LLM
AI agents: read this documentation natively — index: https://docs.openvoidnet.com/llms.txt, full: https://docs.openvoidnet.com/llms-full.txt, this page: https://docs.openvoidnet.com/docs/llm/overview.md.
LLM Overview
AI models through an OpenAI-compatible API, billed per token.
LLM apps serve exactly one AI model each through an OpenAI-compatible endpoint. Billed per token from usage.total_tokens.
At a glance
| Endpoint | POST /v1-beta/llm/{username}/{appname} |
| Body | {"model":"<published-model-id>","messages":[{...}],"stream":false} |
| Billing unit | Tokens (total_tokens / 1M × price) |
| Start here | Buyer Guide to call · Publisher Guide to publish |
Rules that bite
- The
modelyou send must match the app's published model exactly — a read-only listing is available atGET /v1-beta/llm/{username}/{appname}/models(see Buyer Guide). - Text + multimodal: string content plus
image_url,input_audio,video_url,fileparts pass through verbatim to the publisher (20MB body cap on the LLM route). stream:trueis supported (SSE); the gateway injectsstream_options.include_usage:truewhen missing. Calls without a final usage chunk never go unbilled:502pre-response, error-metered post-relay for live streams.- Publishers: serve
POST {serverUrl}/chat/completions({serverUrl}is your configured Server URL, which already ends in/v1), returnusage.total_tokensgreater than zero on every success.