Skip to content
Glitch

Architecture

Glitch consists of two main components: the Platform (central management) and Sensors (edge enforcement). Together they provide comprehensive LLM security with minimal latency.

Overview

Section titled “Overview”
flowchart LR
    A[Your App] --> B[Sensor]
    B --> C[LLM Provider]
    B <--> D[Platform]
    E[Dashboard] --> D


    style A fill:#1a1a2e,stroke:#00d4ff,color:#fff
    style B fill:#1a1a2e,stroke:#00d4ff,color:#fff
    style C fill:#1a1a2e,stroke:#00d4ff,color:#fff
    style D fill:#0d3d4d,stroke:#00d4ff,color:#fff
    style E fill:#1a1a2e,stroke:#00d4ff,color:#fff
  1. You configure policies and projects in the Platform dashboard
  2. Sensors sync configuration automatically
  3. Sensors enforce policies on LLM requests
  4. Logs flow back to the Platform for analysis
  5. You monitor everything in the dashboard

Platform

Section titled “Platform”

The Platform is Glitch’s central control plane. It manages configuration, aggregates logs, provides analytics, and hosts the dashboard.

What the Platform Does

Section titled “What the Platform Does”

Configuration Management

Create and manage policies, projects, and organizations. Changes sync automatically to sensors.

Log Aggregation

Collect and store security events from all sensors. Search, filter, and export logs.

Analytics

Real-time dashboards showing detection rates, blocked requests, and usage trends.

Team Management

Invite team members, assign roles, and manage permissions across your organization.

Organization Structure

Section titled “Organization Structure”

Glitch organizes resources within organizations:

Organization
├── Members (users with roles)
├── Policies (security configurations)
├── Projects (API keys)
└── Logs (security events)

Roles

Section titled “Roles”
RolePermissions
OwnerFull access, billing, delete org
AdminManage resources, invite members
MemberCreate/read resources
ViewerRead-only access

Dashboard Features

Section titled “Dashboard Features”

Policy Management

Section titled “Policy Management”
  • Create and edit security policies
  • Configure detector thresholds per environment
  • Set up allow/deny lists

Project Management

Section titled “Project Management”
  • Create projects for different applications
  • Generate and rotate API keys
  • Assign policies to projects

Log Explorer

Section titled “Log Explorer”
  • Real-time log streaming
  • Full-text search across all events
  • Filter by detector, action, time range
  • Export to CSV/JSON

Analytics

Section titled “Analytics”
  • Request volume over time
  • Detection rates by category
  • Blocked vs. flagged breakdown
  • Top triggered detectors

Log Retention

Section titled “Log Retention”

Security events are stored with the following default retention:

PlanRetention
Free7 days
Pro30 days
EnterpriseCustom (up to unlimited)

Sensors

Section titled “Sensors”

Sensors are Glitch’s edge components that intercept LLM requests, run detection policies, and forward logs to the Platform.

What Sensors Do

Section titled “What Sensors Do”
  • Act as an OpenAI-compatible proxy between your app and the LLM
  • Execute detection policies with minimal latency
  • Forward security events to the Platform for analysis

Deployment Options

Section titled “Deployment Options”

Hosted Sensor

Use our managed sensor infrastructure for zero-ops deployment.

Available on all plans

Self-Hosted

Run sensors in your own infrastructure for lowest latency and data sovereignty.

Enterprise only — Contact sales

Getting Started

Section titled “Getting Started”

Point your application to the Glitch API instead of the LLM provider directly:

Terminal window
# Instead of:
# https://api.openai.com/v1/chat/completions


# Use:
https://api.golabrat.ai/v1/chat/completions

That’s it! Include your Glitch API key in requests and detection is handled automatically.

Request Flow

Section titled “Request Flow”
  1. Your application sends a request to the sensor
  2. Sensor validates your API key and loads the project’s policy
  3. Input detection runs on the request content
  4. If blocked → return 403 Forbidden immediately
  5. If passed → forward to LLM provider
  6. Output detection runs on the response
  7. If blocked → return 403 Forbidden
  8. If passed → return response to your application
  9. Log event sent to Platform asynchronously

Configuration Sync

Section titled “Configuration Sync”

Policy changes take effect automatically:

  • No restarts or redeployments needed
  • Changes propagate within seconds
  • Configuration cached for resilience

Performance

Section titled “Performance”
OperationLatency
Signature detection~11µs
Content moderation~50-100ms
Log forwardingAsync (0ms impact)

Self-Hosted Sensors

Section titled “Self-Hosted Sensors”

Benefits of self-hosted sensors:

  • Lowest latency — Run sensors in the same region as your application
  • Data sovereignty — Keep all data within your infrastructure
  • Custom scaling — Scale independently based on your traffic patterns

Next Steps

Section titled “Next Steps”