Extracting information from documents.
Module haystack_experimental.components.extractors.llm_metadata_extractor
LLMProvider
Currently LLM providers supported by LLMMetadataExtractor.
LLMProvider.from_str
@staticmethod
def from_str(string: str) -> "LLMProvider"
Convert a string to a LLMProvider enum.
LLMMetadataExtractor
Extracts metadata from documents using a Large Language Model (LLM) from OpenAI.
The metadata is extracted by providing a prompt to an LLM that generates the metadata.
This component expects as input a list of documents and a prompt. The prompt should have a variable called
document that will point to a single document in the list of documents. So to access the content of the document,
you can use {{ document.content }} in the prompt.
The component will run the LLM on each document in the list and extract metadata from the document. The metadata
will be added to the document's metadata field. If the LLM fails to extract metadata from a document, the document
will be added to the failed_documents list. The failed documents will have the keys metadata_extraction_error and
metadata_extraction_response in their metadata. These documents can be re-run with another extractor to
extract metadata by using the metadata_extraction_response and metadata_extraction_error in the prompt.
from haystack import Document
from haystack_experimental.components.extractors.llm_metadata_extractor import LLMMetadataExtractor
NER_PROMPT = '''
-Goal-
Given text and a list of entity types, identify all entities of those types from the text.
-Steps-
1. Identify all entities. For each identified entity, extract the following information:
- entity_name: Name of the entity, capitalized
- entity_type: One of the following types: [organization, product, service, industry]
Format each entity as a JSON like: {"entity": <entity_name>, "entity_type": <entity_type>}
2. Return output in a single list with all the entities identified in steps 1.
-Examples-
######################
Example 1:
entity_types: [organization, person, partnership, financial metric, product, service, industry, investment strategy, market trend]
text: Another area of strength is our co-brand issuance. Visa is the primary network partner for eight of the top
10 co-brand partnerships in the US today and we are pleased that Visa has finalized a multi-year extension of
our successful credit co-branded partnership with Alaska Airlines, a portfolio that benefits from a loyal customer
base and high cross-border usage.
We have also had significant co-brand momentum in CEMEA. First, we launched a new co-brand card in partnership
with Qatar Airways, British Airways and the National Bank of Kuwait. Second, we expanded our strong global
Marriott relationship to launch Qatar's first hospitality co-branded card with Qatar Islamic Bank. Across the
United Arab Emirates, we now have exclusive agreements with all the leading airlines marked by a recent
agreement with Emirates Skywards.
And we also signed an inaugural Airline co-brand agreement in Morocco with Royal Air Maroc. Now newer digital
issuers are equally
------------------------
output:
{"entities": [{"entity": "Visa", "entity_type": "company"}, {"entity": "Alaska Airlines", "entity_type": "company"}, {"entity": "Qatar Airways", "entity_type": "company"}, {"entity": "British Airways", "entity_type": "company"}, {"entity": "National Bank of Kuwait", "entity_type": "company"}, {"entity": "Marriott", "entity_type": "company"}, {"entity": "Qatar Islamic Bank", "entity_type": "company"}, {"entity": "Emirates Skywards", "entity_type": "company"}, {"entity": "Royal Air Maroc", "entity_type": "company"}]}
#############################
-Real Data-
######################
entity_types: [company, organization, person, country, product, service]
text: {{ document.content }}
######################
output:
'''
docs = [
Document(content="deepset was founded in 2018 in Berlin, and is known for its Haystack framework"),
Document(content="Hugging Face is a company founded in Paris, France and is known for its Transformers library")
]
extractor = LLMMetadataExtractor(
prompt=NER_PROMPT,
generator_api="openai",
generator_api_params={
"generation_kwargs": {
"max_tokens": 500,
"temperature": 0.0,
"seed": 0,
"response_format": {"type": "json_object"},
},
"max_retries": 1,
"timeout": 60.0,
},
expected_keys=["entities"],
raise_on_failure=False,
)
extractor.warm_up()
extractor.run(documents=docs)
>> {'documents': [
Document(id=.., content: 'deepset was founded in 2018 in Berlin, and is known for its Haystack framework',
meta: {'entities': [{'entity': 'deepset', 'entity_type': 'company'}, {'entity': 'Berlin', 'entity_type': 'city'},
{'entity': 'Haystack', 'entity_type': 'product'}]}),
Document(id=.., content: 'Hugging Face is a company founded in Paris, France and is known for its Transformers library',
meta: {'entities': [
{'entity': 'Hugging Face', 'entity_type': 'company'}, {'entity': 'Paris', 'entity_type': 'city'},
{'entity': 'France', 'entity_type': 'country'}, {'entity': 'Transformers', 'entity_type': 'product'}
]})
]
'failed_documents': []
}
>>
LLMMetadataExtractor.__init__
def __init__(prompt: str,
generator_api: Union[str, LLMProvider],
generator_api_params: Optional[Dict[str, Any]] = None,
expected_keys: Optional[List[str]] = None,
page_range: Optional[List[Union[str, int]]] = None,
raise_on_failure: bool = False,
max_workers: int = 3)
Initializes the LLMMetadataExtractor.
Arguments:
prompt: The prompt to be used for the LLM.generator_api: The API provider for the LLM. Currently supported providers are: "openai", "openai_azure", "aws_bedrock", "google_vertex"generator_api_params: The parameters for the LLM generator.expected_keys: The keys expected in the JSON output from the LLM.page_range: A range of pages to extract metadata from. For example, page_range=['1', '3'] will extract metadata from the first and third pages of each document. It also accepts printable range strings, e.g.: ['1-3', '5', '8', '10-12'] will extract metadata from pages 1, 2, 3, 5, 8, 10, 11, 12. If None, metadata will be extracted from the entire document for each document in the documents list. This parameter is optional and can be overridden in therunmethod.raise_on_failure: Whether to raise an error on failure during the execution of the Generator or validation of the JSON output.max_workers: The maximum number of workers to use in the thread pool executor.
LLMMetadataExtractor.warm_up
def warm_up()
Warm up the LLM provider component.
LLMMetadataExtractor.to_dict
def to_dict() -> Dict[str, Any]
Serializes the component to a dictionary.
Returns:
Dictionary with serialized data.
LLMMetadataExtractor.from_dict
@classmethod
def from_dict(cls, data: Dict[str, Any]) -> "LLMMetadataExtractor"
Deserializes the component from a dictionary.
Arguments:
data: Dictionary with serialized data.
Returns:
An instance of the component.
LLMMetadataExtractor.run
@component.output_types(documents=List[Document],
failed_documents=List[Document])
def run(documents: List[Document],
page_range: Optional[List[Union[str, int]]] = None)
Extract metadata from documents using a Large Language Model.
If page_range is provided, the metadata will be extracted from the specified range of pages. This component
will split the documents into pages and extract metadata from the specified range of pages. The metadata will be
extracted from the entire document if page_range is not provided.
The original documents will be returned updated with the extracted metadata.
Arguments:
documents: List of documents to extract metadata from.page_range: A range of pages to extract metadata from. For example, page_range=['1', '3'] will extract metadata from the first and third pages of each document. It also accepts printable range strings, e.g.: ['1-3', '5', '8', '10-12'] will extract metadata from pages 1, 2, 3, 5, 8, 10, 11, 12. If None, metadata will be extracted from the entire document for each document in the documents list.
Returns:
A dictionary with the keys:
- "documents": A list of documents that were successfully updated with the extracted metadata.
- "failed_documents": A list of documents that failed to extract metadata. These documents will have "metadata_extraction_error" and "metadata_extraction_response" in their metadata. These documents can be re-run with the extractor to extract metadata.