Mixedbread (direct)

Direct Mixedbread AI API calls (mixedbread-ai SDK or raw HTTP, no framework). All requests go to https://api.mixedbread.com/ (confirmed live against Mixedbread's own docs) — the vendor's published surface is embeddings and reranking, reached under /v1 (per public API reference; the exact path names and usage-field layout could not be independently re-verified live in this session — the deeper API-reference pages 404'd/ 403'd on fetch — so treat the JSONPaths below as the standard gen_ai.* convention, not vendor-confirmed field names). Mixedbread's embeddings route is commonly used as an OpenAI-SDK drop-in (same request/response shape as OpenAI's /v1/embeddings). Mixedbread has no admin/billing API (usage is dashboard-only), so spend is only capturable via this customer-side OTLP push. Token accounting is input-only: embeddings and rerank calls have no completion step, so there is no meaningful output_tokens metric — one that could only ever read zero would be worse than omitting it. input_tokens and total_tokens are what this connector reports instead. No telara-observability provider wrapper exists for Mixedbread today (the package's optional-dependencies only cover langchain/langgraph/ crewai/autogen/llamaindex/openai-agents/openai/anthropic/azure-openai) — spans must come from the customer's own OTel instrumentation around the Mixedbread SDK/HTTP call. Spans should be tagged telara.span.kind=provider so the cost rollup excludes them when an active framework adapter already covers the same call (double-count guard).