AiCreateProfileInput

Input for creating a new profile — the same shape as Profile without the engine-generated fields (id, createdAt).

Name	Type	Description	Notes
name	String	User-defined profile display name.	[required] [example: OpenAI GPT-4o]
providerType	AiProviderType	Provider type for this profile. Use external to delegate all HTTP transport to PlatformAdapter.externalFetch while reusing an existing provider's response parser — see Profile.basedOn for the format selector.	[required] [example: openai]
basedOn	AiBuiltinProviderType	Selects the response-format parser used by the external provider. Ignored for any other providerType. Supported values are openai, anthropic, mistral and openrouter. Remaining values (genai, stabilityai, …) are accepted by the type but not yet implemented; passing one raises an error at request time.	[optional] [example: openai] [enum: anthropic, ollama, openai, openaicompatible, together, openrouter, genai, deepseek, xai, lm-studio, mistral, groq, zhipu, stabilityai, gpt4all, onlyoffice, external]
baseUrl	String	Base URL of the provider API.	[required] [example: https://api.openai.com/v1]
key	String	API key or token. Optional for local providers.	[optional] [example: sk-your-provider-api-key]
headers	Map	Extra HTTP headers sent with every request to this provider. Merged into the SDK client's default headers; an explicit Authorization here wins over the one derived from key. Honoured by the OpenAI-family providers.	[optional] [example: {X-Organization=acme}]
modelId	String	Selected model ID within this provider.	[required] [example: gpt-4o]
reasoning	Boolean	Whether extended thinking is enabled for this profile's model.	[optional] [example: false]
reasoningSupport	AiReasoningSupport	Extended-thinking capabilities of the selected model as reported by the provider's catalogue at save time (see Model.reasoningSupport). When present the composer's Effort row follows it exactly; when absent the provider's id-based table answers. Hosts persist it with the rest of the profile.	[optional]
capabilities	BigDecimal	Bitmask of capabilities supported by the selected model.	[optional] [example: 7]
canUseTool	Boolean	Result of the live tool-capability probe performed at create time and on changes to modelId / providerType / baseUrl. undefined means the probe has never run for this profile (legacy record).	[optional] [example: true]
useResponsesApi	Boolean	Result of the live Responses-API probe (parallel to canUseTool). true means the model speaks /v1/responses and the OpenAI provider must route through client.responses.create — required for gpt-5+ reasoning models that reject reasoning_effort together with tools on /v1/chat/completions. Probed at create time and whenever modelId / providerType / baseUrl change. undefined means the probe never ran (legacy record) — readers treat that as false.	[optional] [example: false]
isCloudProvider	Boolean	Whether this profile uses a cloud-hosted provider (e.g. ONLYOFFICE DocSpace).	[optional] [example: true]
useProxy	Boolean	Route every provider request through the host's fetchProxy instead of the global fetch. Useful when the host runs the widget in a sandbox without direct network access (CORS, custom auth, etc.). Has no effect when the PlatformAdapter.fetchProxy is not configured.	[optional] [example: false]