Retrivora AI RAG Engine — Start building today!

Trust & Security Control

Security Architecture & Safeguards

System Security Verified & Audited: July 2026

Zero Raw Secret Exposure

Retrivora eliminates developer key leakage. End-user applications use only a cryptographic RETRIVORA_LICENSE_KEY. Upstream LLM provider keys remain vaulted server-side.

Multi-Tenant Vector Isolation

Every workspace gets a dedicated, isolated Pinecone vector database namespace (retrivora-<projectId>). Cross-tenant queries are blocked at the router layer.

High-Availability HA Failover

The llm.retrivora.com gateway automatically reroutes queries on upstream 429 rate limit triggers (Groq $\rightarrow$ Google Gemini $\rightarrow$ Hugging Face) to maintain 99.9% uptime.

Adaptive Quota & Rate Defense

Per-project RPM rate-limiting and daily token quotas prevent DDoS attempts and unexpected billing surges.

1. Cryptographic License Verification

When an end-user application makes a streaming chat or embedding request to Retrivora, the request includes a signed JWT license key in the Authorization header (Bearer LK_FREE_...).

The gateway verifies the RSA key signature in-memory in under 5 milliseconds, extracting the verified projectId, user email, and tier entitlements before forwarding payloads.

2. Responsible Vulnerability Disclosure

We welcome security researchers and community members to report potential vulnerabilities. If you discover a vulnerability or security issue, please notify our team responsibly:

Security Contact & PGP Key

Email security issues to security@retrivora.com. Please include reproduction steps, HTTP logs, and target endpoints. We respond within 24 hours.