Overview
Retry failed calls on a secondary provider (e.g., OpenAI → Anthropic on transient 5xx).
Problem
Single-provider dependency causes outages and SLA breaches.
Solution
AI Hub routes a one-time fallback retry on eligible errors and records the final provider in usage logs.
How it works
Store credentials for multiple providers. If the primary returns transient 5xx, AI Hub tries the secondary and returns the successful content with accurate cost attribution.
Who is this for
SRE / Ops
Engineering
Product
Expected outcomes
- Higher uptime for AI-backed workflows
- Transparent cost attribution during fallbacks
Key metrics
AI-related incidents
Baseline
6 count/qtr
Target
1 count/qtr
Successful fallbacks
Baseline
0 % of eligible
Target
95 % of eligible
Gallery
Case studies
E-commerce app rides out LLM outage
Fallback kept PDP generation online during a 45-min incident.
Retail Mid-market LATAM
Security impact
- Prompts/outputs; audit of provider switches · PII: none
Compliance
- SOC2
- Change management logging
Availability & next steps
Pro
Enterprise