Self-improving document extraction
Our document extraction harness can now adapt to the documents you send it. On two public datasets, accuracy rose by about 7 to 10 percentage points.
The Uprelic Blog
Agent guard: rules in plain words for your AI agent
Check your agent's tool calls, their results and its replies against rules you write as sentences.
strom-base: a bigger model for the hard decisions
We added a second, larger model to Strom. Here is where it beats strom-fast, where it doesn't, and what it costs.
Strom's first 10 days
Users in 14 European countries, interest from three more regions, and use cases we did not expect.
Extracting document data using decision models
Decision models let you extract structured data from documents and images with low cost and latency. We discuss how we built our document extraction API using our decision model, Strom.
Analysing invoices using Strom
A fast and inexpensive approach to extracting structured data from invoice PDFs.
Prompt caching in Strom
A quick primer about how we use caching at Strom, and how to optimize your requests for maximum cache reuse.
Why we built Strom, an EU alternative decision model
Small decision models lose accuracy. Large ones need datacenter GPUs and heavy post-training. None of the open models matched Jev overall, and Jev can't see images. Our data had to stay in the EU. So we trained our own, and now it's an API.