crewAIInc
Key error: "description"
On my first run after successfully installing everything i got this, and it kept coming no matter what company i asked for -
Thought: Do I need to use a tool? Yes
Action: yahoo_finance_news
Action Input: TSLATraceback (most recent call last):
File "E:\Project\crewAI-examples\stock_analysis\main.py", line 54, in
To fix it i changed the yahoo_finance_news.py as follows:
From this : doc_strings = [ "\n".join([doc.metadata["title"], doc.metadata["description"]]) for doc in docs if query in doc.metadata["description"] or query in doc.metadata["title"] ]
to this : def _format_results(docs: Iterable[Document], query: str) -> str: doc_strings = [] for doc in docs: title = doc.metadata.get("title", "") description = doc.metadata.get("description", "") if query in description or query in title: doc_strings.append("\n".join([title, description]))
So... is this an issue upstream with yahoo_finance_news?
Anyway, thanks for the code, it works :) I'll paste it here more formattet to make it more accessible for others:
In yahoo_finance_news.py (find the full path in your error message), replace the method on line 60 with:
@staticmethod
def _format_results(docs: Iterable[Document], query: str) -> str:
doc_strings = []
for doc in docs:
title = doc.metadata.get("title", "")
description = doc.metadata.get("description", "")
if query in description or query in title:
doc_strings.append("\n".join([title, description]))
return "\n\n".join(doc_strings)
Sorry about that, messing around, didnt see what i was clicking - so as i mentioned, the issue is in the description field not existing in yahoo finance api. The above is more of a workaround - Heres a working fix that actually gets the links, and tries to get the description should they ever decide to add it:
from typing import Iterable, Optional
from requests.exceptions import HTTPError, ReadTimeout
from urllib3.exceptions import ConnectionError
from langchain.callbacks.manager import CallbackManagerForToolRun
from langchain.document_loaders.web_base import WebBaseLoader
from langchain.schema import Document
from langchain.tools.base import BaseTool
import yfinance as yf
class YahooFinanceNewsTool(BaseTool):
"""Tool that searches financial news on Yahoo Finance."""
name = "yahoo_finance_news"
description = (
"Useful for when you need to find financial news "
"about a public company. "
"Input should be a company ticker. "
"For example, AAPL for Apple, MSFT for Microsoft."
)
top_k: int = 10
"""The number of results to return."""
def _run(
self,
query: str,
run_manager: Optional[CallbackManagerForToolRun] = None,
) -> str:
"""Use the Yahoo Finance News tool."""
company = yf.Ticker(query)
try:
if company.isin is None:
return f"Company ticker {query} not found."
except (HTTPError, ReadTimeout, ConnectionError):
return f"Company ticker {query} not found."
try:
news = company.news
if not news:
return f"No news found for company with ticker {query}."
formatted_news = [self._format_news_item(item) for item in news]
return "\n\n".join(formatted_news)
except (HTTPError, ReadTimeout, ConnectionError) as e:
return f"Error fetching news for ticker {query}: {e}"
@staticmethod
def _format_news_item(news_item):
"""Format a single news item."""
title = news_item.get('title')
link = news_item.get('link')
description = news_item.get('summary') if 'summary' in news_item else 'No description available'
return f"Title: {title}\nLink: {link}\nDescription: {description}\n"
On a side note - you can do away with a lot of langchain stuff if you want. I just set it up as a tool and import from there, and create a yahoo_news_agent who only uses this and the browser_Tools to analyze the articles and it works great.
@cranyy feel free to open a PR with your solution! I'm happy to approve it!