5 pagesopens at /ai

AI

A product with a model in it: tokens and cost, two assistants, an image studio and the voice jobs.

/ai

Install

The whole dashboard — every page under one sidebar, the sample data they read — is one command. Or take a page at a time: each one brings the sidebar with it.

npx shadcn@latest add @vibra/ai

/aiAI cost and latency
npx shadcn@latest add @vibra/ai-overview
/ai/assistantAssistant
npx shadcn@latest add @vibra/ai-assistant
/ai/assistant-v2Assistant v2
npx shadcn@latest add @vibra/ai-assistant-v2
/ai/imagesImage studio
npx shadcn@latest add @vibra/ai-images
/ai/voiceVoice
npx shadcn@latest add @vibra/ai-voice

Widgets

Single cards cut from these pages, each installed on its own.

npx shadcn@latest add @vibra/widget-ai-overview-usage-summary
Open
From AI cost and latency
Usage headline
What the window cost across every provider, the requests and how many did not return 200, the tokens in and out, and the P95 latency beside the median; reads headline().
npx shadcn@latest add @vibra/widget-ai-overview-latency-chart
Open
From AI cost and latency
Latency
The median and 95th-percentile response time, a point a day over the thirty-day window; reads latencyByDay().
npx shadcn@latest add @vibra/widget-ai-overview-provider-mix-chart
Open
From AI cost and latency
Spend by provider
What each hosted-model provider billed over the window, in whole dollars, ranked with a bar each; reads providerMix().
npx shadcn@latest add @vibra/widget-ai-overview-status-mix
Open
From AI cost and latency
Status codes
How every request in the window ended — OK, rate limited or server error — as one bar with its counts and shares; reads statusMix().
npx shadcn@latest add @vibra/widget-ai-overview-traces-table
Open
From AI cost and latency
Traces
The newest twelve requests, the slowest twelve or every one that failed, and a panel on the one clicked with its tokens, cost and log record; reads tracesByView().