BidAgent

Security and trust

Security controls belong in the platform and database—not only in the interface.

BidAgent security overview covering tenant isolation, forced PostgreSQL row-level security, upload quarantine, secrets, provenance, and human approval boundaries.

Tenant isolation

Application tables use forced PostgreSQL row-level security, request-scoped tenant context, and separate runtime roles for tenant-plane and platform operations.

  • Forced RLS
  • Dedicated database roles
  • Platform-route controls

Document safety

Uploads enter quarantine, file types are validated, and downstream processing is blocked until the configured malware-scanning workflow returns an acceptable verdict.

  • Quarantine bucket
  • Scan-verdict gate
  • Restricted document promotion

AI and human boundaries

Generated outputs are labeled and connected to source excerpts. Construction, legal, pricing, and final submission decisions require authorized human judgment.

  • Source provenance
  • Unverified-result labels
  • Review and approval records

Frequently asked questions

Is BidAgent production-ready?

The current public release is an indexable pre-beta. Production release still requires the remaining scanner, rate-limit, authentication-hardening, corpus, and live authorization gates documented by the project.

Where are database passwords stored?

Database credentials are stored in Cloudflare Hyperdrive configuration rather than application source or public configuration.