Final_Assignment / agent.py
znacer's picture
download attached files
dd46710 unverified
import math
import os
import re
import numexpr
from langchain_community.document_loaders import ArxivLoader
import requests
from smolagents import (
CodeAgent,
GoogleSearchTool,
InferenceClientModel,
WikipediaSearchTool,
tool,
)
hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
@tool
def calculator(expression: str) -> str:
"""Calculate expression using Python's numexpr library.
Expression should be a single line mathematical expression
that solves the problem.
Examples:
"37593 * 67" for "37593 times 67"
"37593**(1/5)" for "37593^(1/5)"
Args:
expression: the expression to be compute
"""
local_dict = {"pi": math.pi, "e": math.e}
return str(
numexpr.evaluate(
expression.strip(),
global_dict={}, # restrict access to globals
local_dict=local_dict, # add common mathematical functions
)
)
@tool
def arxiv_search_tool(query: str) -> str:
"""
Searches Arxiv and returns a summary or full text of the given topic of the 5 most likely related papers, along with the page URL.
Args:
query: the topic to be searched
"""
docs = ArxivLoader(query=query, load_max_docs=5).load()
if len(docs) > 0:
return "\n".join(
[
f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
for doc in docs
]
)
return "No content found"
@tool
def download_attached_files(task_id: str) -> str:
"""
Get the first attached file if exists and save it. Returns the filename.
Args:
task_id: this id of the task
"""
api_url = DEFAULT_API_URL
file_url = f"{api_url}/files/{task_id}"
response = requests.get(file_url, timeout=15)
if response.status_code == 200:
cd = response.headers.get("content-disposition")
if cd:
fname = re.findall("filename=(.+)", cd)
if len(fname) > 0:
with open(fname[0], "wb") as file:
file.write(response.content)
return f"{fname[0]} saved "
return "no file attached to this task"
tools = [
calculator,
arxiv_search_tool,
download_attached_files,
GoogleSearchTool(provider="serper"),
WikipediaSearchTool(),
]
model_id = "Qwen/Qwen3-235B-A22B"
model = InferenceClientModel(
model_id=model_id, token=hf_token
) # You can choose to not pass any model_id to InferenceClientModel to use a default model
agent = CodeAgent(
model=model,
tools=tools,
add_base_tools=True,
additional_authorized_imports=["pandas", "numpy", "csv", "subprocess"],
)