portfolio-press · cockpit

Use case 1 — Trust-ladder publishing

LLM drafts a portfolio report → deterministic gate verifies every figure → you approve → it publishes. Live runs, real AWS.

Open pipeline cockpit

Use case 2 — Verified chart commentary

The engine computes volatility metrics, the LLM narrates the chart, the gate re-verifies every number. Regenerate: prose changes, figures don't.

Open risk demo

Use case 3 — Ask the data, in writing

A sentence compiles into a formal query from a whitelist. You see the compiled contract and approve it before anything runs; the engine executes it and the figures come back with the dataset they came from. The model never touches a number and never reaches the data.

Open the query console

Use case 4 — Ask the reports (agent)

A workflow runs a path we chose in advance. Here the model chooses the path: it picks which tools to call over the published reports, the pipeline journal and the real AWS bill. The route it took is shown, and the same gate still checks every number it quotes.

Open the agent