Search engines are only as good as the data you feed them. For the Algolia Agent Studio Challenge, I wanted to build more than just a chatbot; I wanted to build a living pipeline.
Introducing CineAgent, a voice-powered cinema concierge grounded by Trawl—a custom-built, high-concurrency ETL engine written in Go.
🚀 The Vision: Beyond Static Data
Most AI agents are limited by their training data or static JSON uploads. CineAgent is different. It uses a "Trawler" approach to bridge the gap between unstructured web content and conversational discovery.
The project solves three specific problems:
- Data Freshness: Trawl (the backend) can scrape and index thousands of records concurrently.
- Grounding: The AI Agent is strictly grounded in a curated index of 250+ cinematic masterpieces.
- The "Empty State" Problem: In standard search, "No Results" is a dead end. In CineAgent, "No Results" is the beginning of a conversation.
🛠 The Tech Stack
- The Engine (Trawl): Built with Go, utilizing Goroutines and Channels for high-concurrency web scraping and batch indexing via the Algolia Go API Client v5.
- The Brain: Algolia Agent Studio, configured as a "Film Discovery Expert" to handle complex, natural language queries.
- The Face: A React frontend styled with Chakra UI v3 for a premium, immersive dark-mode experience.
- The Interaction: Web Speech API for hands-free discovery and the Algolia InstantSearch Chat Widget (Beta).
💡 Key Features & Originality
1. The Trawl ETL Pipeline (Go)
While most participants might upload a static file, I built an ETL (Extract, Transform, Load) engine. Trawl follows a concurrent worker-pool pattern. It scrapes metadata (Title, Year, Rating, High-res posters), normalizes it into Go structs, and pipes it into Algolia in optimized batches.
2. The Hybrid "Voice-to-Agent" UI
I implemented a "NoResultsBoundary" logic. If the local search grid can't find a specific niche movie, the UI doesn't just show an error. It proactively invites the user to "Ask CineAgent AI," handing off the search query to the conversational agent.
3. Voice-First Discovery
Users can click the microphone icon to search. If the voice input is a short keyword (e.g., "Inception"), the movie grid updates. If the input is a complex question (e.g., "What is the best sci-fi movie from the 90s?"), CineAgent opens the conversational chat automatically.
📺 Demo & Visuals
Live URL: trawl-ui.vercel.app
![Image Description: A sleek dark-mode movie library with a floating AI Assistant bubble.]
🧠 Lessons Learned
Integrating a Beta widget like the Algolia Chat component into a modern React 18 / Chakra UI v3 environment was a challenge in state management. I learned that isolating the AI handshake from the main search lifecycle is key to maintaining a responsive, "hang-free" user experience.
🔗 Links
- GitHub Repository:
Frontend (UI)
github.com/ntekim/trawl-ui
Backend
github.com/ntekim/trawl
-
Algolia Agent ID:
4ddf0636-38b2-4578-a98f-8715efde81ca
Top comments (0)