What Is Web Retrieval and Why It's Optional
Web retrieval allows AI to augment answers with current information from the internet—such as recent news, technical documentation, or public data sources.
This is entirely optional. By default, HybridLLM operates using only your uploaded knowledge bases and Swiss-hosted open-source models. Web retrieval must be explicitly enabled per workspace.
When enabled, web retrieval routes requests through Swiss infrastructure with IP anonymization to protect your organization's identity.
How Anonymization Works
When web retrieval is enabled, all outbound requests are routed through Swiss proxy servers that strip identifying information:
Your organization's IP address is replaced with a Swiss proxy IP.
No user identifiers, session cookies, or organizational metadata are included in web requests.
The external website sees only a generic request from a Swiss IP, with no way to trace it back to your organization or user.
This protects your competitive intelligence, client information, and research activities from external tracking.
Workspace-Level Policy Controls
Admins can configure web retrieval policies at the workspace level:
Enable or disable web retrieval entirely for specific workspaces (e.g., allow for research team, disable for client-facing legal work).
Restrict which users can trigger web retrieval (Owner and Admin roles only, or extend to all Members).
Set audit log retention for web retrieval actions (recommended: 90 days minimum).
Review web retrieval usage through audit logs to ensure compliance with your internal policies.
Audit Trails and Compliance
Every web retrieval action is logged with:
User ID and role (Owner, Admin, or Member).
Workspace ID and name.
Search query or URL requested.
Timestamp and anonymized proxy IP used.
AI response based on retrieved content.
These logs are immutable and can be exported for compliance audits or GDPR data subject access requests.
When to Enable Web Retrieval
Enable web retrieval for: Research projects requiring current data, technical support needing documentation lookups, or market intelligence teams analyzing public information.
Disable web retrieval for: Client-facing work with confidential data, legal matters protected by privilege, or healthcare scenarios with patient information.
Best practice: Create separate workspaces for different use cases, enabling web retrieval only where appropriate.