Skip to content

Release Notes v0.2.0

Release Date: December 14, 2025

This is a major release that introduces the complete 4-agent architecture, Docker deployment, multi-LLM provider support, and real web search integration.

Highlights

  • 4-Agent Architecture - Complete multi-agent pipeline with Research, Synthesis, Verification, and Orchestration agents
  • Docker Deployment - Full containerization with docker-compose support
  • Multi-LLM Support - Unified interface for Gemini, Claude, OpenAI, xAI Grok, and Ollama
  • Real Web Search - Serper/SerpAPI integration for live search results

New Features

Synthesis Agent (agents/synthesis/)

New LLM-powered agent for extracting statistics from web content:

  • Fetches webpage content from URLs provided by Research agent
  • Uses LLM analysis to extract numerical statistics
  • Finds verbatim excerpts containing statistics
  • Creates structured CandidateStatistic objects with metadata
  • Runs on port 8004

4-Agent Architecture Documentation

Added comprehensive architecture documentation (4_AGENT_ARCHITECTURE.md) describing:

  • Agent responsibilities and communication patterns
  • Research → Synthesis → Verification pipeline
  • Orchestration strategies (ADK and Eino)
  • Port assignments and API contracts

Docker Support

Complete containerization for easy deployment:

  • Dockerfile - Multi-stage build for optimized images
  • docker-compose.yml - Full stack orchestration
  • docker-entrypoint.sh - Flexible agent startup script
  • DOCKER.md - Comprehensive Docker deployment guide
  • .dockerignore - Optimized build context

Serper/SerpAPI Integration

Real web search capabilities via the metasearch library:

  • pkg/search/service.go - Unified search service
  • Support for Serper and SerpAPI providers
  • Configurable result limits and filtering
  • SEARCH_INTEGRATION.md - Search provider documentation

Multi-LLM Provider Support

Unified LLM interface supporting multiple providers:

  • Gemini (Google) - Default provider
  • Claude (Anthropic) - Via gollm adapter
  • OpenAI - GPT-4 and GPT-3.5 support
  • xAI Grok - xAI's Grok models
  • Ollama - Local LLM deployment

New files: - pkg/llm/adapters/gollm_adapter.go - Multi-provider adapter - pkg/llm/factory.go - Enhanced model factory - MULTI_LLM_SUPPORT.md - Provider configuration guide - LLM_INTEGRATION.md - LLM integration documentation

Direct LLM Search Mode

Alternative search mode using LLM memory (for quick results without web verification):

  • pkg/direct/llm_search.go - Direct search implementation
  • --direct CLI flag for direct mode
  • Optional web verification of LLM-generated statistics

Base Agent Package

New shared agent utilities:

  • pkg/agent/base.go - Common agent functionality
  • Standardized health checks and configuration

Enhancements

  • Orchestration agent improvements - Better coordination logic and error handling
  • Research agent refactoring - Cleaner separation of concerns
  • Eino orchestration updates - Improved graph-based workflow
  • JSON extraction - Enhanced extractJSONFromMarkdown() function

Documentation

  • 4_AGENT_ARCHITECTURE.md - Complete architecture documentation
  • DOCKER.md - Docker deployment guide
  • LLM_INTEGRATION.md - LLM provider integration
  • MULTI_LLM_SUPPORT.md - Multi-provider configuration
  • SEARCH_INTEGRATION.md - Search provider setup
  • Updated README.md with new features and examples

Dependencies

  • Added github.com/grokify/gollm for multi-LLM support
  • Added github.com/grokify/metasearch for web search
  • Various go.mod updates for compatibility

Bug Fixes

  • Fixed search functionality (ad14475)
  • Fixed LLM support for synthesis and verification agents (99c9447)
  • Various golangci-lint fixes

Breaking Changes

  • Removed pre-built binaries from bin/ directory (now built from source)
  • Research agent API changes for 4-agent architecture compatibility

Migration Guide

From v0.1.0

  1. Update environment variables: 

    # Search provider (required for web search)
export SERPER_API_KEY="your-key"
# or
export SERPAPI_API_KEY="your-key"

  2. Docker deployment (recommended): 

    docker-compose up -d

  3. Local development: 

    make install
make run-all-eino  # Recommended
# or
make run-all       # ADK orchestration

LLM Provider Configuration

# Gemini (default)
export GOOGLE_API_KEY="your-key"

# Claude
export LLM_PROVIDER="claude"
export ANTHROPIC_API_KEY="your-key"

# OpenAI
export LLM_PROVIDER="openai"
export OPENAI_API_KEY="your-key"

# xAI Grok
export LLM_PROVIDER="xai"
export XAI_API_KEY="your-key"

# Ollama (local)
export LLM_PROVIDER="ollama"
export OLLAMA_URL="http://localhost:11434"

Contributors

  • John Wang (@grokify)

Full Changelog: v0.1.0...v0.2.0