AI Kernel

The AI Kernel is the core processing engine that orchestrates the full pipeline from query to response.

What the AI Kernel Does

The AI Kernel coordinates all processing steps for each query:

• Routing — Determines which LLM handles the query, using routing rules or Smart Routing
• Workflows — Executes multi-model workflows when a routing rule references a workflow
• Quality Scoring — Computes faithfulness, answer relevance, and context relevance scores for each response
• Fact Extraction — Extracts atomic facts from responses and stores them in the Context Engine
• Webhook Dispatch — Fires event-driven webhooks when configured

How It Works

When a query arrives, the AI Kernel:

1. Checks the semantic cache for a match
2. Evaluates restriction rules and guardrails
3. Selects the target model via routing rules or Smart Routing
4. Applies context compression if enabled
5. Sends the query to the LLM provider
6. Receives the response
7. Scores the response quality
8. Extracts and stores atomic facts
9. Logs the full pipeline to the query log
10. Fires webhooks for subscribed events

The AI Kernel handles all of this transparently — users and agents simply send a request and receive a response.