Why local¶
One-liner: All data and features live on your machine. Switch the chat model and the table stays the same. Evidence is re-checkable, results are reproducible.
Wrong vs right¶
| Wrong | Right |
|---|---|
| “I asked a web AI; it said the golden cross is a classic trend.” | “The feature store and trades on my machine printed this table: golden-cross recent-window CAGR −3.1%.” |
A web AI gives you “a story from its training data” that you cannot re-check. Local gives you a table printed by “trades you downloaded + features you computed + a locked court”, which you can re-check line by line.
What “local” means¶
- Data: monthly parquet in
data/parquet_data/. - Features: monthly parquet in
feature_store/. - Court: locked scripts in
scripts/. - Results: KPI tables and logs in
results/.
Switch the AI model — as long as these files are there, the table stays the same.