Brazilian Law Research MCP Server

import contextlib import logging from typing import TYPE_CHECKING, Self, override from patchright.async_api import TimeoutError from pydantic import field_validator from brlaw_mcp_server.domain.base import BaseLegalPrecedent if TYPE_CHECKING: from patchright.async_api import Page _LOGGER = logging.getLogger(__name__) class TstLegalPrecedent(BaseLegalPrecedent): """Model for a legal precedent from the Tribunal Superior do Trabalho (TST).""" @field_validator("summary") @classmethod def _remove_style_elements_from_summary(cls, v: str) -> str: """Remove style elements from the summary. On the TST website, the summary is split into multiple elements. In some cases, there's a <style> element among the <p> elements. Its data is not relevant at all to the summary, so we need to filter it out.""" if not v.startswith("<!--"): return v for char_idx, char in enumerate(v): if char == ">" and v[char_idx - 2 : char_idx + 1] == "-->": return v[char_idx + 1 :].strip() raise RuntimeError( "Could not find the end of the style element inside the summary" ) @override @classmethod async def research( cls, browser: "Page", *, summary_search_prompt: str, desired_page: int = 1 ) -> "list[Self]": _LOGGER.info( "Starting research for legal precedents authored by the TST with the summary search prompt %s", repr(summary_search_prompt), ) await browser.goto("https://jurisprudencia.tst.jus.br/") with contextlib.suppress(TimeoutError): await ( browser.locator("span[class^='jss']") .filter(has_text="Fechar") .click(timeout=1000) ) locator_summary_input = browser.locator("#campoTxtEmenta") await locator_summary_input.fill(summary_search_prompt) await locator_summary_input.press("Enter") await browser.locator("circle").wait_for(state="hidden", timeout=1000 * 30) precedents = [ cls(summary=text) for locator in await browser.locator("div[id^=celulaLeiaMaisAcordao]").all() if (text := await locator.text_content()) is not None ] _LOGGER.info( "Found %d legal precedents", len(precedents), ) return precedents