Finsights

FinSights — from company facts to explainable exposure maps, powered by Prometheux.

A side project (braindb) ingests news/filings into verified company entities + facts. One SQL query exports them; an LLM classifies each co-mention into a clean relation (supplies / invests_in / partners_with / competes_with); then Prometheux does what SQL and LLMs can't — recursive, explainable reasoning that derives the indirect connections and risk exposure across the network.

It discovers, with a proof chain, that a West-Texas gas field and a value stock (Smart Sand) are both exposed to the AI-datacenter capex boom three hops away. A simple HTML/JS dashboard shows it live: direct relations, Prometheux-derived indirect links, AI-exposure highlighting, PageRank centrality.

I chose that as I am very much interested into investing, agents, and the management of their memory and context. I learned about the usability of graph databases and rules, and what value tools like Prometheux can add

Stack: braindb (memory/extraction) · LLM (fact→relation) · Prometheux (Vadalog reasoning) · vanilla JS

Built With

• braindb
• claudecode
• prometheux
• python