Konthora is an autonomous voice-driven enterprise operations engine built for the AssemblyAI Voice Agent Hackathon 2026. It turns real-time spoken prompts into production-ready, cryptographically verified business documents in sub-850 milliseconds. Powered by AssemblyAI's managed Voice Agent API, Konthora captures 24kHz PCM audio over a bidirectional WebSocket with sub-100ms VAD barge-in detection. It interprets natural multilingual speech (English, Bangla, and code-switched Banglish), executes 12 specialized API tools via a FastAPI backend, and renders 24 dynamic document types — including commercial invoices, purchase orders, HR offer letters, legal NDAs, inventory logs, and financial reports. Key Features: 1. 24 Enterprise Document Schemas: Instant generation with itemized line items, automated multi-currency conversions (USD, BDT, EUR, GBP, INR, JPY), and dynamic tax/discount calculations. 2. Real-Time Voice Revisions: Statefully revise any active document field or line item mid-conversation using speech commands. 3. Cryptographic Verification: Every generated document is sealed with a SHA-256 hash and a scannable QR audit payload. High-value transactions (>$10,000) trigger a CFO approval passkey gate. 4. Full-Duplex Architecture: Built-in Acoustic Echo Cancellation (AEC) and microphone muting prevent feedback loops, enabling seamless voice interruptions. 5. Edge Infrastructure: Next.js 14 frontend hosted on Vercel, FastAPI backend with SQLite on Railway, and local Kokoro-82M neural TTS feedback. Konthora eliminates manual data entry and multi-click dashboards, providing a zero-friction, voice-native interface for enterprise logistics, commercial finance, and HR management.
Category tags: