Struct LLMRequest 

pub struct LLMRequest {
    pub input: Vec<APIInputMessage>,
    pub model: String,
    pub tools: Vec<AzureLLMToolDefinition>,
    pub tool_choice: Option<LLMToolChoice>,
    pub parallel_tool_calls: Option<bool>,
    pub max_output_tokens: Option<i32>,
    pub text: Option<RequestTextOptions>,
    pub prompt_cache_key: Option<String>,
    pub params: LLMRequestParams,
}

Fields

input: Vec<APIInputMessage>

model: String

tools: Vec<AzureLLMToolDefinition>

tool_choice: Option<LLMToolChoice>

parallel_tool_calls: Option<bool>

max_output_tokens: Option<i32>

text: Option<RequestTextOptions>

prompt_cache_key: Option<String>

Routes requests sharing a prompt prefix to the same cache entry. None for a model that has no Azure prompt cache.

params: LLMRequestParams

Implementations

impl LLMRequest

pub fn new( model: String, input: Vec<APIInputMessage>, params: LLMRequestParams, ) -> Self

A request with no tools, tool choice, cache key, or output cap — the shape every one-shot LLM call outside a chatbot turn starts from. Override fields via struct-update syntax.

pub(super) async fn build_and_insert_incoming_user_message_to_db( conn: &mut PgConnection, chatbot_configuration_id: Uuid, conversation_id: Uuid, message: &str, page_context: Option<ChatbotPageContext>, user_context: &ChatbotTurnContext, app_config: &ApplicationConfiguration, ) -> ChatbotResult<Self>

Writes the learner’s new message to the conversation and builds the request for the turn that answers it.

The write happens in one transaction with abort_pending_client_tool_calls, so a client answering a suspended tool call at the same moment either gets its answer in before the learner moved on, or finds the call already aborted. page_context is recorded ahead of the message it gives context for.

pub(super) async fn build_from_conversation( conn: &mut PgConnection, configuration: &ChatbotConfiguration, conversation_id: Uuid, user_context: &ChatbotTurnContext, app_config: &ApplicationConfiguration, ) -> ChatbotResult<Self>

Builds the request for a turn from the conversation exactly as it is stored, writing nothing.

Both the turn that follows a new user message and a resumed turn go through here, a resumed one adding nothing of its own beyond the tool output that woke it. The tools the request offers depend on user_context, so a caller who has lost a role is not offered a tool that needs it again.

fn assemble( configuration: &ChatbotConfiguration, conversation_id: Uuid, inputs: TurnInputs, app_config: &ApplicationConfiguration, enabled_tool_categories: &EnabledToolCategories, ) -> ChatbotResult<Self>

Assembles the request’s shape — grounding instruction, tool list, tool_choice, params, prompt cache key, and the system-message prepend — from data TurnInputs::load already read from the database.

Trait Implementations

impl Clone for LLMRequest

fn clone(&self) -> LLMRequest

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LLMRequest

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for LLMRequest

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for LLMRequest

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

where __S: Serializer,

Serialize this value into the given Serde serializer.