phucdev's picture
Rework wiki search by directly using Wikimedia API and RetrievalQA chain
823bd24
raw
history blame
15.3 kB
import base64
import os
from typing import Optional
import pandas as pd
import requests
import whisper
from bs4 import BeautifulSoup
from datetime import datetime
from dotenv import find_dotenv, load_dotenv
from langchain.chains import RetrievalQA
from langchain.chat_models import init_chat_model
from langchain_community.document_loaders import (
UnstructuredPDFLoader, UnstructuredPowerPointLoader,
UnstructuredWordDocumentLoader, WebBaseLoader)
from langchain_community.tools import DuckDuckGoSearchRun
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.tools import tool
from langchain.schema import Document
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_huggingface.embeddings import HuggingFaceEmbeddings
from markdownify import markdownify as md
from youtube_transcript_api import YouTubeTranscriptApi
from yt_dlp import YoutubeDL
UNWANTED_SECTIONS = {
"references",
"external links",
"further reading",
"see also",
"notes",
}
@tool
def get_weather_info(location: str) -> str:
"""Fetches weather information for a given location.
Usage:
```
# Initialize the tool
weather_info_tool = Tool(
name="get_weather_info",
func=get_weather_info,
description="Fetches weather information for a given location.")
```
"""
load_dotenv(find_dotenv())
api_key = os.getenv("OPENWEATHERMAP_API_KEY")
url = (
f"https://api.openweathermap.org/data/2.5/"
f"weather?q={location}&appid={api_key}&units=metric"
)
res = requests.get(url, timeout=15)
data = res.json()
humidity = data["main"]["humidity"]
pressure = data["main"]["pressure"]
wind = data["wind"]["speed"]
description = data["weather"][0]["description"]
temp = data["main"]["temp"]
min_temp = data["main"]["temp_min"]
max_temp = data["main"]["temp_max"]
return (
f"Weather in {location}: {description}, "
f"Temperature: {temp}°C, Min: {min_temp}°C, Max: {max_temp}°C, "
f"Humidity: {humidity}%, Pressure: {pressure} hPa, "
f"Wind Speed: {wind} m/s"
)
@tool
def add(a: int, b: int) -> int:
"""Adds two numbers together.
Args:
a (int): The first number.
b (int): The second number.
"""
return a + b
@tool
def get_sum(list_of_numbers: list[int]) -> int:
"""Sums a list of numbers.
Args:
list_of_numbers (list[int]): The list of numbers to sum.
"""
return sum(list_of_numbers)
@tool
def subtract(a: int, b: int) -> int:
"""Subtracts the second number from the first.
Args:
a (int): The first number.
b (int): The second number.
"""
return a - b
@tool
def multiply(a: int, b: int) -> int:
"""Multiplies two numbers together.
Args:
a (int): The first number.
b (int): The second number.
"""
return a * b
@tool
def divide(a: int, b: int) -> float:
"""Divides the first number by the second.
Args:
a (int): The first number.
b (int): The second number.
"""
if b == 0:
raise ValueError("Cannot divide by zero.")
return a / b
@tool
def get_current_time_and_date() -> str:
"""Returns the current time and date in ISO format."""
return datetime.now().isoformat()
@tool
def reverse_text(text: str) -> str:
"""Reverses the given text.
Args:
text (str): The text to reverse.
"""
return text[::-1]
def build_retriever(text: str):
"""Builds a retriever from the given text.
Args:
text (str): The text to be used for retrieval.
"""
splitter = RecursiveCharacterTextSplitter(
separators=["\n### ", "\n## ", "\n# "],
chunk_size=1000,
chunk_overlap=200,
)
chunks = splitter.split_text(text)
docs = [
Document(page_content=chunk)
for chunk in chunks
]
hf_embed = HuggingFaceEmbeddings(
model_name="sentence-transformers/all-MiniLM-L6-v2"
)
index = FAISS.from_documents(docs, hf_embed)
return index.as_retriever(search_kwargs={"k": 3})
def get_retrieval_qa(text: str):
"""Creates a RetrievalQA instance for the given text.
Args:
text (str): The text to be used for retrieval.
