sail-inference-with-voyage

To trace LLM inference calls, you attribute Sail inference model requests to the correct Voyage execution context and dashboard trace by applying request-time header propagation and agent scoping. This ensures every model call appears under the correct execution step for observability.

Inference attribution for background agent workflows is the process of attaching Voyage, span, and agent context to Sail inference requests. It solves the problem of untracked or incorrectly attributed model calls by connecting them to the correct dashboard trace.

Yes, you can use OpenAI-compatible clients with Voyage tracing. The Skill helps wrap compatible OpenAI clients and handles background versus synchronous inference behavior to avoid response attribution mistakes during dashboard observability.

Sail inference model calls may not show up in the correct dashboard trace due to missing request-time Voyage header propagation, incorrect span and agent scoping, or improper response association handling. Reliable observability requires applying all three context attribution mechanisms.

Yes, you need request-time Voyage header propagation to debug model-call attribution. Along with span and agent scoping and response association handling, it is a required mechanism for connecting background LLM calls to the correct execution context.

The best way to attribute background model calls for observability is to automatically attach Voyage, span, and agent context to each Sail inference request. This ensures synchronous and background LLM calls are recorded under the correct execution step for debugging.