langchain.callbacks.argilla_callback.ArgillaCallbackHandler

class langchain.callbacks.argilla_callback.ArgillaCallbackHandler(dataset_name: str, workspace_name: Optional[str] = None, api_url: Optional[str] = None, api_key: Optional[str] = None)

Bases: BaseCallbackHandler

Callback Handler that logs into Argilla.

Parameters

• dataset_name – name of the FeedbackDataset in Argilla. Note that it must exist in advance. If you need help on how to create a FeedbackDataset in Argilla, please visit https://docs.argilla.io/en/latest/guides/llms/practical_guides/use_argilla_callback_in_langchain.html.

• workspace_name – name of the workspace in Argilla where the specified FeedbackDataset lives in. Defaults to None, which means that the default workspace will be used.

• api_url – URL of the Argilla Server that we want to use, and where the FeedbackDataset lives in. Defaults to None, which means that either ARGILLA_API_URL environment variable or the default http://localhost:6900 will be used.

• api_key – API Key to connect to the Argilla Server. Defaults to None, which means that either ARGILLA_API_KEY environment variable or the default argilla.apikey will be used.

Raises

• ImportError – if the argilla package is not installed.

• ConnectionError – if the connection to Argilla fails.

• FileNotFoundError – if the FeedbackDataset retrieval from Argilla fails.

Examples

>>> from langchain.llms import OpenAI
>>> from langchain.callbacks import ArgillaCallbackHandler
>>> argilla_callback = ArgillaCallbackHandler(
...     dataset_name="my-dataset",
...     workspace_name="my-workspace",
...     api_url="http://localhost:6900",
...     api_key="argilla.apikey",
... )
>>> llm = OpenAI(
...     temperature=0,
...     callbacks=[argilla_callback],
...     verbose=True,
...     openai_api_key="API_KEY_HERE",
... )
>>> llm.generate([
...     "What is the best NLP-annotation tool out there? (no bias at all)",
... ])
"Argilla, no doubt about it."

Initializes the ArgillaCallbackHandler.

Parameters

• dataset_name – name of the FeedbackDataset in Argilla. Note that it must exist in advance. If you need help on how to create a FeedbackDataset in Argilla, please visit https://docs.argilla.io/en/latest/guides/llms/practical_guides/use_argilla_callback_in_langchain.html.

• workspace_name – name of the workspace in Argilla where the specified FeedbackDataset lives in. Defaults to None, which means that the default workspace will be used.

• api_url – URL of the Argilla Server that we want to use, and where the FeedbackDataset lives in. Defaults to None, which means that either ARGILLA_API_URL environment variable or the default http://localhost:6900 will be used.

• api_key – API Key to connect to the Argilla Server. Defaults to None, which means that either ARGILLA_API_KEY environment variable or the default argilla.apikey will be used.

Raises

• ImportError – if the argilla package is not installed.

• ConnectionError – if the connection to Argilla fails.

• FileNotFoundError – if the FeedbackDataset retrieval from Argilla fails.

Methods

__init__(dataset_name[, workspace_name, ...])	Initializes the ArgillaCallbackHandler.
on_agent_action(action, **kwargs)	Do nothing when agent takes a specific action.
on_agent_finish(finish, **kwargs)	Do nothing
on_chain_end(outputs, **kwargs)	If either the parent_run_id or the run_id is in self.prompts, then log the outputs to Argilla, and pop the run from self.prompts.
on_chain_error(error, **kwargs)	Do nothing when LLM chain outputs an error.
on_chain_start(serialized, inputs, **kwargs)	If the key input is in inputs, then save it in self.prompts using either the parent_run_id or the run_id as the key.
on_chat_model_start(serialized, messages, *, ...)	Run when a chat model starts running.
on_llm_end(response, **kwargs)	Log records to Argilla when an LLM ends.
on_llm_error(error, **kwargs)	Do nothing when LLM outputs an error.
on_llm_new_token(token, **kwargs)	Do nothing when a new token is generated.
on_llm_start(serialized, prompts, **kwargs)	Save the prompts in memory when an LLM starts.
on_retriever_end(documents, *, run_id[, ...])	Run when Retriever ends running.
on_retriever_error(error, *, run_id[, ...])	Run when Retriever errors.
on_retriever_start(serialized, query, *, run_id)	Run when Retriever starts running.
on_text(text, **kwargs)	Do nothing
on_tool_end(output[, observation_prefix, ...])	Do nothing when tool ends.
on_tool_error(error, **kwargs)	Do nothing when tool outputs an error.
on_tool_start(serialized, input_str, **kwargs)	Do nothing when tool starts.