"""
retriever = build_retriever(text)
llm = init_chat_model("groq:meta-llama/llama-4-scout-17b-16e-instruct")
return RetrievalQA.from_chain_type(
llm=llm,
chain_type="stuff",
retriever=retriever,
return_source_documents=True,
)
def clean_html(html: str) -> str:
soup = BeautifulSoup(html, "html.parser")
# 1. Remove <script> & <style>
for tag in soup(["script", "style"]):
tag.decompose()
# 2. Drop whole <section> blocks whose first heading is unwanted
for sec in soup.find_all("section"):
h = sec.find(["h1","h2","h3","h4","h5","h6"])
if h and any(h.get_text(strip=True).lower().startswith(u) for u in UNWANTED_SECTIONS):
sec.decompose()
# 3. Additional filtering by CSS selector
for selector in [".toc", ".navbox", ".vertical-navbox", ".hatnote", ".reflist", ".mw-references-wrap"]:
for el in soup.select(selector):
el.decompose()
# 4. Isolate the main content container if present
main = soup.find("div", class_="mw-parser-output")
return str(main or soup)
def get_wikipedia_article(query: str, lang: str = "en") -> str:
"""Fetches a Wikipedia article for a given query and returns its content in Markdown format.
Args:
query (str): The search query.
lang (str): The language code for the search. Default is "en".
"""
headers = {
'User-Agent': 'MyLLMAgent ([email protected])'
}
# Step 1: Search
search_url = f"https://api.wikimedia.org/core/v1/wikipedia/{lang}/search/page"
search_params = {'q': query, 'limit': 1}
search_response = requests.get(search_url, headers=headers, params=search_params, timeout=15)
if search_response.status_code != 200:
return f"Search error: {search_response.status_code}"
results = search_response.json().get("pages", [])
if not results:
return "No results found."
page = results[0]
page_key = page["key"]
# Step 2: Get the wiki page, only keep relevant content and convert to Markdown
content_url = f"https://api.wikimedia.org/core/v1/wikipedia/{lang}/page/{page_key}/html"
content_response = requests.get(content_url, timeout=15)
if content_response.status_code != 200:
return f"Content fetch error: {content_response.status_code}"
html = clean_html(content_response.text)
markdown = md(
html,
heading_style="ATX",
bullets="*+-",
table_infer_header=True,
strip=['a', 'span']
)
return markdown
@tool
def wiki_search(query: str, question: str, lang: str="en") -> str:
"""Searches Wikipedia for a specific article and answers a question based on its content.
The function retrieves a Wikipedia article based on the provided query, converts it to Markdown,
and uses a retrieval-based QA system to answer the specified question.
Args:
query (str): A concise topic name with optional keywords, ideally matching the relevant Wikipedia page title.
question (str): The question to answer using the article.
lang (str): Language code for the Wikipedia edition to search (default: "en").
"""
markdown = get_wikipedia_article(query, lang)
qa = get_retrieval_qa(markdown)
return qa.invoke(question)
@tool
def web_search(query: str) -> str:
"""Searches the web for a given query and returns the first result.
Args:
query (str): The search query.
"""
search_tool = DuckDuckGoSearchRun()
results = search_tool.invoke(query)
if results:
return results
else:
return "No results found."
@tool
def visit_website(url: str) -> str:
"""Visits a website and returns the content.
Args:
url (str): The URL of the website to visit.
"""
loader = WebBaseLoader(url)
documents = loader.load()
if documents:
return documents[0].page_content
else:
return "No content found."
@tool
def get_youtube_transcript(video_url: str, return_timestamps: bool = False) -> str:
"""Fetches the transcript of a YouTube video.
Args:
video_url (str): The URL of the YouTube video.
return_timestamps (bool): If True, returns timestamps with the transcript. Otherwise, returns only the text.
"""
try:
video_id = video_url.split("v=")[-1]
transcript = YouTubeTranscriptApi.get_transcript(video_id)
if return_timestamps:
sentences = []
for t in transcript:
start = t["start"]
end = start + t["duration"]
sentences.append(f"{start:.2f} - {end:.2f}: {t['text']}")
return "\n".join(sentences)
else:
return "\n".join([t["text"] for t in transcript])
except Exception as e:
return f"Error fetching transcript: {e}"
@tool
def get_youtube_video_info(video_url: str) -> str:
"""Fetches information about a YouTube video.
Args:
video_url (str): The URL of the YouTube video.
