How to Summarise GitHub with CrewAI

import os

import requests
from crewai import Agent, Crew, Task
from crewai.tools import BaseTool
from pydantic import BaseModel, Field

ENCONVERT_API_KEY = os.environ["ENCONVERT_API_KEY"]
GITHUB_URL = os.environ["GITHUB_URL"]  # a public GitHub repo, issue, or pull request URL
# CrewAI's default LLM reads OPENAI_API_KEY from the environment


class GitHubUrlInput(BaseModel):
    url: str = Field(description="The GitHub repository, issue, or pull request URL")


class GitHubDistillTool(BaseTool):
    name: str = "GitHub content summariser"
    description: str = (
        "Summarises a GitHub repository, issue, or pull request URL "
        "into clean prose using EnConvert Distill."
    )
    args_schema: type[BaseModel] = GitHubUrlInput

    def _run(self, url: str) -> str:
        # Step 1: Distill the GitHub page into a summary
        response = requests.post(
            "https://api.enconvert.com/v2/distill",
            headers={
                "X-API-Key": ENCONVERT_API_KEY,
                "Content-Type": "application/json",
            },
            json={
                "urls": [url],
                "schema": {
                    "summary": (
                        "a clear summary of the page. For a repository: its "
                        "purpose and README. For an issue: what is raised, its "
                        "status, and any proposed fix. For a pull request: "
                        "what it changes and its review status"
                    ),
                },
            },
            timeout=120,
        )
        response.raise_for_status()

        # Distill reports failures per URL, so check the item, not just the HTTP status
        item = response.json()["results"][0]
        summary = (item.get("data") or {}).get("summary")
        if item.get("error") or not summary:
            raise RuntimeError(
                f"Distill failed for {url}: {item.get('error') or item.get('warnings')}"
            )
        return summary


github_tool = GitHubDistillTool()

# Step 2: Give the tool to a CrewAI agent
analyst = Agent(
    role="GitHub content analyst",
    goal="Summarise GitHub repositories, issues, and pull requests into usable context",
    backstory="You read GitHub content and produce clean summaries for downstream tasks.",
    tools=[github_tool],
    verbose=True,
)

task = Task(
    description=(
        f"Summarise the GitHub page at {GITHUB_URL} "
        "and report what it covers and whether a fix has been proposed."
    ),
    expected_output="A short summary of the page, its current status, and any proposed fix.",
    agent=analyst,
)

# Step 3: Run the crew
crew = Crew(agents=[analyst], tasks=[task])
result = crew.kickoff()
print(result.raw)