contextgem

Why we included this project

Most extraction tools either return free-form summaries or force you to hand-write prompts and parsing code for every document type. ContextGem takes a different route: you define 'aspects' for the specific sections you want pulled out, such as clauses or terms in a contract, and 'concepts' for the concrete fields you want inferred, like dates, numbers, or labels. These pieces chain into reusable pipelines, so the same extraction logic runs consistently across a whole batch of files, which matters for legal and financial work where the format varies but the fields you need do not. It works with cloud LLM providers and local models, and the project offers practical guidance on tuning accuracy against cost and latency. For engineers who want dependable typed output without reinventing extraction logic for each new document type, it is a solid fit.