OpenAI
OpenAI's GPT models and API — one of the providers I integrate, from NovaMX's classification engine to Akashi's OpenAI adapter.
OpenAI provides the GPT family of models and an API for chat, structured output, tool calling and embeddings.
How I use it
- NovaMX CRM — OpenAI models power the LLM-based lead and property classification engine.
- Akashi —
akashi_openaiis one of the provider adapters; aFallbackModelcan fail over between OpenAI, Google Gemini and Anthropic Claude based on the error type. - Labs — langgraph-ts-agents runs on an OpenAI or Anthropic key; langgraph-langserve-starter and local-rag-chatbot switch to OpenAI with a single environment variable.
I design around providers rather than for one: the recurring pattern across my projects is a provider-neutral seam with OpenAI as one interchangeable option.
Why it matters
Keeping OpenAI behind an interface means a rate limit, outage or price change does not become a rewrite. In Akashi that is enforced in code: failover is decided by error type, and mid-stream failover is blocked because re-running an already emitted stream would duplicate output.