ClickLess AI
Travel conversational agent that replaces clicks with conversations
Overview
Travel planning is broken — you open 10 tabs, compare prices, check weather, read guides, and still feel uncertain. ClickLess AI replaces that entire workflow with a single conversation. Ask it "Plan a week in Tokyo in April" and it scrapes real-time flights, hotel prices, weather forecasts, and cultural tips, then synthesizes everything into a coherent trip brief.
The Problem
Travel planning is a fragmented, click-intensive process. Users manually navigate multiple platforms — comparing flights on one site, checking hotels on another, reading cultural guides, and consulting weather forecasts separately. Despite advances in conversational AI, no existing system automated this end-to-end research workflow from a single natural language utterance.
The Approach
ClickLess AI uses Playwright-based browser automation to scrape real-time data from flights, hotels, weather, and cultural sources. An LLM backbone handles multi-turn dialogue management, structured intent extraction, and response synthesis. The system uses a two-track data acquisition strategy: live scraping at query time and pre-cached knowledge bases for demo destinations.
The architecture includes a FastAPI backend with async parallel scraper orchestration, spaCy NER for deterministic entity detection (dates, currencies, locations), and LLM-based semantic extraction for unstructured content like cultural guides.
Key Decisions
Using Playwright over simple HTTP scraping was essential for handling JavaScript-rendered pages and bot detection layers on modern travel sites. The parallel scraping architecture minimizes end-to-end response time while maintaining stable element detection through explicit waits.
Comparing classical NLP baselines (spaCy + TF-IDF) against LLM-based extraction allowed us to identify where each approach excels — deterministic parsing for structured data, LLM for unstructured narrative content.
Impact
The system aggregates data from 5+ sources into a coherent trip brief in under 30 seconds. It eliminates the multi-tab, multi-click workflow that typically takes 30-60 minutes of manual research, reducing travel planning to a single conversational interaction.