Data Flow

How data moves through the Xilos platform from input to output.

Input

Data enters Xilos when:

• An AI agent sends a chat completion or query request via the API
• A routine triggers a scheduled query
• A workflow stage processes its input

Processing

Each query flows through the processing pipeline:

Input → Authenticate → Guardrails → Restrictions → Cache Check
    → Route → Compress → LLM Call → Quality Score → Fact Extract
    → Log → Webhooks → Output

Storage

Data is stored in:

• Query Log — Full query history with routing decisions, governance actions, and responses
• Semantic Cache — Cached responses for similar future queries
• Context Engine — Extracted atomic facts for enriching future queries
• Audit Trail — Immutable record of all governance and security events

Output

Data leaves Xilos when:

• A response is returned to the requesting agent
• A webhook fires to notify an external system
• A SIEM export forwards governance events
• A report is generated and exported
• Training data is exported for fine-tuning

Info: All data storage is scoped to the organization. No data is shared between organizations.