Apps & data pipelines
Build a small knowledge graph without embeddings: Jev checks mentions and links, retrieval follows the graph, and Claude Haiku writes a cited answer.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
The proof of concept uses 331 author-written chunks about three fiction franchises. A local interface can replay saved results. Questions and text go to TypeSafe and Bedrock; deployment and calls cost money. It is explicitly not intended for production.
Use this idea to help people find answers in a small collection of text. Ask a developer to build a map that connects names in the text. Jev checks those connections. The program follows the map to find evidence, then another AI uses that evidence to write an answer.
The local page can replay saved results without new calls. Running the full example uses Amazon’s cloud and answer-writing service, plus TypeSafe’s Jev service. These can cost money and send text outside your system. The small fiction sample is not a broad search test, and the maker explicitly rules out production use.