LangChain module.

Classes

utils.aiter.NoLock()	Dummy lock that provides the proper interface but no protection
utils.aiter.Tee(iterable[, n, lock])	Create n separate asynchronous iterators over iterable
utils.aiter.atee	alias of Tee
utils.formatting.StrictFormatter()	A subclass of formatter that checks for extra keys.
utils.iter.NoLock()	Dummy lock that provides the proper interface but no protection
utils.iter.Tee(iterable[, n, lock])	Create n separate asynchronous iterators over iterable
utils.iter.safetee	alias of Tee

Functions

utils.aiter.py_anext(iterator[, default])	Pure-Python implementation of anext() for testing purposes.
utils.aiter.tee_peer(iterator, buffer, ...)	An individual iterator of a tee()
utils.env.env_var_is_set(env_var)	Check if an environment variable is set.
utils.env.get_from_dict_or_env(data, key, ...)	Get a value from a dictionary or an environment variable.
utils.env.get_from_env(key, env_key[, default])	Get a value from a dictionary or an environment variable.
utils.html.extract_sub_links(raw_html, url, *)	Extract all links from a raw html string and convert into absolute paths.
utils.html.find_all_links(raw_html, *[, pattern])	Extract all links from a raw html string.
utils.input.get_bolded_text(text)	Get bolded text.
utils.input.get_color_mapping(items[, ...])	Get mapping for items to a support color.
utils.input.get_colored_text(text, color)	Get colored text.
utils.input.print_text(text[, color, end, file])	Print text with highlighting and no end characters.
utils.iter.batch_iterate(size, iterable)	Utility batching function.
utils.iter.tee_peer(iterator, buffer, peers, ...)	An individual iterator of a tee()
utils.json_schema.dereference_refs(schema_obj, *)	Try to substitute $refs in JSON Schema.
utils.loading.try_load_from_hub(path, ...)	Load configuration from hub.
utils.pydantic.get_pydantic_major_version()	Get the major version of Pydantic.
utils.strings.comma_list(items)	Convert a list to a comma-separated string.
utils.strings.stringify_dict(data)	Stringify a dictionary.
utils.strings.stringify_value(val)	Stringify a value.
utils.utils.build_extra_kwargs(extra_kwargs, ...)	Build extra kwargs from values and extra_kwargs.
utils.utils.check_package_version(package[, ...])	Check the version of a package.
utils.utils.convert_to_secret_str(value)	Convert a string to a SecretStr if needed.
utils.utils.get_pydantic_field_names(...)	Get field names, including aliases, for a pydantic class.
utils.utils.guard_import(module_name, *[, ...])	Dynamically imports a module and raises a helpful exception if the module is not installed.
utils.utils.mock_now(dt_value)	Context manager for mocking out datetime.now() in unit tests.
utils.utils.raise_for_status_with_text(response)	Raise an error with the response text.
utils.utils.xor_args(*arg_groups)	Validate specified keyword args are mutually exclusive.

langchain_core.vectorstores

Classes

vectorstores.VectorStore()	Interface for vector store.
vectorstores.VectorStoreRetriever	Base Retriever class for VectorStore.