Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.mirobody.ai/llms.txt

Use this file to discover all available pages before exploring further.

OpenAI is Mirobody’s global cluster default LLM backend — covers GPT-5 / GPT-4o / o-series reasoning / embeddings.
China cluster note: OpenAI is not directly reachable from mainland China. You’d need your own compliant egress; Mirobody does not provide a proxy. On the CN cluster prefer DashScope or Volcengine Ark instead.
Official docs: OpenAI Platform · Models · API Keys · Pricing

​
1. Register and get a key

1

Sign up on OpenAI Platform

Go to platform.openai.com. Default organization is fine for a single key.
2

Add a payment method

OpenAI is pre-paid. Add a card under Billing → Payment methods; new accounts get $5 in trial credit.
3

Create a secret key

On API Keys, create a secret key (sk-...). Shown once — save immediately.

​
2. Endpoint info

FieldValue
Base URLhttps://api.openai.com/v1
AuthAuthorization: Bearer sk-...
RegionGlobal, auto-routed
ProtocolNative OpenAI

​
3. Available models

Mirobody providerUnderlying model idNotes
generalgpt-5 / gpt-5-miniGPT-5, strong general purpose
expertgpt-5-pro or o3-proDeep reasoning
— (when global cluster lands)gpt-4o, gpt-4o-miniMultimodal (vision/audio)
—o3 / o4-miniReasoning series
—text-embedding-3-large / -smallEmbeddings
See OpenAI Models for the live list.

​
4. Use it in Mirobody (global cluster)

curl -N -X POST <global-cluster-base-url>/api/chat \
  -H "Authorization: Bearer $TOKEN" \
  -H "X-LLM-Provider: openai" \
  -H "X-LLM-Api-Key: sk-..." \
  -H "Content-Type: application/json" \
  -d '{"question":"hi","agent":"App","provider":"general"}'

​
Custom base URL (Azure OpenAI / self-hosted relay)

For Azure OpenAI, self-hosted LiteLLM, or any OpenAI-compatible gateway: 
X-LLM-Base-Url: https://your-azure-endpoint.openai.azure.com/openai/deployments/<deployment>
Mirobody will use that URL in place of https://api.openai.com/v1.

​
Troubleshooting

  • invalid_api_key → key revoked or rotated; regenerate.
  • insufficient_quota → balance exhausted or monthly cap hit.
  • Mainland-China timeouts → expected; switch to DashScope / Volcengine.
  • Region-restricted model → some preview models are not GA in every region; check the model card.