Guides
Practical how-to content for platform engineers and AI teams — routing strategy, compliance, budget management, evals, and self-hosted deployment.
Migrating from a direct OpenAI integration to a gateway
A step-by-step guide to moving your existing OpenAI API calls behind ManyLayers Gateway — with zero code changes for most applications, and a phased plan for larger migrations.
Rolling out per-team LLM budgets
Configure spend limits, alert thresholds, and hard caps per team using ManyLayers Gateway's budget engine — without touching application code.
Choosing a routing strategy
Latency-first, cost-first, or quality-first? A decision framework for configuring single, fallback, canary, and conditional routing across 40+ providers.
A practical webhook and alerting setup for budget overruns
Configure ManyLayers Gateway to send real-time budget alerts to Slack, PagerDuty, or a custom handler — so your team knows before a hard cap blocks production traffic.
Building a PII-safe prompting program
How to deploy ManyLayers Gateway's PII firewall across your organization — redaction modes, allow-list patterns, and audit logging without storing sensitive data.
Designing team and role structures for an AI platform
How to map your organization's teams and permission needs onto ManyLayers' team, role, and API key hierarchy — covering common enterprise org patterns and the access controls that serve them.
Connecting your knowledge sources
A practical overview of ManyLayers Workspace connectors — how to ingest documents, databases, and SaaS tools into your knowledge base for RAG-powered AI workflows.
Running evals your team will trust
Set up ELO-ranked model comparisons, golden datasets, and continuous eval scoring in ManyLayers Workspace — fed from real gateway traffic during canary rollouts.
Setting up canary rollouts for a new model version
A step-by-step guide to configuring weighted routing in ManyLayers Gateway to safely roll a new model version from 5% to 100% of production traffic.
Going air-gapped: a self-hosting checklist
Everything your team needs to deploy ManyLayers in a fully isolated environment — network requirements, license activation, model weight distribution, and operational readiness.