"""
try:
ydl_opts = {
"quiet": True,
"skip_download": True,
}
with YoutubeDL(ydl_opts) as ydl:
info = ydl.extract_info(video_url, download=False)
video_info = {
"Title": info.get("title"),
"Description": info.get("description"),
"Uploader": info.get("uploader"),
"Upload date": info.get("upload_date"),
"Duration": info.get("duration"),
"View count": info.get("view_count"),
"Like count": info.get("like_count"),
}
video_info_filtered = {k: v for k, v in video_info.items() if v is not None}
video_info_str = "\n".join(
[f"{k}: {v}" for k, v in video_info_filtered.items()]
)
return video_info_str
except Exception as e:
return f"Error fetching video info: {e}"
def encode_image(image_path):
with open(image_path, "rb") as image_file:
return base64.b64encode(image_file.read()).decode("utf-8")
@tool
def ask_about_image(image_path: str, question: str) -> str:
"""Performs vision-based question answering on an image.
Args:
image_path (str): The path to the image file.
question (str): Your question about the image, as a natural language sentence. Provide as much context as possible.
"""
load_dotenv(find_dotenv())
llm = init_chat_model("groq:meta-llama/llama-4-maverick-17b-128e-instruct")
prompt = ChatPromptTemplate(
[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Please write a concise caption for the image that helps answer the following question: {question}",
},
{
"type": "image_url",
"image_url": {
"url": "data:image/{image_format};base64,{base64_image}",
},
},
],
}
]
)
file_suffix = os.path.splitext(image_path)[-1]
if file_suffix == ".png":
image_format = "png"
else:
# We could handle other formats explicitly, but for simplicity we assume JPEG
image_format = "jpeg"
chain = prompt | llm
response = chain.invoke(
{
"question": question,
"base64_image": encode_image(image_path),
"image_format": image_format,
}
)
return response.text()
def transcribe_audio(audio_path: str) -> str:
"""Transcribes audio to text.
Args:
audio_path (str): The path to the audio file.
"""
model = whisper.load_model("base")
result = model.transcribe(audio_path)
text = result.text
return text
def get_table_description(table: pd.DataFrame) -> str:
"""Generates a description of the table. If applicable, calculates sum and mean of numeric
columns.
Args:
table (pd.DataFrame): The table to describe.
"""
if table.empty:
return "The table is empty."
description = []
total_sum = 0
for column in table.select_dtypes(include=[int, float]).columns:
column_sum = table[column].sum()
column_mean = table[column].mean()
description.append(
f"Column '{column}': Sum = {column_sum}, Mean = {column_mean:.2f}"
)
total_sum += column_sum
if total_sum:
description.append(f"Total Sum of all numeric columns: {total_sum}")
if description:
description = "\n".join(description)
else:
description = "No numeric columns to summarize."
# Add the number of rows and columns
description += f"\n\nTable has {table.shape[0]} rows and {table.shape[1]} columns."
df_as_markdown = table.to_markdown()
description += f"\n\nTable:\n{df_as_markdown}"
return description
@tool
def inspect_file_as_text(file_path: str) -> str:
"""This tool reads a file as markdown text. It handles [".csv", ".xlsx", ".pptx", ".wav",
".mp3", ".m4a", ".flac", ".pdf", ".docx"], and all other types of text files. IT DOES NOT
HANDLE IMAGES.
Args:
file_path (str): The path to the file you want to read as text. If it is an image, use `vision_qa` tool.
"""
# TODO we could also pass the file content to a retrieval chain
try:
suffix = os.path.splitext(file_path)[-1]
if suffix in [".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff"]:
raise Exception(
"Cannot use inspect_file_as_text tool with images: use `vision_qa` tool instead!"
)
if suffix in [".csv", ".tsv", ".xlsx"]:
if suffix == ".csv":
df = pd.read_csv(file_path)
elif suffix == ".tsv":
df = pd.read_csv(file_path, sep="\t")
elif suffix == ".xlsx":
df = pd.read_excel(file_path)
else:
raise Exception(f"Unsupported file type: {suffix}")
table_description = get_table_description(df)
return table_description
elif suffix == ".pptx":
doc = UnstructuredPowerPointLoader(file_path)
return doc.load()[0].page_content
elif suffix == ".pdf":
doc = UnstructuredPDFLoader(file_path)
return doc.load()[0].page_content
elif suffix == ".docx":
doc = UnstructuredWordDocumentLoader(file_path)
return doc.load()[0].page_content
elif suffix in [".wav", ".mp3", ".m4a", ".flac"]:
return transcribe_audio(file_path)
else:
# All other text files
with open(file_path, "r", encoding="utf-8") as file:
content = file.read()
return content
except Exception as e:
return f"Error file: {e}"