Attributes

ignore_agent	Whether to ignore agent callbacks.
ignore_chain	Whether to ignore chain callbacks.
ignore_chat_model	Whether to ignore chat model callbacks.
ignore_llm	Whether to ignore LLM callbacks.
ignore_retriever	Whether to ignore retriever callbacks.
ignore_retry	Whether to ignore retry callbacks.
raise_error
run_inline

on_agent_action(action: AgentAction, **kwargs: Any) → Any

Do nothing when agent takes a specific action.

on_agent_finish(finish: AgentFinish, **kwargs: Any) → None

Do nothing

on_chain_end(outputs: Dict[str, Any], **kwargs: Any) → None

If either the parent_run_id or the run_id is in self.prompts, then log the outputs to Argilla, and pop the run from self.prompts. The behavior differs if the output is a list or not.

on_chain_error(error: Union[Exception, KeyboardInterrupt], **kwargs: Any) → None

Do nothing when LLM chain outputs an error.

on_chain_start(serialized: Dict[str, Any], inputs: Dict[str, Any], **kwargs: Any) → None

If the key input is in inputs, then save it in self.prompts using either the parent_run_id or the run_id as the key. This is done so that we don’t log the same input prompt twice, once when the LLM starts and once when the chain starts.

on_chat_model_start(serialized: Dict[str, Any], messages: List[List[BaseMessage]], *, run_id: UUID, parent_run_id: Optional[UUID] = None, tags: Optional[List[str]] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs: Any) → Any

Run when a chat model starts running.

on_llm_end(response: LLMResult, **kwargs: Any) → None

Log records to Argilla when an LLM ends.

on_llm_error(error: Union[Exception, KeyboardInterrupt], **kwargs: Any) → None

Do nothing when LLM outputs an error.

on_llm_new_token(token: str, **kwargs: Any) → None

Do nothing when a new token is generated.

on_llm_start(serialized: Dict[str, Any], prompts: List[str], **kwargs: Any) → None

Save the prompts in memory when an LLM starts.

on_retriever_end(documents: Sequence[Document], *, run_id: UUID, parent_run_id: Optional[UUID] = None, **kwargs: Any) → Any

Run when Retriever ends running.

on_retriever_error(error: Union[Exception, KeyboardInterrupt], *, run_id: UUID, parent_run_id: Optional[UUID] = None, **kwargs: Any) → Any

Run when Retriever errors.

on_retriever_start(serialized: Dict[str, Any], query: str, *, run_id: UUID, parent_run_id: Optional[UUID] = None, tags: Optional[List[str]] = None, metadata: Optional[Dict[str, Any]] = None, **kwargs: Any) → Any

Run when Retriever starts running.

on_text(text: str, **kwargs: Any) → None

Do nothing

on_tool_end(output: str, observation_prefix: Optional[str] = None, llm_prefix: Optional[str] = None, **kwargs: Any) → None

Do nothing when tool ends.

on_tool_error(error: Union[Exception, KeyboardInterrupt], **kwargs: Any) → None

Do nothing when tool outputs an error.

on_tool_start(serialized: Dict[str, Any], input_str: str, **kwargs: Any) → None

Do nothing when tool starts.

property ignore_agent: bool

Whether to ignore agent callbacks.

property ignore_chain: bool

Whether to ignore chain callbacks.

property ignore_chat_model: bool

Whether to ignore chat model callbacks.

property ignore_llm: bool

Whether to ignore LLM callbacks.

property ignore_retriever: bool

Whether to ignore retriever callbacks.

property ignore_retry: bool

Whether to ignore retry callbacks.

raise_error: bool = False

run_inline: bool = False

Examples using ArgillaCallbackHandler

• Argilla