Passer au contenu
DeepTokenSuite créative IA
AccueilTableau de bordModèlesInvitesTableau de classementDocumentationTarifsEntreprisesBlog
    Routing

    Routing & Fallback

    Multi-provider routing with health-aware fallback

    Define routing rules across providers and regions. Requests automatically fall over to healthy backends when upstreams degrade, time out, or rate limit.

    Configure RoutingView Pricing
    Reduced single-provider riskLatency-aware routingPer-model routing rules

    Aperçu des capacités

    multi

    providers per model

    live

    health-aware routing

    1 flow

    no client code change

    Avantages clés

    Points de valeur clés liés aux workflows de production pratiques.

    Avantage

    01

    Reduced single-provider risk

    When one upstream goes down, traffic shifts to fallback providers instead of impacting your end users.

    Avantage

    02

    Latency-aware routing

    Prefer providers with the lowest observed latency and success rate over a rolling window.

    Avantage

    03

    Per-model routing rules

    Tune routing priority, fallback chain, and timeout policy independently for each model.

    Workflow recommandé

    Un parcours simple en 3 étapes pour intégrer cette capacité aux appels de passerelle, politiques de routage et contrôles de facturation.

    Étape 1

    Define route

    Map each model id to a primary provider plus an ordered fallback chain.

    1

    Étape 2

    Set policy

    Pick timeouts, retry limits, and health thresholds that match your latency and reliability budget.

    2

    Étape 3

    Monitor and tune

    Inspect provider health and fallback rate in the dashboard and refine rules as traffic shifts.

    3

    FAQ

    What triggers a fallback?

    Upstream timeouts, 5xx responses, rate limit errors, or breaching the health threshold all trigger fallback to the next provider in the chain.

    Can I disable fallback for a specific request?

    Yes. Per-request and per-key policy overrides let you force a specific provider when needed.

    Étape suivante

    Prêt à mettre cette fonctionnalité en production ?

    Commencez depuis la documentation, le catalogue de modèles ou le tableau de bord et validez cette capacité avec du trafic de production réel.

    Configure RoutingRetour aux fonctionnalités