Inferagate

Playbooks for operating Inferagate in production.

Guided operating paths for provider setup, model rollout, routing, guardrails, spend controls, observability, identity, and go-live readiness.

Playbook library

Everything a customer needs after landing in the product.

Open in console
01

Workspace activation

Activate a tenant, confirm the license tier, set the workspace owner, and verify the first operator can sign in.

  • License and edition check
  • Owner, admin, security, and viewer roles
  • MFA and SSO readiness
02

Provider connection

Connect managed APIs or self-hosted runtimes with the right endpoint, auth type, adapter, and readiness checks.

  • OpenAI, Anthropic, Bedrock, OpenRouter, Hugging Face, Groq
  • Ollama and vLLM local/self-hosted lanes
  • Credential status and quick test
03

Model catalog rollout

Scan provider models, choose approved IDs, classify capabilities, and expose only what each route should use.

  • Model family, capability, region, and price metadata
  • Primary, fallback, and restricted roles
  • Multi-model selection by family
04

Routing and promotion

Build route drafts, simulate candidate order, compare latency and cost behavior, then promote only after review.

  • Priority, weighted, latency, and cost-aware strategy
  • Fallback lane and timeout design
  • Preview before live promotion
05

Guardrails and policy

Create policy bundles that protect runtime traffic without burying operators in one-off rules.

  • PII, PHI, secrets, prompt injection, harmful content
  • Block, redact, warn, monitor, or approval flows
  • Diff review without exposing prompt content unnecessarily
06

Usage, tokens, and billing

Track platform usage, model consumption, token volume, budgets, plan limits, and customer pressure signals.

  • Tenant, route, model, key, and project-level usage
  • Budget alerts and usage export
  • No third-party infrastructure cost assumptions
07

Observability and integrations

Route metadata, health, policy outcomes, and audit events into the tools the customer already uses.

  • Splunk, OTLP, Datadog, Elastic, OpenSearch
  • Finance exports and customer success alerts
  • Prompt content never exported by default
08

Audit and access control

Manage operators, roles, activation, suspension, deletion, and review every sensitive administrative action.

  • Owner, admin, security, and viewer permissions
  • Role changes and account status lifecycle
  • Audit trail for customer operations
09

Go-live readiness

Confirm runtime health, model coverage, route readiness, keys, traffic evidence, and rollback posture before production use.

  • Gateway, database, cache, provider, and guardrail checks
  • Smoke tests and validation evidence
  • Rollback and incident response handoff

API quickstart stays here for builders.

Developers still get the minimum call shape when they need to send traffic through the gateway.

curl 
curl https://gateway.example.com/v1/chat/completions \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "chat",
    "messages": [{ "role": "user", "content": "Explain the rollout risk" }]
  }'