Struct CloudAiRequest 

pub struct CloudAiRequest {

    pub model: String,
    pub prompt: String,
    pub max_tokens: Option<u32>,
    pub temperature: Option<f32>,
    pub api_key: String,
    pub system: Option<String>,
    pub base_url: Option<String>,
    pub top_p: Option<f32>,
    pub top_k: Option<u32>,
    pub stop: Option<Vec<String>>,
    pub stream: bool,
    pub call_id: Option<String>,
    pub reasoning: Option<bool>,
    pub images: Vec<String>,
    pub tools: Vec<ToolSpec>,
    pub response_schema: Option<Value>,
}

Fields

model: String

prompt: String

max_tokens: Option<u32>

temperature: Option<f32>

api_key: String

system: Option<String>

Optional system-role text. Each provider maps it to its native shape:

• Claude: top-level system field on the request payload.
• OpenAI: prepended {role:"system"} message.
• Gemini: top-level system_instruction.

None sends a single user-role message with no system text - the cloud-AI node path, which has no per-task system prompt.

base_url: Option<String>

Override endpoint for the OpenAI-compatible local provider (e.g. http://localhost:11434/v1/chat/completions). Cloud providers ignore this and use their fixed ENDPOINT const; the local provider requires it.

top_p: Option<f32>

Nucleus sampling cutoff (0..=1). Forwarded by the OpenAI-compatible body builder (top_p); cloud providers with their own builders (Claude, Gemini) ignore it. None means “let the server decide”.

top_k: Option<u32>

Top-K sampling cutoff. Non-standard in OpenAI’s API but accepted by llama.cpp / Ollama / LM Studio; the OpenAI-compatible body builder forwards it verbatim when set.

stop: Option<Vec<String>>

Stop sequences (any element halts generation). Forwarded as the OpenAI stop field. Empty None means no caller-supplied stops.

stream: bool

When true the caller is invoking the provider’s invoke_stream path and expects per-chunk events. Providers always set stream: true in the HTTP payload when this is set.

call_id: Option<String>

Caller-supplied correlation id stamped onto every emitted LlmStreamEvent. Required when stream = true; harmless otherwise.

reasoning: Option<bool>

Per-request reasoning/thinking toggle for models that support it. Some(true) asks the provider to enable extended thinking, Some(false) disables it, None leaves the server/model default. Each provider maps it to its native shape (OpenAI-compat chat_template_kwargs, Claude thinking, Gemini thinkingConfig).

images: Vec<String>

Image references for vision models - http(s) URLs or data: URLs (the executor resolves local paths to base64 data: URLs before building the request). Empty for text-only calls.

tools: Vec<ToolSpec>

Tool/function specs the model may call. Empty disables tool use; the actual call loop runs in CloudAiProvider::invoke_tools.

response_schema: Option<Value>

Optional JSON Schema for structured output. When set, the OpenAI-compatible body builder adds a response_format: json_schema so the model is constrained to matching JSON (providers with their own body builders - Claude / Gemini - ignore it and rely on the prompt). The "structured" task sets this from the node’s outputSchema.

Trait Implementations

impl Clone for CloudAiRequest

fn clone(&self) -> CloudAiRequest

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for CloudAiRequest

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for CloudAiRequest

fn default() -> CloudAiRequest

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for CloudAiRequest

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for CloudAiRequest

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.