
invest-ai, View code (opens in a new tab)
Investment analysis with a local LLM and figures you can trace
It assesses whether an asset looks like a good investment using live market data and a local LLM on Ollama. The code computes the figures and the model explains them. A verifier flags or drops any model figure that is not in the data, and every claim cites its source.
- 1,200+ tests: 931 backend and 286 frontend.
- It publishes its own backtest even though the result is unflattering: the price factors of its score showed no edge, and none of the 12 preregistered tests cleared the threshold.
- Chat with tool calling over live data, and hybrid RAG (FTS5 and embeddings).
- Private by construction: the model runs locally and the server only listens on loopback.
- Rust
- axum
- SQLite
- React
- TypeScript
- Vite
- Ollama
- RAG




