我正建構 Unified AI Gateway，免再切換不同 LLM API key
Stop juggling LLM API keys: I’m building a Unified AI Gateway 🚀
作者正驗證 Unified LLM API Gateway 的構想，計劃以單一代理連接 40+ AI providers 和 250+ models。閘道兼容 OpenAI SDK，並提供備援路由及統一錢包；作者正徵詢社羣意見，尚未決定是否投入建構 MVP。
Hey everyone 👋,
If you’re building AI apps today, you probably know the headache of API fragmentation. You might want to use GPT-4o for reasoning, Claude 3.5 for coding, and Llama 3 for fast, cheap processing.
But doing this currently means:
❌ Managing 5+ different provider accounts.
❌ Rewriting code to match different SDKs.
❌ Dealing with separate billing, rate limits, and declined cards.
❌ App downtime when a provider's server crashes.
I am validating an idea for a Unified LLM API Gateway to solve this. It acts as a single proxy between your code and 40+ AI providers (250+ models).
How it works:
Drop-in Replacement: 100% compatible with the OpenAI SDK. Just change the baseURL and API key. Zero code rewrites.
Smart Fallback: If a provider has an outage or hits a rate limit, it instantly routes your request to a backup model. 99.99% uptime.
Unified Wallet: Top up once, pay for all models.
Before I fully commit to building the MVP, I want to hear from the community:
Is this a pain point you currently face? Would a unified gateway save you engineering hours?
Let me know your thoughts or feature requests in the comments 👇
來源：DEV Community · dev.to