I built a local-first intelligence analysis tool that computes probabilistic assessments from live public data sources (GDELT, Google News, Wikipedia,private web search).
Every prediction has traceable evidence chains. Sources are weighted by reliability tier (World Bank 0.93, state media 0.25). Probabilities shift as new evidence arrives -- Bayesian math, not vibes.
Ships with 5 demo assessments on Russia/Ukraine and a 1010-node knowledge graph. Click "Ingest Now" to pull fresh news and watch the numbers move.
docker compose up
http://localhost:8888
No cloud. No accounts. No telemetry.
Optional: run Ollama locally for AI-assisted analysis (summarize, challenge, what-if Questions).