Risk Dimensions sends your database login (role + password) to you privately. It's read-only on the firm's data; Contributors also get append rights to shared tables on the same login. Keep it private — never commit it or paste it into a notebook cell.
The database is reached through a secure Cloudflare tunnel. Install the small client once:
winget install --id Cloudflare.cloudflared
(macOS: brew install cloudflared.)
In a terminal, start the tunnel and leave the window open while you work:
cloudflared access tcp --hostname db.riskdimensions.io --url localhost:5432
Your laptop's localhost:5432 is now wired to the database.
Open rd_sandbox_starter.ipynb in Jupyter or VS Code (the rd_metrics.py calc library sits beside it and just imports). In Step 1, set the host to localhost.
Run top to bottom. Step 1 prompts for your credential (via getpass — nothing is saved to the file). Then: pull candidates → compute with rd_metrics → rank & comment with your own LLM → (Contributor) push your picks → read the feedback.
rd_metrics.py to the session (or set the download URL in Step 0b). The write step always needs the database credential.
getpass only — never hard-code a key in a cell or commit one. Your access has a set term and renews with the program; if a credential is ever exposed, tell us and we revoke it instantly.