Use SDKs, OpenTelemetry, decorators, or REST APIs to instrument internal models, LLM gateways, and providers that do not have a native AnoSys integration.
Normalize model calls from hosted providers, internal models, and gateways into comparable records for latency, quality, cost, safety, and workflow context.
Prompt and completion metadata
Model parameters, latency, errors, and retries
Token usage, cost, and provider context
Quality, safety, relevance, and custom evals
Workflow, user, team, and business outcome context
Use the linked guide to install the collector or SDK, verify the endpoint, and send a small trace or event payload. Once data lands, connect it to dashboards, evals, alerts, and investigations.
@anosys_trace("custom-model-call")
def call_model(prompt):
return provider.generate(prompt)No. You can use OpenTelemetry, REST, SDK instrumentation, or decorators depending on your environment.
Yes. Payload capture can be controlled so sensitive prompts and completions are redacted or excluded.
Yes. AnoSys can capture token counts, provider cost, model cost, and custom cost attributes.
Yes. If your LLM workflow uses tools, functions, or external calls, AnoSys can capture them as spans or custom events.
Yes. Enterprise teams can apply governance, access, audit, and sensitive data controls.