How I designed a self-hosted reverse proxy that scans LLM traffic for PII, secrets, and prompt injection in under 2ms. Architecture decisions, bypass cases, and what I'd do differently.
Tamga (Go reverse proxy) scans LLM prompts in <2ms across OpenAI/Anthropic/Azure self-hosted; 62 adversarial test vectors, 47% still bypass policy. Self-hosted multi-provider avoids cloud gateway lock-in (Lakera, Portkey); performance <2ms enables GDPR/KVKK compliance enforcement for regulated environments.
TL;DR — I spent 6 months building Tamga,
an open-source reverse proxy that sits between your application and LLM
providers (OpenAI, Anthropic, Azure) and enforces a security policy on
every prompt in under 2ms. This post walks through the architecture
decisions, the 62 adversarial test vectors I built, where 29 of them
The Problem With the rise of ChatGPT and enterprise LLM integrations, a new attack vector...
How I designed a 5,000-line Node.js reverse proxy that captures traffic from Claude Code, Codex, DeepSeek, and 13 other providers…
A recall reference built from working through a 7-layer prompt-injection challenge. Focus: how each...
CostGuard's proxy endpoint makes an autonomous decision on every LLM call that passes through it. It...
I built a scanner that fires prompt-injection probes at a self-hosted AI agent and checks whether it...
Six weeks ago I shipped Lunaris Guard v0.1 — a dual-head classifier for prompt injection and content...
