Skip to content

Release Notes v0.1.0

Release Date: December 13, 2025

This is the initial release of Statistics Agent Team, a multi-agent system for finding and verifying statistics from reputable web sources.

Overview

Statistics Agent Team is built with Google ADK (Agent Development Kit) and Eino, implementing a sophisticated multi-agent architecture that leverages LLMs and web search to find verifiable statistics from well-known and respected publishers.

Features

3-Agent Architecture

The system implements three specialized agents:

Research Agent (agents/research/)

  • Built with Google ADK and Gemini 2.0 Flash model
  • Executes web searches for statistics on given topics
  • Prioritizes reputable sources (academic, government, research organizations)
  • Extracts candidate statistics with context
  • Returns structured data for verification
  • Port: 8001

Verification Agent (agents/verification/)

  • Built with Google ADK and Gemini 2.0 Flash model
  • Fetches actual source content from URLs
  • Searches for verbatim excerpts in source
  • Validates numerical values match exactly
  • Flags hallucinations and discrepancies
  • Returns verification results with reasons
  • Port: 8002

Orchestration Agents (Two Options)

ADK Orchestration (agents/orchestration/) - Built with Google ADK and Gemini 2.0 Flash model - LLM-based decision making for workflow coordination - Implements adaptive retry logic - Dynamic quality control - Port: 8000

Eino Orchestration (agents/orchestration-eino/) - Recommended - Deterministic graph-based workflow - Type-safe orchestration with compile-time checks - Predictable, reproducible behavior - Faster and lower cost (no LLM for orchestration) - Port: 8003

MCP Server Integration

  • Full MCP (Model Context Protocol) server implementation
  • Integration with Claude Code and other MCP clients
  • Exposes Eino orchestration via MCP tools
  • mcp/server/main.go - Server implementation
  • MCP_SERVER.md - Integration documentation

Multi-LLM Provider Support

Configurable LLM providers with unified interface: - Gemini (default) - Google's Gemini 2.0 Flash - Claude - Anthropic Claude models - OpenAI - GPT-4 and GPT-3.5 - Ollama - Local LLM deployment

Configuration via environment variables documented in LLM_CONFIGURATION.md.

Structured Output

Statistics returned in JSON format with complete metadata:

{
  "name": "Global temperature increase since pre-industrial times",
  "value": 1.1,
  "unit": "°C",
  "source": "IPCC Sixth Assessment Report",
  "source_url": "https://www.ipcc.ch/...",
  "excerpt": "Global surface temperature has increased by approximately 1.1°C...",
  "verified": true,
  "date_found": "2025-12-13T10:30:00Z"
}

Core Capabilities

  • Multi-agent orchestration with workflow coordination
  • Google ADK integration for LLM-based agents
  • Eino framework for deterministic graph orchestration
  • Source verification to prevent hallucinations
  • Reputable source prioritization
  • HTTP APIs for all agents
  • Retry logic for ensuring quality results
  • Function tools for structured agent capabilities

Project Structure

stats-agent-team/
├── agents/
│   ├── orchestration/       # ADK-based orchestration agent
│   ├── orchestration-eino/  # Eino-based orchestration agent
│   ├── research/            # Research agent
│   └── verification/        # Verification agent
├── mcp/
│   └── server/              # MCP server implementation
├── pkg/
│   ├── config/              # Configuration management
│   ├── httpclient/          # HTTP client utilities
│   ├── llm/                 # LLM factory and adapters
│   ├── models/              # Data models
│   └── orchestration/       # Eino orchestration logic
├── main.go                  # CLI entry point
└── Makefile                 # Build and run targets

Documentation

  • README.md - Project overview and quick start
  • README_EINO.md - Eino orchestrator details
  • LLM_CONFIGURATION.md - LLM provider setup
  • MCP_SERVER.md - MCP integration guide

Requirements

  • Go 1.21 or higher
  • LLM API key (Gemini, Claude, OpenAI, or Ollama)

Quick Start

# Install dependencies
make install

# Set API key
export GOOGLE_API_KEY="your-api-key"

# Run with Eino orchestration (recommended)
make run-eino

# Or run all agents
make run-all

CI/CD

  • GitHub Actions workflows for build and lint
  • Dependabot configuration for dependency updates
  • golangci-lint integration for code quality

Contributors

  • John Wang (@grokify)

Repository: https://github.com/grokify/stats-agent-team