The ICCN AI Agent Implementation consists of two major sub-projects: an AI Research Agent and a Ramalan Agent (Gamification Data Collector), both designed to serve the needs of the ICCN organization.
The AI Research Agent leverages LangGraph's multi-step reasoning with the ReAct paradigm, integrated with Tavily for real-time web research. It automates the synthesis of comprehensive Markdown reports with citations, working across multi-model LLMs for optimal results in different reasoning tasks.
The Ramalan Agent is a gamified task-oriented agent designed for user profiling and data collection. It uses Redis for session management across multi-turn conversations and implements async background tasks for UI responsiveness, creating an engaging experience for users while gathering valuable data.
┌──────────────┐ ┌──────────────┐
│ PLANNER │────▶│ REASONER │◀─── ReAct LOOP
└──────────────┘ └──────┬───────┘
│
┌─────────────┼──────────────┐
▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌────────────┐
│ TAVILY │ │ ANALYZE │ │ EVALUATE │
└──────────┘ └──────────┘ └────────────┘
│
▼
┌──────────────┐
│ SYNTHESIZER │────▶ MD Report
└──────────────┘ReAct Reasoning Loop → Multi-Step Research → Report Synthesis