
ZeroRoute
Architected an open-source, multi-cloud AI gateway that aggregates 8 free-tier LLM providers into a resilient, zero-downtime OpenAI-compatible API with sub-8ms auto-failover, exact caching, and an embeddable RAG chatbot widget.
About
ZeroRoute is a lightweight, zero-dependency multi-cloud AI gateway designed for solo founders, developers, and startups. It pools free-tier cloud quotas across 8 major AI providers (Groq, SambaNova, Mistral AI, Google Gemini, NVIDIA NIM, Cloudflare Workers AI, OpenRouter, and Cohere) into a single, high-availability, OpenAI-compatible endpoint. Features include millisecond-level circuit breaker failover on rate limits (HTTP 429 / 5xx), a self-healing background heartbeat, SHA-256 in-memory caching for zero-latency repeats, an AES-256-GCM encrypted credential vault, multi-tenant traffic analytics, and a 1-line drop-in embeddable chatbot widget (widget.js) with custom markdown RAG capabilities. It deploys seamlessly across Cloudflare Pages, Vercel Serverless, Render, Docker, and Node.js with universal persistence across Local Disk, Cloudflare KV, and Upstash Redis.
Key Features
- Multi-cloud routing aggregating 8 free-tier AI providers (Groq, SambaNova, Mistral, Gemini, NVIDIA NIM, OpenRouter, Cloudflare Workers AI, Cohere)
- Sub-8ms circuit breaker & automatic multi-hop failover on HTTP 429/5xx errors with self-healing background health checks
- SHA-256 in-memory exact prompt caching serving repeated queries at 0ms latency
- Drop-in OpenAI-compatible endpoints (/v1/chat/completions, /v1/models) supporting real-time Server-Sent Events (SSE) streaming
- Zero-dependency 1-line embeddable RAG chatbot widget (widget.js) supporting dynamic themes and custom Markdown knowledge bases
- Multi-tenant website traffic analytics and real-time provider quota/cooldown monitoring
- Enterprise AES-256-GCM credential encryption at rest with strict separation of ADMIN_KEY and ROUTER_API_KEY
- Zero runtime dependencies with universal edge/cloud deployment across Cloudflare Pages, Vercel, Render, Docker, and Node.js


