OpenAI-Compatible AI Governance Gateway
Drop-in integration
SAIG exposes an OpenAI-compatible /v1/chat/completions endpoint. Any application using the OpenAI SDK or API can integrate by changing only the base URL and API key — no SDK changes, no code rewrite, no agent installation.
What you get
- PII detection and reversible anonymization on every request
- Deterministic policy enforcement before the request reaches the LLM
- Provider routing across OpenAI, Azure OpenAI, Anthropic Claude, and Google Gemini
- Signed audit evidence with SHA-256 hash chain and Ed25519 signatures
- Governance metadata in X-SAIG-* response headers
Model aliases
Decouple your application from provider-specific model names. Use saig-default, saig-fast, saig-sensitive-data, or saig-low-cost. Switch providers without touching application code.
Quick start
from openai import OpenAI
client = OpenAI(
base_url="https://api.your-domain.eu/v1",
api_key="your-saig-api-key",
)
response = client.chat.completions.create(
model="saig-default",
messages=[{"role": "user", "content": "Summarize this report."}],
)
print(response.choices[0].message.content)Works with Python, TypeScript, Java, Go, cURL, and any OpenAI-compatible client.
Frequently Asked Questions
Do I need to change my code?
For OpenAI-compatible clients, change only the base URL and API key. No SDK changes, no code rewrite.
Which providers does SAIG support?
SAIG routes to OpenAI, Azure OpenAI, Anthropic Claude, and Google Gemini through a unified API.
What are model aliases?
Model aliases like saig-default and saig-sensitive-data decouple your application from provider model names. Switch providers without code changes.
Does SAIG add latency?
SAIG adds governance processing time for PII detection, policy evaluation, and audit record creation. Observe mode lets you measure this before enforcing.