Blog
Updates, announcements, and engineering insights
Nano Banana 2 Is Here: Reasoning-Guided Image Generation, Now on Cencori
Google's Nano Banana 2 (Gemini 3.1 Flash Image) just dropped — blazing-fast, 4K-capable image generation built on reasoning, not diffusion. It's already live on Cencori.
Introducing Gemini 3.1 Pro: Enhanced Reasoning & Agentic Workflows
Google just released Gemini 3.1 Pro and it's already live on Cencori. Experience 1M context, elite reasoning, and optimized agentic tool use.
Run OpenClaw With Cencori
OpenClaw now works with Cencori using the standard OpenAI-compatible setup: base URL + API key. No custom header required.
Moving From Database to Redis
As AI traffic grows, database-backed rate limiting becomes a bottleneck. Here's how we migrated to Redis for sub-millisecond enforcement.
Semantic Caching
LLM APIs are expensive and slow. We implemented a semantic caching layer in the Cencori AI Gateway to make repeatable requests instant and free.
Tokenization: How We Let AI Use Context Without Seeing Your Data
Masking PII breaks AI context. Blocking it kills the feature entirely. We built a third option: tokenize sensitive data into named placeholders, let the LLM reason about them, then restore the real values only for the end user.
Cencori + Vercel AI SDK
The AI Infrastructure
SDK v1.0.2: Multi-Provider Image Generation & AI SDK Parity
Generate images with GPT Image 1.5 and Gemini, use function calling, get structured output, and analyze images - all through one unified API.
Building a Production-Ready AI Security Layer: How We Catch What Your LLM Shouldn't Say
When everyone's focused on what goes into an AI, we learned the hard way that what comes out can be just as dangerous. Here's how we built an output scanner that catches PII leakage and instruction attacks.
Cencori + TanStack AI
Cencori is now an official adapter for TanStack AI. Use any AI provider with built-in security and observability.
Introducing Cencori
Today, we're excited to introduce Cencori — the AI infrastructure platform for production applications.