Tri2Direct AI Travel
Enterprise conversational AI hotel recommendation system built with a dual-layer LangGraph, Django Auth, FastAPI, and Groq (Llama 3).
Overview
Tri2Direct is a highly intelligent travel assistant designed to understand natural language queries, filter real-time database records, apply semantic vector ranking, and enrich results with live API data. The backend utilizes a dual-layer decoupled architecture: Django handles User Authentication/JWTs, while an asynchronous FastAPI server manages the high-speed conversational AI loop. The intelligence is driven by a 2-Tier LangGraph state machine featuring a Travel Orchestrator for fast classification and a Hotel Agent for multi-stage RAG pipelines. Features include direct PostgreSQL integration for zero-staleness queries, a hybrid SQL-Vector semantic search, Smart Session Pagination, and a Tavily Web Search fallback when the dataset lacks direct results.
What it does
- 0ms staleness · live DB
- Hybrid SQL + vector
- Smart session pagination