AI ENGINE

Two frontier models. One router. Zero guesswork.

Not every alert needs an expensive reasoning model. Our AI router evaluates task complexity and directs the workload to the right model family, Claude or GPT, balancing speed, cost and accuracy.

gauge-terminal / ai engine

System UI Render Pending

Task-complexity routing

Simple normalisation tasks route to fast, lightweight models. Complex correlation and triage route to frontier reasoning models automatically.

Enforced JSON output

No fragile text parsing. The engine enforces strict JSON schemas so AI outputs map cleanly to your dashboard and ticketing systems.

Tenant isolation

Data never bleeds between clients. Prompts are constructed strictly within the boundary of a single tenant, ensuring absolute privacy.

Confidence thresholds

The engine scores its own certainty. High-confidence alerts pass through; borderline cases queue for human approval tiers.

Full audit log

Every decision the AI makes is recorded. Review the exact prompt, the raw model response, and the routing logic at any time.

Pipeline Architecture

Alert In

Router

LLM

Model Tier

Confidence Gate

Auto Approval

Human Tier