Built from the ground up for strict data isolation
When you connect your documentation to an AI platform, trust is paramount. Here is how Saathi Bot isolates, encrypts, and defends your data at every architectural layer.
Multi-Tenant Row-Level Security (RLS)
Every document chunk, vector embedding, and visitor conversation is stored in a multi-tenant PostgreSQL database protected by native Row-Level Security (RLS) policies. Queries enforce a mandatory bot_id partition filter before vector math occurs. Organization A can never query, infer, or leak Organization B's data.
Zero-Server Key Vault Guarantee
With our Bring-Your-Own-Key (BYOK) architecture, you maintain direct control over your AI model accounts. When you test in Playground mode, your private LLM API keys (Google Gemini, Groq, OpenAI, Claude, DeepSeek) are stored strictly in client-side device memory and are never written to our central PostgreSQL database.
Origin Whitelisting & CORS Defense
Embeddable widgets can be restricted to specific authorized domain hostnames and wildcard subdomains (*.yourcompany.com). Inbound chat requests verify HTTP headers and refuse to serve chat tokens to unauthorized origins, preventing widget hijacking and token theft.
Crawler SSRF Shielding
When our web ingestion engine indexes your public documentation, all URLs undergo pre-flight DNS resolution against our hardened blocklist. Any domain resolving to loopback addresses, internal RFC 1918 subnets, or cloud metadata endpoints (169.254.169.254) is dropped immediately before network calls begin.
Encryption & Transport Standards
In Transit
TLS 1.3 / HTTPS
All client-to-server traffic is forced over encrypted HTTPS. HTTP requests are redirected permanently (301).
At Rest
AES-256 Storage
Underlying PostgreSQL volumes and binary storage buckets are encrypted using industry-standard AES-256 block ciphers.
Authentication
Signed HS256 JWTs
Visitor sessions are cryptographically signed using HS256 secret keys, preventing session spoofing and unauthorized replay.
Responsible Vulnerability Disclosure
We welcome independent security researchers to inspect our platform. If you discover a security vulnerability or potential tenant isolation weakness, we commit to prompt verification, triage, and resolution.
Email our direct security triage inbox with proof-of-concept steps:
security@saathibot.com