You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
langchain/libs/community/langchain_community/utilities/arcee.py

256 lines
8.5 KiB
Python

# This module contains utility classes and functions for interacting with Arcee API.
# For more information and updates, refer to the Arcee utils page:
# [https://github.com/arcee-ai/arcee-python/blob/main/arcee/dalm.py]
from enum import Enum
from typing import Any, Dict, List, Literal, Mapping, Optional, Union
import requests
from langchain_core.pydantic_v1 import BaseModel, SecretStr, root_validator
from langchain_core.retrievers import Document
class ArceeRoute(str, Enum):
"""Routes available for the Arcee API as enumerator."""
generate = "models/generate"
retrieve = "models/retrieve"
model_training_status = "models/status/{id_or_name}"
class DALMFilterType(str, Enum):
"""Filter types available for a DALM retrieval as enumerator."""
fuzzy_search = "fuzzy_search"
strict_search = "strict_search"
class DALMFilter(BaseModel):
"""Filters available for a DALM retrieval and generation.
Arguments:
field_name: The field to filter on. Can be 'document' or 'name' to filter
on your document's raw text or title. Any other field will be presumed
to be a metadata field you included when uploading your context data
filter_type: Currently 'fuzzy_search' and 'strict_search' are supported.
'fuzzy_search' means a fuzzy search on the provided field is performed.
The exact strict doesn't need to exist in the document
for this to find a match.
Very useful for scanning a document for some keyword terms.
'strict_search' means that the exact string must appear
in the provided field.
This is NOT an exact eq filter. ie a document with content
"the happy dog crossed the street" will match on a strict_search of
"dog" but won't match on "the dog".
Python equivalent of `return search_string in full_string`.
value: The actual value to search for in the context data/metadata
"""
field_name: str
filter_type: DALMFilterType
value: str
_is_metadata: bool = False
@root_validator()
def set_meta(cls, values: Dict) -> Dict:
"""document and name are reserved arcee keys. Anything else is metadata"""
values["_is_meta"] = values.get("field_name") not in ["document", "name"]
return values
class ArceeDocumentSource(BaseModel):
"""Source of an Arcee document."""
document: str
name: str
id: str
class ArceeDocument(BaseModel):
"""Arcee document."""
index: str
id: str
score: float
source: ArceeDocumentSource
class ArceeDocumentAdapter:
"""Adapter for Arcee documents"""
@classmethod
def adapt(cls, arcee_document: ArceeDocument) -> Document:
"""Adapts an `ArceeDocument` to a langchain's `Document` object."""
return Document(
page_content=arcee_document.source.document,
metadata={
# arcee document; source metadata
"name": arcee_document.source.name,
"source_id": arcee_document.source.id,
# arcee document metadata
"index": arcee_document.index,
"id": arcee_document.id,
"score": arcee_document.score,
},
)
class ArceeWrapper:
"""Wrapper for Arcee API.
For more details, see: https://www.arcee.ai/
"""
def __init__(
self,
arcee_api_key: Union[str, SecretStr],
arcee_api_url: str,
arcee_api_version: str,
model_kwargs: Optional[Dict[str, Any]],
model_name: str,
):
"""Initialize ArceeWrapper.
Arguments:
arcee_api_key: API key for Arcee API.
arcee_api_url: URL for Arcee API.
arcee_api_version: Version of Arcee API.
model_kwargs: Keyword arguments for Arcee API.
model_name: Name of an Arcee model.
"""
if isinstance(arcee_api_key, str):
arcee_api_key_ = SecretStr(arcee_api_key)
else:
arcee_api_key_ = arcee_api_key
self.arcee_api_key: SecretStr = arcee_api_key_
self.model_kwargs = model_kwargs
self.arcee_api_url = arcee_api_url
self.arcee_api_version = arcee_api_version
try:
route = ArceeRoute.model_training_status.value.format(id_or_name=model_name)
response = self._make_request("get", route)
self.model_id = response.get("model_id")
self.model_training_status = response.get("status")
except Exception as e:
raise ValueError(
f"Error while validating model training status for '{model_name}': {e}"
) from e
def validate_model_training_status(self) -> None:
if self.model_training_status != "training_complete":
raise Exception(
f"Model {self.model_id} is not ready. "
"Please wait for training to complete."
)
def _make_request(
self,
method: Literal["post", "get"],
route: Union[ArceeRoute, str],
body: Optional[Mapping[str, Any]] = None,
params: Optional[dict] = None,
headers: Optional[dict] = None,
) -> dict:
"""Make a request to the Arcee API
Args:
method: The HTTP method to use
route: The route to call
body: The body of the request
params: The query params of the request
headers: The headers of the request
"""
headers = self._make_request_headers(headers=headers)
url = self._make_request_url(route=route)
req_type = getattr(requests, method)
response = req_type(url, json=body, params=params, headers=headers)
if response.status_code not in (200, 201):
raise Exception(f"Failed to make request. Response: {response.text}")
return response.json()
def _make_request_headers(self, headers: Optional[Dict] = None) -> Dict:
headers = headers or {}
if not isinstance(self.arcee_api_key, SecretStr):
raise TypeError(
f"arcee_api_key must be a SecretStr. Got {type(self.arcee_api_key)}"
)
api_key = self.arcee_api_key.get_secret_value()
internal_headers = {
"X-Token": api_key,
"Content-Type": "application/json",
}
headers.update(internal_headers)
return headers
def _make_request_url(self, route: Union[ArceeRoute, str]) -> str:
return f"{self.arcee_api_url}/{self.arcee_api_version}/{route}"
def _make_request_body_for_models(
self, prompt: str, **kwargs: Mapping[str, Any]
) -> Mapping[str, Any]:
"""Make the request body for generate/retrieve models endpoint"""
_model_kwargs = self.model_kwargs or {}
_params = {**_model_kwargs, **kwargs}
filters = [DALMFilter(**f) for f in _params.get("filters", [])]
return dict(
model_id=self.model_id,
query=prompt,
size=_params.get("size", 3),
filters=filters,
id=self.model_id,
)
def generate(
self,
prompt: str,
**kwargs: Any,
) -> str:
"""Generate text from Arcee DALM.
Args:
prompt: Prompt to generate text from.
size: The max number of context results to retrieve. Defaults to 3.
(Can be less if filters are provided).
filters: Filters to apply to the context dataset.
"""
response = self._make_request(
method="post",
route=ArceeRoute.generate.value,
body=self._make_request_body_for_models(
prompt=prompt,
**kwargs,
),
)
return response["text"]
def retrieve(
self,
query: str,
**kwargs: Any,
) -> List[Document]:
"""Retrieve {size} contexts with your retriever for a given query
Args:
query: Query to submit to the model
size: The max number of context results to retrieve. Defaults to 3.
(Can be less if filters are provided).
filters: Filters to apply to the context dataset.
"""
response = self._make_request(
method="post",
route=ArceeRoute.retrieve.value,
body=self._make_request_body_for_models(
prompt=query,
**kwargs,
),
)
return [
ArceeDocumentAdapter.adapt(ArceeDocument(**doc))
for doc in response["results"]
]