Aeion AI + OpenAI
Configure OpenAI as a provider in Aeion AI. Covers chat completions, image generation (DALL-E 3 / GPT Image), TTS, STT (Whisper), and embeddings. Pair with Azure OpenAI as a separate provider for regional tenancy + BAA requirements.
Setup — 2 Minutes
- Generate an API key at
https://platform.openai.com/api-keys
Models Available — Multi-Modality
OpenAI's surface, all callable through the same engine.
Azure OpenAI Alternative — When + Why
Azure OpenAI Service is the same models hosted by Microsoft in their Azure regions, with three operational differences:
Cost Math
Real per-1M-token rates (2026).
Common Gotchas
- Rate limits — OpenAI's tier system has different RPM / TPM at each tier. Production deployments should be on at least Tier 4. Configure a fallback chain (e.g.
fallback: ["anthropic", "google"]) so a 429 on OpenAI doesn't fail the request outright.
FAQ
For production deployments: at least Tier 4. Aeion's adapter doesn't depend on tier features beyond standard chat completions.
Not yet — Aeion's OpenAI adapter doesn't currently implement the bidirectional Realtime flow. Use `generateSpeech` + `generateTranscription` for a turn-based voice pipeline instead.
Yes — pass a schema and Aeion uses OpenAI's `response_format: { type: "json_schema" }` under the hood for native structured output.
Yes — Aeion provides a unified tool-use abstraction across providers. Define tools with a schema; matched calls dispatch to your handler functions.
Yes — they're separate providers with separate configs. Use Azure for HIPAA/EU workloads, direct OpenAI for everything else.