What is smolagents?
Hugging Face's lightweight framework for building AI agents. Minimal, code-first approach to agent development.
Searching Google Scholar with smolagents
This integration lets your smolagents agent search Google Scholar in real time via the Scavio API. The agent gets back structured JSON with paper titles, paper URLs, SERP snippets -- ready for reasoning and decision-making.
Setup
Bash
pip install smolagents requestsCode Example
Here is a complete smolagents agent that searches Google Scholar using Scavio:
Python
from smolagents import tool, CodeAgent, HfApiModel
import json
import os
import requests
# Scavio has no Google Scholar endpoint, so citation counts and author lists are not available. This runs a Google web search - narrow it with site:arxiv.org or filetype:pdf.
@tool
def scavio_search(query: str) -> str:
"""Query Google Scholar through the Scavio API (POST /api/v2/google).
Args:
query: What to look for.
"""
response = requests.post(
"https://api.scavio.dev/api/v2/google",
headers={
"Authorization": "Bearer " + os.environ["SCAVIO_API_KEY"],
"Content-Type": "application/json",
},
json={"query": query},
timeout=60,
)
response.raise_for_status()
return json.dumps(response.json())
agent = CodeAgent(tools=[scavio_search], model=HfApiModel())
result = agent.run("retrieval augmented generation 2024")
print(result)Full Working Example
A production-ready example with error handling:
Python
from smolagents import tool, CodeAgent, HfApiModel
import json
import os
import requests
# Scavio has no Google Scholar endpoint, so citation counts and author lists are not available. This runs a Google web search - narrow it with site:arxiv.org or filetype:pdf.
@tool
def scavio_search(query: str) -> str:
"""Query Google Scholar through the Scavio API (POST /api/v2/google).
Args:
query: What to look for.
"""
response = requests.post(
"https://api.scavio.dev/api/v2/google",
headers={
"Authorization": "Bearer " + os.environ["SCAVIO_API_KEY"],
"Content-Type": "application/json",
},
json={"query": query},
timeout=60,
)
response.raise_for_status()
return json.dumps(response.json())
agent = CodeAgent(tools=[scavio_search], model=HfApiModel())
result = agent.run("retrieval augmented generation 2024")
print(result)Pricing
Scavio offers a free tier with 50 credits on signup (1 credit per search). No credit card required. This is enough to build and test your smolagents integration. Paid plans start at $30/month for higher volumes.