Is Not Diamond Down?

If Prismix shows Not Diamond as "Operational" but you're having issues:

• Router defaulting to single model despite multi-model config (API key for secondary provider missing) — Not Diamond routes to alternative models only if valid API keys are configured for each provider. If the Anthropic or Gemini key is missing or expired in your Not Diamond settings, the router silently falls back to whichever provider has a valid key. Verify all provider API keys are present and valid in the Not Diamond dashboard under provider settings.
• Latency increase when preferred provider has degraded status — when a preferred model provider (e.g., OpenAI) is experiencing degraded performance, Not Diamond's router may continue sending traffic there before detecting the issue, increasing response latency. Enable automatic fallback in your routing configuration so that slow or error responses from a degraded provider trigger immediate rerouting to an available alternative.
• Routing decision not matching expected model (task signature changed) — Not Diamond's routing algorithm classifies requests by task type based on prompt structure and content signals. If you refactor your prompts significantly, the task classifier may re-categorize requests and route them to a different model than before. Review the routing decision logs in the Not Diamond dashboard to see which task class is being assigned to your requests.
• SDK not respecting max_cost budget per call — the max_cost budget parameter is a soft routing hint, not a hard cap. If all models within budget are unavailable or rate-limited, Not Diamond may select a more expensive model to fulfill the request. Monitor per-call cost via the dashboard and set hard budget limits at the account level if cost control is critical.
• Streaming incompatible with certain provider combinations — not all provider and model combinations in Not Diamond's routing pool support streaming responses. If streaming is enabled and the router selects a model that doesn't support it for your account tier, the response falls back to non-streaming delivery. Check which models in your routing pool have streaming enabled and explicitly include or exclude models that match your streaming requirements.
• Model preference cache stale (clear via dashboard) — Not Diamond caches routing preferences and performance metrics per model to optimize routing decisions over time. If a provider resolves an outage or you update your model preferences, the cache may still route to a previously favored (now suboptimal) model. Force a cache refresh from the Not Diamond dashboard to immediately apply updated routing preferences.