Live on RapidAPI

Sentinel

A dedicated supervision layer for LLM applications. Sentinel acts as a firewall for AI, ensuring responses meet safety, compliance, and quality standards before they reach your users.

Get API Key View on GitHub
Live Interactive Demo

Sentinel AI Supervisor

Test the supervision engine. Enter a user prompt and a simulated LLM response to see how Sentinel judges compliance.

* Connects to Live Production API (GCP)
Open full dashboard →

The Architecture of Trust

Deploying LLMs in production carries risk. "Jailbreaks," hallucinations, and compliance violations can damage brand reputation. Sentinel mitigates this by introducing a human-in-the-loop inspired automated workflow.

System Flow

  1. 1
    Request Interception Your application sends the user prompt and the LLM's draft response to Sentinel.
  2. 2
    Policy Evaluation Sentinel runs specialized "Supervisor Agents" (e.g., Healthcare Compliance, PII Scrubber) against the draft.
  3. 3
    Verdict & Action Returns a PASS, FAIL, or FIX verdict. If 'FIX', it provides specific instructions for regeneration.

Tech Stack

  • FastAPI (Async Python)
  • Google Cloud Run (Serverless)
  • RapidAPI (Gateway & Billing)
  • Pydantic (Validation)

Monetization & API Design

Sentinel is designed as a SaaS product. It leverages a Freemium model via RapidAPI.

  • Hard Limits: Free tier users are capped to prevent cloud cost overruns.
  • Security: Direct access to the backend is blocked via a X-RapidAPI-Proxy-Secret handshake, ensuring all traffic must pass through the monetization gateway.
  • Scalability: Deployed on Cloud Run, it scales to zero when unused and auto-scales to handle traffic spikes.

Production Integration

To use Sentinel in your own production environment, subscribe via RapidAPI to get your credentialed endpoint.

cURL Example (RapidAPI) Production 
curl --request POST \
  --url https://sentinel-ai.p.rapidapi.com/supervise \
  --header 'Content-Type: application/json' \
  --header 'x-rapidapi-host: sentinel-ai.p.rapidapi.com' \
  --header 'x-rapidapi-key: YOUR_RAPIDAPI_KEY' \
  --data '{
    "prompt": "How do I secure an API?",
    "draft": "Use a proxy layer and validate secrets..."
  }'