⟨u · v⟩ Multi-Tenant Edge Inference Gateway

DotProd AI Inference Platform

OpenAI SDK Compatible · Workers AI + Modal vLLM

High-precision AI inference wired directly into your harness of choice.

Issue tenant-isolated virtual API keys in seconds, stream tokens over standard OpenAI /v1/chat/completions endpoints, enforce atomic pre-inference credit holds on SQLite Durable Objects, and route seamlessly across Cloudflare Workers AI and dedicated Modal/vLLM GPU clusters.

Drop-In OpenAI SDK Integration

Zero client changes for Python, TypeScript, evaluation harnesses, or cURL.


              

Early Access · Kimi K3.1 Cohort Open

Proposed waitlist terms: $2.00 / 1M input · $0.25 / 1M cached · $4.00 / 1M output, or $600/mo unlimited. Reserve a spot →

01 · Atomic Ledger

Single-Threaded Credit Holds

Every tenant runs an isolated SQLite Durable Object that reserves conservative UTF-8 byte ceilings prior to upstream dispatch and settles exact cached vs. uncached tokens.

02 · Hybrid Compute

Workers AI + Modal vLLM

Instant global edge inference on Cloudflare Workers AI with pre-wired Modal Proxy Auth (Modal-Key / Modal-Secret) for dedicated vLLM clusters.

03 · Policy & Privacy

Envelope-Encrypted R2 Captures

Input/output policy screening with AES-256-GCM envelope-encrypted R2 storage and MFA-enforced, ticket-audited reviewer access.

04 · Prepaid Billing

Stripe Top-Ups & Reconciliation

Instant Stripe Checkout credit packages with webhook signature verification, payment deduplication, incremental refund deltas, and dispute protection.

Live Hosted Models & Metered Token Pricing

Prices in USD per 1,000,000 tokens. Prefix-cached input tokens are metered independently.

DotProd public model rate card
Model Identifier Uncached Input / 1M Cached Input / 1M Output / 1M

Save Your New DotProd Virtual API Key

This secret key is displayed only once. Copy it now or load it directly into the playground.


      

Revoke Virtual API Key?

Clients using this key will immediately receive HTTP 401 Unauthorized.