What Was Shipped in v0.5.10 at a Glance
We eliminated tight ORM couplings to keep the core library 100% self-contained (dependencies: {}) while giving developers unlimited flexibility to store billing records anywhere.
1. Streamlined Database Sinks & Universal DIY Adapters
Earlier versions bundled heavy ORM schema definitions. In 0.5.10, we removed hard-coded Prisma and raw SQL couplings, establishing an unopinionated, pluggable sink model:
Pass your Supabase client directly. Automatically spools usage rows to vibez_usage or custom tables with optional balance verification.
vibezcheck.supabase(client)Connect Drizzle, Kysely, MongoDB, ClickHouse, or webhooks using a single asynchronous callback. Fully isolated so database errors never disrupt inference.
vibezcheck.database(async fn)Send usage data directly to Metronome’s /v1/ingest endpoint with zero third-party dependencies using native HTTP fetch.
vibezcheck.metronome({ apiKey })2. Serverless Lifecycle Spooling (Zero Dropped Events)
Serverless workers (Next.js App Router on Vercel, AWS Lambda, Cloudflare Workers) terminate immediately after streaming the final token. VibezCheck 0.5.10 guarantees zero dropped events:
Native Lifecycle Hooks (after & waitUntil)
VibezCheck automatically binds to globalThis.after() on Next.js/Vercel and globalThis.waitUntil() on Cloudflare Workers, keeping execution alive off the user-critical path until all queued events are persisted.
Explicit vibezcheck.flush() Guarantee
For standalone workers, microservices, or scheduled jobs, calling await vibezcheck.flush() drains all active batch queues synchronously before the process terminates.
3. Standardized <VibezCheck /> React HUD
The primary telemetry component is now cleanly named <VibezCheck /> (with <VibezCheckHUD /> preserved as an alias for complete backward compatibility):
import { VibezCheck, VibezReceipt } from 'vibezcheck/ui';
<VibezCheck messages={messages} />messages.$0.0028) and Tokens (1,420 tok).Bundled 700+ Models with Dynamic Sync
Every package release comes pre-bundled with exact rates for over 700 models across OpenAI, Anthropic, Google Gemini, DeepSeek, Meta Llama, Mistral, and AWS Bedrock. Plus, non-blocking background sync automatically ingests the latest rate cards from Vercel AI Gateway and Stripe Metronome.
| Model Identifier | Provider / Family | Input / 1M | Output / 1M | Latency / Overhead |
|---|---|---|---|---|
| gpt-4o-mini | OpenAI | $0.15 | $0.60 | 0ms (In-Memory) |
| claude-3-7-sonnet | Anthropic | $3.00 | $15.00 | 0ms (In-Memory) |
| gemini-2.0-flash | $0.10 | $0.40 | 0ms (In-Memory) | |
| deepseek-r1 | DeepSeek | $0.55 | $2.19 | 0ms (In-Memory) |
Ready to ship production AI billing?
Upgrade in seconds. Zero runtime dependencies added, all previous adapters and components remain fully functional with deprecation-free aliases.