OpenAI
Streaming text output through the server-side BYOK adapter.
Interface: Responses API
Inspect evidence + limits →Evidence observatory
A wired adapter is not the same as an exact model run. Every entry separates DecaCap implementation evidence, vendor documentation, user-side requirements, and the verification that still has to happen on a real account or device.

Streaming text output through the server-side BYOK adapter.
Interface: Responses API
Inspect evidence + limits →Streaming text output with a separate system instruction and bounded message history.
Interface: Messages API
Inspect evidence + limits →Streaming text output through the Google Generative Language endpoint.
Interface: Generative Language streamGenerateContent
Inspect evidence + limits →Non-streaming final text output through the Moonshot chat-completions endpoint.
Interface: Moonshot OpenAI-compatible chat completions
Inspect evidence + limits →Streaming text output through Perplexity’s OpenAI-compatible endpoint.
Interface: Chat Completions
Inspect evidence + limits →Streaming text output through DeepSeek’s OpenAI-compatible endpoint.
Interface: Chat Completions
Inspect evidence + limits →Streaming text output through xAI’s OpenAI-compatible endpoint.
Interface: Chat Completions
Inspect evidence + limits →Local text inference from any entitled Workbench box through the user-paired loopback companion.
Interface: Device-local /api/chat and /api/tags
Inspect evidence + limits →Local text inference through the user’s companion when the runtime follows the supported OpenAI-compatible response contract.
Interface: User-configured loopback chat-completions endpoint
Inspect evidence + limits →An explicitly started local fine-tuning command using GoldCap-prepared data on compatible Apple silicon systems.
Interface: Local command launched by the companion
Inspect evidence + limits →