CrewAI

"""CrewAI with Boxline: a CrewAI tool (BaseTool) that reads pages through Boxline, used by a one-agent crew on yourown OpenAI key. Each page renders in a real browser inside a Boxline sandbox, never on this computer.    pip install -r requirements.txt && python main.py     (BOXLINE_API_KEY and OPENAI_API_KEY; CREW_MODEL, QUESTION)Writes output/result.json: the answer and every tool call."""import jsonimport osfrom pathlib import Pathfrom crewai import LLM, Agent, Crew, Taskfrom crewai.tools import BaseToolfrom pydantic import BaseModel, Fieldfrom boxline import Boxlineout = Path(os.environ.get("OUTPUT_DIR", "output"))question = os.environ.get("QUESTION", 'On books.toscrape.com, what does "Sapiens: A Brief History of Humankind" cost, and how many are in stock?')bx = Boxline()calls = []class FetchInput(BaseModel):    url: str = Field(..., description="an http(s) address")class BoxlineFetchTool(BaseTool):    name: str = "boxline_fetch"    description: str = "Open a web page in a real browser (in a Boxline sandbox) and return it as Markdown with its links."    args_schema: type[BaseModel] = FetchInput    def _run(self, url: str) -> str:        calls.append({"tool": self.name, "input": {"url": url}})        page = bx.fetch(url, format="markdown", links=True)        return json.dumps({"url": page["finalUrl"], "status": page["status"], "title": page["title"], "content": page["content"][:12_000],                           "links": (page.get("links") or [])[:80]})checker = Agent(    role="Web researcher",    goal="Answer questions about web pages from what the pages say",    backstory="You read pages with the boxline_fetch tool and quote exact figures.",    tools=[BoxlineFetchTool()],    llm=LLM(model=os.environ.get("CREW_MODEL", "openai/gpt-6-luna")),    verbose=False,)task = Task(description=question, expected_output="The answer with the exact figures, and the address of the page they came from.", agent=checker)output = Crew(agents=[checker], tasks=[task], verbose=False).kickoff()answer = output.rawfor c in calls:    print(f"-> {c['tool']} {json.dumps(c['input'])}")print(f"\n{answer}")out.mkdir(parents=True, exist_ok=True)usage = output.token_usage.total_tokens if output.token_usage else 0(out / "result.json").write_text(json.dumps({"question": question, "answer": answer, "calls": calls, "usage": {"ownModelTokens": usage}}, indent=2, ensure_ascii=False))

# Python 3.10 to 3.13 or newerexport BOXLINE_API_KEY=…        # your Boxline API keyexport OPENAI_API_KEY=…         # your own OpenAI keypip install boxline-sdk crewai==1.15.23python main.py