model for OpenAI-shaped chat, completions, embeddings, images, audio, and vector stores. Use these guides to put that deployment in an app you already run.
Connection settings
model / deployment arguments must be the id from Models, not a public OpenAI or Hub name.
Choose a path
LangChain
ChatOpenAI and OpenAIEmbeddings against your deployment.LlamaIndex
LLM + embeddings settings for indexes and query engines.
Embeddings + FAISS
Embed on SynapsAI, retrieve locally with FAISS.
Gradio / FastAPI
Keep the API key on the server; proxy from a UI or API.
Migrate from OpenAI
Environment variables, endpoint map, and differences (agents, custom containers, error types).
Minimal REST check
/v1/agent/..., /v1/vector_stores) — call those from your backend, not from browser bundles.
Production notes
- Retry 429 with exponential backoff — Rate limits.
- Scope keys per app — API keys.
- Never ship keys in client-side code — Security.