fpa-portfolio-learnlisted

# Portfolio Learn (Loop B) ## Overview Loop A makes the model better at one client. This makes your *practice* compound: client #10 starts smarter than client #1 because your library carries what generalized across #1–9. Everything is local - your own book, on your own machine. **Core principle:** self-improving, never self-ratifying - propose, you accept. The objective metric is cross-client: does a pattern learned on some clients fail to degrade the *others*' backtest? ## Setup A portfolio manifest `~/.fpa/portfolio.yaml` lists your clients + a business-type tag: ```yaml library: ~/.fpa/library clients: - { path: ~/clients/acme, type: d2c-inventory } - { path: ~/clients/peak, type: d2c-inventory } - { path: ~/clients/haul, type: trucking } ``` ## Workflow 1. **Load** the manifest (`pyfpa.load_portfolio`). 2. For each business-type with at least 3 clients: - **Priors:** let `type_clients = pyfpa.portfolio.clients_of_type(portfolio, type)`. `pyfpa.mine_priors(portfolio, type)` finds drivers that cluster tightly; validate each with `pyfpa.validate_prior(driver, type_clients)` (leave-one-out). Surface validated ones first (by cross-client delta), then unvalidated/judgment. - **Skills:** `pyfpa.find_recurring_skills(portfolio, type)` for recurring generated skills. Also weigh recurring **structural corrections** across clients (read each `.fpa/corrections/` for `type: structural`) - a human-authored pattern that repeats is str