Czech Intelligent Agent (CIA)
Team led by a CzechFabric AI Engineer who scaled public data AI (FastAPI, MCP, RAG) to the ChatGPT Store, plus a Project Manager.
YouTube Video
Project Description
Czech Intelligent Agent (CIA) is a real-time voice and text AI assistant that understands Czech public transport and air-quality conditions, performs tool-based reasoning through an MCP server, autonomously monitors delays using n8n workflows, and responds with natural ElevenLabs speech. The agent runs end-to-end: users speak or type, the system transcribes, reasons, calls real tools (departures, routing, air quality), and streams audio replies instantly.
Working Prototype:
CIA is fully functional: Chainlit chat, real-time WebSocket voice interaction, live MCP data, multilingual LLM reasoning, n8n automation, and ElevenLabs TTS all run as a complete system in Docker.
Technical Complexity & Integration:
The agent integrates multiple advanced components: OpenAI Whisper STT, ElevenLabs TTS streaming, CzechFabric MCP tools (transport, air quality, routing), a voice conversation loop with VAD and interrupt handling, n8n workflow automation, Redis session management, and a FastAPI backend orchestrating everything.
Innovation & Creativity:
CIA transforms public transport into a fully interactive voice agent. Instead of checking apps, users simply ask out loud and get immediate, real-time, tool-grounded answers. The design uses TONE serialization to drastically cut LLM token costs and supports autonomous monitoring workflows triggered by user commands.
Real-World Impact:
CIA removes friction in mobility, improves accessibility for visually-impaired users, supports bilingual Czech/English users, and allows autonomous delay notifications. The system can be deployed in trams, stations, or mobile phones.
Theme Alignment:
Built exactly around the hackathon theme: a multimodal conversational agent that speaks, listens, reasons, and uses tools. It showcases deep integration of ElevenLabs, OpenAI, n8n, and MCP APIs in a unified smart agent.
Technologies Used:
FastAPI-swx_api, Chainlit, Docker, Redis, PostgreSQL, CzechFabric MCP, ElevenLabs TTS, OpenAI Whisper STT, GPT-4o-mini, n8n workflows, WebSockets, and TONE serialization.
Prior Work
Parts of the MCP transport tools were originally developed in the CzechFabric SDK prior to the hackathon. The hackathon work consists of integrating these tools into a real conversational agent, implementing real-time audio streaming, the entire voice pipeline, the LLM reasoning loop, ElevenLabs TTS, n8n automation, Chainlit UI integration, all backend orchestration, and the new “CIA” agent experience.
Team
Products & Tools
Additional Links
Presentation Slides