Your workspace, explained.

NEXORA uses the enabled GenMB AI service through window.genmb.ai.complete. The platform manages the supported model and server-side provider credentials. You do not need to obtain or paste an API key, and no secret is included in this frontend. Keep AI Text Generation enabled in the Add-ons panel.

The Free plan allows 100 AI requests per hour, shared by this app’s visitors. Pro allows 1,000 per hour. Each request can run for up to 60 seconds and generate up to 8,192 output tokens; NEXORA defaults to 2,048. Regenerating an answer uses another request. This enabled SDK returns a complete response rather than a streaming response, so NEXORA shows an honest thinking indicator—never simulated streaming. Stop waiting discards the pending answer locally; it does not cancel provider usage. Live web search is not available on this plan. Answers are AI-generated, not verified current information.

Conversations and preferences are saved in the app’s built-in key-value store after real awaited writes. Without an authentication capability, NEXORA uses a random browser workspace identifier and AES-GCM encryption before writing to shared storage. Your encryption key remains in localStorage on this device; plaintext conversations are never stored in shared keys. This is browser-local access, not account-level authorization: shared records can still be deleted by someone who discovers their keys, and malicious scripts in this browser could access the encryption key. Clearing browser data loses access to the saved workspace. There is no cross-device sync. Export important conversations. Your message, recent conversation context and attached text are sent to the AI service to generate answers. Do not enter secrets or highly sensitive information.

On GenMB, leave AI Text Generation and Key-Value Store enabled in Add-ons, preview the application, then publish it with the platform’s publishing controls. No custom server or API-key configuration is required for this integration. The supplied README has the project commands: npm install, npm run dev, and npm run build. Vite creates a dist folder and HashRouter makes deep links work without server rewrites. A standalone static host does not include GenMB’s injected services: use GenMB deployment for working AI and persistence. If you migrate to another host, implement your own server-side provider integration and storage adapter first; keep provider keys in that server’s environment, never the browser.

The built-in key-value store allows 600 reads and 180 writes per minute shared across this app, with a maximum value size of 100 KB. NEXORA encrypts and stores each conversation separately, loads history once, and saves at the beginning and end of a turn. A chat that exceeds the encrypted record limit must be exported and continued in a new conversation. Text attachments are processed locally and limited to 40 KB; PDFs, images and binary documents are not supported by this text-only AI service. If a request fails, its actual error appears in the chat and you can retry. A failed history read never prevents writing a new message. Check Add-ons and your plan’s usage if a rate-limit message appears.