Agents4GovApps

Laboratory of Computational Intelligence (LABIC – ICMC/USP)

Overview

Agents4Gov is a research and development project from LABIC – Institute of Mathematics and Computer Sciences (ICMC/USP) focused on building LLM-based tools to support and modernize public sector services. The project emphasizes local Large Language Models (LLMs) for privacy, data anonymization, and the development and evaluation of tools for use in government and institutional environments.

Installation

# Base package
pip install agents4gov-apps

# With OpenML tools
pip install agents4gov-apps[openml]

# With browser-based tools (Lattes)
pip install agents4gov-apps[browser]

# With GNews collector
pip install agents4gov-apps[gnews]

Quick start

from agents4gov_apps import load_tool_instance
import json

# OpenAlex
tool = load_tool_instance("openalex_doi")
result = json.loads(tool.get_openalex_metadata_by_doi(doi="10.1038/s41586-021-03819-2"))

# GNews (async, with per-window progress log)
import asyncio, logging
from agents4gov_apps.gnews_collector import console_emitter
logging.basicConfig(level=logging.INFO)
tool = load_tool_instance("gnews_collector")
result = asyncio.run(tool.collect_general_news(
    query='"ICMC USP"',
    start_year_month="2023-01",
    end_year_month="2024-12",
    __event_emitter__=console_emitter(),
))

Research pipelines

End-to-end scripts that drive the bundled tools for concrete research tasks. They use only the public API (load_tool_instance + the tool's documented methods) so they keep working with any pip install agents4gov-apps version.

Gender-violence news monitoring (10-year window)

Crawls Brazilian Portuguese news for a curated list of gender-violence queries (read from an Excel file with query + categoria da query columns), then produces initial indicators: frequency by category over time, top publishers, spike detection.

Pipeline files (under scripts/):

• collect_violencia_genero.py — drives gnews_collector.collect_general_news one query at a time across monthly windows. Idempotent (skips parquets already on disk via the library's built-in checkpoint), supports SerpAPI primary + gnews fallback, and exports an Excel workbook with one sheet per category.
• analyze_violencia_genero.py — reads the parquets, joins the category from the queries Excel, and writes a report workbook with freq_yearly, freq_monthly, top_publishers, top_publishers_por_cat, and spikes sheets, plus optional PNG/SVG plots.

Usage:

pip install agents4gov-apps[gnews] openpyxl python-dotenv matplotlib

# 1) Smoke test (one query, one month)
SERPAPI_KEY=... python3 scripts/collect_violencia_genero.py \
    --auto-fallback --limit-queries 1 --start 2024-01 --end 2024-01

# 2) Full collection (81 queries × 121 months, ~10-14h on SerpAPI)
SERPAPI_KEY=... python3 scripts/collect_violencia_genero.py --auto-fallback \
    2>&1 | tee collect.log

# 3) Indicators
python3 scripts/analyze_violencia_genero.py \
    --output-dir ./gnews_output_violencia_genero \
    --queries-xlsx ./gnews_queries_violencia_genero.xlsx

The collector is resumable — interrupt with Ctrl+C and re-run; it skips every window whose parquet is already on disk.

Adapt the same shape to other query sets by replacing the Excel file (any two-column query + categoria da query workbook works).

Documentation

• Developer Guide — architecture, standard interface, how to create and use tools across Open WebUI, LangChain, OpenAI, and other frameworks
• Tool Protocol — quick-reference packaging contract
• Available Tools — full list of tools with source links