Eden AI (direct)

Direct Eden AI API calls (edenai SDK or raw HTTP, no framework). Eden AI is a multi-provider aggregator — "one API to access OpenAI, Mistral, Gemini, Claude, and dozens of specialized AI providers" for chat, embeddings, translation and more — at base URL https://api.edenai.run/v2/ (confirmed partially live: Eden AI's own docs state "Both share the same base URL (https://api." before the fetched excerpt was truncated; the domain matches the well-documented api.edenai.run). Eden AI also publishes an OpenAI-compatible endpoint ("drop-in replacement — change the base URL"), and per its pricing docs charges "no markup on provider pricing — you pay exactly what the underlying provider charges, plus a [platform fee]", so cost is provider-rate-based even though it is billed through Eden AI. Because Eden AI fronts chat/completion models as well as embeddings, an output_tokens metric is meaningful here and is kept. IMPORTANT for chargeback: the OTLP spans this connector ingests carry Eden AI's own request/response envelope. Whichever underlying provider actually served the call, cost and usage land on Eden AI as the billing entity in this pipeline, not on the underlying model vendor directly — a real limitation for per-vendor chargeback. Per-underlying-provider attribution is only possible if the customer's own instrumentation additionally captures the provider/model Eden AI reports back (gen_ai.response.model), which this connector does try via model_fallback_paths. Eden AI has no separate admin/billing API beyond its dashboard, so spend is only capturable via this customer-side OTLP push. No telara-observability provider wrapper exists for Eden AI 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 Eden AI SDK/ HTTP call, or from the OpenAI instrumentation package if they call Eden AI through its OpenAI-compatible endpoint. 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).