README.md

Tourist🤳

An open-source, low-cost, serverless application for SERP extraction and web scraping.

Work on LLM projects without worrying about credits, subscriptions, or rate-limits. Tourist is a free alternative to many mainstream SERP API services. Run Tourist on your machine or deploy it into your own AWS account.

Important

Tourist is in early development. Features and API's may change unexpectedly.

Overview

Tourist has both Service and Client components. The Service (HTTP API) handles requests from the Client (your app, agent, or scraper scripts). You're in control of both components! None of your data is ever processed or stored by third parties.

Service

Local deployment (for testing...)

Tip

Docker is recommended for running Tourist locally to handle dependencies for headless browsing.

1. Have Docker installed
2. docker pull ghcr.io/pogzyb/tourist:latest
3. docker run -p 8000:8000 ghcr.io/pogzyb/tourist:latest

If the service came up correctly, you should see:

INFO:     Started server process [1]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Check the docs at http://localhost:8000/docs

AWS deployment (for real...)

Deploy your own instance of Tourist into AWS with Terraform:

1. Have Docker installed
2. Clone this repo
3. Have an AWS account with credentials copied to .env.aws in the root of this project
4. make tourist-iac-interactive
5. terraform apply - deploys the infrastructure into your AWS account

Use your endpoint: https://<uuid>.execute-api.us-east-1.amazonaws.com/main (available in terraform outputs)

Warning

Tourist uses serverless infrastructure to keep costs extremely low; however these costs will not be $0.00 depending on how heavily you use your API.

Important

Tourist uses the X-SECRET authorization header to protect your API, you should set this value in terraform/aws/terraform.tfvars

Client

Build your own LLM tools, web scraping apps, or automated testing workflows with the Tourist client.

Python

You can use the python client to interact with your Tourist service. Check out the examples folder for the complete code.

pip install tourist

LLM Tools

For example, create a LangChain Tool for your LLM Agent.

from bs4 import BeautifulSoup as bs
from tourist.core import TouristScraper
from langchain_core.tools import tool

# Assumes you're running locally,
# change this to your cloud endpoint if you've deployed via terraform.
scraper = TouristScraper(
    "http://localhost:8000",
    secret="doesntmatterlocally",  # authorization secret  
    concurrency=1,  # control concurrent searches/scrapes
)


@tool
def scrape_tool(url: str) -> str:
    """
    A web scraper. 
    Useful for when you need to answer questions related the contents of a website or URL.
    Input should be a URL.
    """
    results = scraper.get_page(url)
    if source_html := results.get("source_html"):
        soup = bs(source_html, "html.parser")
        return soup.get_text()
    else:
        return "Could not scrape that page. Try again."


@tool
def search_tool(query: str) -> str:
    """
    A search tool.
    Useful for when you need to answer questions about current events, people, places, or things.
    Input should be a search query.
    """
    results = scraper.get_serp(query, max_results=3):
    return " ".join([r["content"] for r in results])


# ... use the tools

Selenium

You can submit selenium scripts (human or AI-generated) to Tourist for execution.

from pprint import pprint

from tourist.core import TouristScraper


scraper = TouristScraper("http://localhost:8000", "no-secret")

my_selenium_code = """
# `driver`, a selenium `webdriver.Chrome` object, is available in globals
driver.get("https://www.example.com")
html = driver.page_source
# any key:values stored in `actions_output` will be available in the response
actions_output["html"] = html
actions_output["current_url"] = driver.current_url
"""

result = scraper.get_page_with_actions(my_selenium_code)
assert result["current_url"] == "https://www.example.com/"
pprint(result["html"])

Contributions

This is an open-source project. Please consider adding improvements or features related to your specific use-case. Chances are someone else is also facing the issue or limitation. Some ready-to-do tasks can be found in the source code as TODO/Contribution: ....

To run the Tourist service on your local machine for testing or prototyping:

1. Have Docker installed
2. Clone this repo
3. Add your contributions/modifications
4. make tourist-local - build the container from source code in src/

Credits

Components of this repository were influenced by these projects! Check them out.

• awslabs/aws-lambda-web-adapter
• umihico/docker-selenium-lambda