Spead AI — Enterprise AI Assistant Platform
Overview
Spead AI is the web client of Caliana's enterprise AI assistant platform, designed for Indonesian-first organizations. Built with Next.js 14 (Pages Router) and React 18, it ships three core modules: AI Partner, a general-purpose chat assistant with streaming responses and file attachment; Doc Assistant, a retrieval-augmented (RAG) assistant that answers questions against an uploaded company knowledge base with document/tag filters and cited sources; and Doc Builder, which generates professional documents from templates and prompts. The frontend streams model responses over Server-Sent Events, parses source citations, and tracks asynchronous document vectorization through Socket.IO. It is built on shadcn/ui and assistant-ui chat primitives with a unified custom chat runtime shared across modules, TanStack Query for server state, and a full bilingual (Indonesian/English) interface with dark mode. Beyond the AI experience, it includes a management suite for knowledge base documents, generation templates, departments, roles, and user accounts, with department- and role-based access to document labels. Access is protected by JWT sessions with edge middleware, super-admin gating, and RSA + AES encryption of sensitive payloads, while consumption is metered through an AI token quota with real-time warnings.
Key Capabilities
- AI Partner — conversational assistant with streaming responses and file attachments
- Doc Assistant — RAG chat over a company knowledge base with cited sources
- Doc Builder — AI document generation from templates and prompts
- Knowledge Base management with document upload, labeling, approval, and revectorization
- Asynchronous document vectorization with real-time Socket.IO status
- Document/tag filtering for targeted retrieval
- Template, department, role, and account management
- Department- and role-based access control over document labels
- AI token quota metering with real-time usage warnings
- Bilingual UI (Indonesian/English) with dark mode