Token Management
API tokens are the primary authentication mechanism for the Infere platform. Each token is a secret key that carries its own configuration - model routing, a spending budget, rate limits, an IP allowlist, optional LLM security scanning, and an optionally deployed prompt.
Token Types
- Personal token - owned by you in your personal workspace.
- Organization token - owned by an organization and assignable to a specific member.
- Workspace default token - marked as the default for its workspace, used by the Playground and in-app tools.
Token Features
- Model routing - routing mode (balanced, cost-optimized, speed-optimized, quality-first, throughput-optimized), optional Auto-Routing, preferred/excluded/fallback models, and provider order.
- Token budgets - spending caps with daily, weekly, or monthly periods, a warning threshold (50–95%), and optional hard limits that block requests.
- Rate limits - requests per minute/hour/day, tokens per minute, and max concurrent requests.
- IP allowlisting - restrict a token to specific IPs or CIDR ranges.
- LLM security scanning - optional prompt-injection / unsafe-content scanning with a threat threshold and block-or-monitor mode.
- PII & compliance - redact or block PII, restrict routing to GDPR/HIPAA-compliant providers, and enforce data residency.
- Prompt deployment - deploy one prompt per token; its messages are prepended to every chat request automatically.
- Configurable expiration and instant revocation (revoke is reversible; delete is permanent).
Budget Configuration
{
"name": "production-api",
"budget": {
"enabled": true,
"limit": 250, // $250 spending cap
"period": "monthly",
"warningThreshold": 80, // flag at 80%
"hardLimit": true // block requests once reached
},
"rateLimits": {
"requestsPerMinute": 600,
"maxConcurrent": 50
}
}
Create tokens with the Quick Dev Token one-click path or the five-step Advanced Wizard (template → configure → routing → security → review). The token secret is shown only once at creation - store it securely.