Voistra
Team led by a Langdemy CEO (GCP, AI, Web3 Full Stack) and an Asahi Shimbun Frontend Lead (NextJS).
Project Description
Voistra（ヴォイスラ）は、声ひとつであらゆるツール・ワークフロー・情報をオーケストレーションする次世代AIアシスタントです。
話しかけるだけで、検索、分析、ブラウザ操作、コード生成、PRレビュー、ワークフロー実行まで、
すべてが自動で動き出します。
あなたの声が、あなたのデジタル世界の指揮者に。
Voistra is a voice-first autonomous agent framework that coordinates browser automation, workflow engines, code tools, and cloud services through natural speech.
Powered by real-time speech AI, LLM reasoning, and tool orchestration,
Voistra turns verbal intent into structured actions — instantly.
Use your voice to build. Deploy. Automate. Review. Navigate.
Voistra handles the details.
Prior Work
Our team had no fully working agent system prior to the hackathon.
We only brought basic conceptual sketches and isolated experiments.
All core functionality showcased in the submission was built entirely during the hackathon.
Prior Work (Before the Hackathon)
We had the following lightweight, non-functional materials prepared in advance:
High-level architectural notes
Rough ideas on combining voice interfaces with LLM tool orchestration.
No working code or integrated components.
Small isolated code experiments
A simple Next.js boilerplate (no agent logic, no APIs).
Basic tests of ElevenLabs TTS (non-realtime, single API call).
No realtime WS, no agent routing, no automation logic.
Design mockups / sketches
Non-interactive UI sketches for a “voice console” interface.
None of these components were connected, production-ready, or functional before the event.