Skip to main content
TealTiger’s SDK-only architecture makes it naturally suited for edge deployment. Since all policy evaluation happens in-process with no external service calls, TealTiger adds minimal latency when deployed at the edge.

When to Use Edge Deployment

Edge deployment is ideal when:
  • Your LLM application runs on edge functions (Cloudflare Workers, Vercel Edge Functions, AWS Lambda@Edge)
  • You need sub-millisecond policy evaluation latency
  • You want guardrails to run before requests reach your origin server
  • You’re building a multi-region application and need policy enforcement in each region

Supported Edge Runtimes

Example: Cloudflare Worker

Performance

TealTiger’s guardrail evaluation typically completes in under 5ms, well within edge function time limits. The SDK has zero external dependencies for core guardrails (prompt injection, PII detection) — no network calls, no model inference.

Limitations

  • Content moderation via OpenAI Moderation API requires network access (opt-in, not default)
  • Edge runtimes may have memory limits — TealEngine’s policy cache should be sized accordingly
  • Behavioral policies (rate limiting, cost tracking) require state that doesn’t persist across edge invocations — use a KV store or Durable Objects for cross-request state