Category: WunderGraph
-
RAG Cost Optimization
The article discusses how costs associated with Retrieval-Augmented Generation (RAG) systems can spiral due to fragmented control and local decision-making in AI workflows. It emphasizes the need for a centralized control layer to manage resource allocation, caching, retrieval depth, and model calls, ultimately aiming to reduce unpredictability in AI spending.
-
The State of Federation 2026 Survey Is Open
Originally posted on the WunderGraph Blog We’re running a survey to understand how organizations are adopting, operating, and scaling federated GraphQL. The goal is straightforward: build a vendor-neutral picture of what Federation looks like in production today, and why some teams haven’t adopted it yet. The required questions take 5–10 minutes. Optional sections cover governance…
-
The Hidden Cost of Non-Compliance in AI
Originally posted on the WunderGraph Blog Executive Framing Every AI deployment now carries audit risk. For high‑risk AI in Europe, logging and documentation are written into law. In the United States, states are building their own playbooks with no consistency. In parts of Asia, certain high‑impact AI systems increasingly face mandatory risk assessments and, in…
-
What a Year at WunderGraph Looks Like
Originally published on the WunderGraph Blog After a year at WunderGraph, one thing stood out. The speed is real, but so is the structure behind it. This is what that combination feels like day-to-day, and why the past twelve months feel fuller than the calendar suggests. We’ve been growing rapidly. We’re 3x bigger than we were a…
-
From 10+ Seconds to Under One: Solving Their Slowest Operation
Originally posted on the WunderGraph Blog TL;DR: A growing personal finance marketplace hit severe planning delays on its Apollo Gateway, with complex queries taking more than 10 seconds and blocking traffic. After migrating to WunderGraph Cosmo and enabling the Cache Warmer, planning latency dropped by roughly 90–95 percent, and previously slow operations became 12–18 times faster. The team was able to…