mirror of
https://github.com/hwchase17/langchain
synced 2024-11-08 07:10:35 +00:00
54adcd9e82
We add a tool and retriever for the [AskNews](https://asknews.app) platform with example notebooks. The retriever can be invoked with: ```py from langchain_community.retrievers import AskNewsRetriever retriever = AskNewsRetriever(k=3) retriever.invoke("impact of fed policy on the tech sector") ``` To retrieve 3 documents in then news related to fed policy impacts on the tech sector. The included notebook also includes deeper details about controlling filters such as category and time, as well as including the retriever in a chain. The tool is quite interesting, as it allows the agent to decide how to obtain the news by forming a query and deciding how far back in time to look for the news: ```py from langchain_community.tools.asknews import AskNewsSearch from langchain import hub from langchain.agents import AgentExecutor, create_openai_functions_agent from langchain_openai import ChatOpenAI tool = AskNewsSearch() instructions = """You are an assistant.""" base_prompt = hub.pull("langchain-ai/openai-functions-template") prompt = base_prompt.partial(instructions=instructions) llm = ChatOpenAI(temperature=0) asknews_tool = AskNewsSearch() tools = [asknews_tool] agent = create_openai_functions_agent(llm, tools, prompt) agent_executor = AgentExecutor( agent=agent, tools=tools, verbose=True, ) agent_executor.invoke({"input": "How is the tech sector being affected by fed policy?"}) ``` --------- Co-authored-by: Emre <e@emre.pm>
147 lines
4.7 KiB
Python
147 lines
4.7 KiB
Python
import os
|
|
import re
|
|
from typing import Any, Dict, List, Literal, Optional
|
|
|
|
from langchain_core.callbacks import (
|
|
AsyncCallbackManagerForRetrieverRun,
|
|
CallbackManagerForRetrieverRun,
|
|
)
|
|
from langchain_core.documents import Document
|
|
from langchain_core.retrievers import BaseRetriever
|
|
|
|
|
|
class AskNewsRetriever(BaseRetriever):
|
|
"""AskNews retriever."""
|
|
|
|
k: int = 10
|
|
offset: int = 0
|
|
start_timestamp: Optional[int] = None
|
|
end_timestamp: Optional[int] = None
|
|
method: Literal["nl", "kw"] = "nl"
|
|
categories: List[
|
|
Literal[
|
|
"All",
|
|
"Business",
|
|
"Crime",
|
|
"Politics",
|
|
"Science",
|
|
"Sports",
|
|
"Technology",
|
|
"Military",
|
|
"Health",
|
|
"Entertainment",
|
|
"Finance",
|
|
"Culture",
|
|
"Climate",
|
|
"Environment",
|
|
"World",
|
|
]
|
|
] = ["All"]
|
|
historical: bool = False
|
|
similarity_score_threshold: float = 0.5
|
|
kwargs: Optional[Dict[str, Any]] = {}
|
|
client_id: Optional[str] = None
|
|
client_secret: Optional[str] = None
|
|
|
|
def _get_relevant_documents(
|
|
self, query: str, *, run_manager: CallbackManagerForRetrieverRun
|
|
) -> List[Document]:
|
|
"""Get documents relevant to a query.
|
|
Args:
|
|
query: String to find relevant documents for
|
|
run_manager: The callbacks handler to use
|
|
Returns:
|
|
List of relevant documents
|
|
"""
|
|
try:
|
|
from asknews_sdk import AskNewsSDK
|
|
except ImportError:
|
|
raise ImportError(
|
|
"AskNews python package not found. "
|
|
"Please install it with `pip install asknews`."
|
|
)
|
|
an_client = AskNewsSDK(
|
|
client_id=self.client_id or os.environ["ASKNEWS_CLIENT_ID"],
|
|
client_secret=self.client_secret or os.environ["ASKNEWS_CLIENT_SECRET"],
|
|
scopes=["news"],
|
|
)
|
|
response = an_client.news.search_news(
|
|
query=query,
|
|
n_articles=self.k,
|
|
start_timestamp=self.start_timestamp,
|
|
end_timestamp=self.end_timestamp,
|
|
method=self.method,
|
|
categories=self.categories,
|
|
historical=self.historical,
|
|
similarity_score_threshold=self.similarity_score_threshold,
|
|
offset=self.offset,
|
|
doc_start_delimiter="<doc>",
|
|
doc_end_delimiter="</doc>",
|
|
return_type="both",
|
|
**self.kwargs,
|
|
)
|
|
|
|
return self._extract_documents(response)
|
|
|
|
async def _aget_relevant_documents(
|
|
self, query: str, *, run_manager: AsyncCallbackManagerForRetrieverRun
|
|
) -> List[Document]:
|
|
"""Asynchronously get documents relevant to a query.
|
|
Args:
|
|
query: String to find relevant documents for
|
|
run_manager: The callbacks handler to use
|
|
Returns:
|
|
List of relevant documents
|
|
"""
|
|
try:
|
|
from asknews_sdk import AsyncAskNewsSDK
|
|
except ImportError:
|
|
raise ImportError(
|
|
"AskNews python package not found. "
|
|
"Please install it with `pip install asknews`."
|
|
)
|
|
an_client = AsyncAskNewsSDK(
|
|
client_id=self.client_id or os.environ["ASKNEWS_CLIENT_ID"],
|
|
client_secret=self.client_secret or os.environ["ASKNEWS_CLIENT_SECRET"],
|
|
scopes=["news"],
|
|
)
|
|
response = await an_client.news.search_news(
|
|
query=query,
|
|
n_articles=self.k,
|
|
start_timestamp=self.start_timestamp,
|
|
end_timestamp=self.end_timestamp,
|
|
method=self.method,
|
|
categories=self.categories,
|
|
historical=self.historical,
|
|
similarity_score_threshold=self.similarity_score_threshold,
|
|
offset=self.offset,
|
|
return_type="both",
|
|
doc_start_delimiter="<doc>",
|
|
doc_end_delimiter="</doc>",
|
|
**self.kwargs,
|
|
)
|
|
|
|
return self._extract_documents(response)
|
|
|
|
def _extract_documents(self, response: Any) -> List[Document]:
|
|
"""Extract documents from an api response."""
|
|
|
|
from asknews_sdk.dto.news import SearchResponse
|
|
|
|
sr: SearchResponse = response
|
|
matches = re.findall(r"<doc>(.*?)</doc>", sr.as_string, re.DOTALL)
|
|
docs = [
|
|
Document(
|
|
page_content=matches[i].strip(),
|
|
metadata={
|
|
"title": sr.as_dicts[i].title,
|
|
"source": str(sr.as_dicts[i].article_url)
|
|
if sr.as_dicts[i].article_url
|
|
else None,
|
|
"images": sr.as_dicts[i].image_url,
|
|
},
|
|
)
|
|
for i in range(len(matches))
|
|
]
|
|
return docs
|