Module llm

Expand description

Agent model port and execution policy, using the shared model contract.

Structs§

Completion: Provider-neutral completion returned to Runtime.
CompletionRequest: One logical LLM invocation.
Continuation: Provider-issued replay data attached to exactly one completed model message. Adapters own the format and compatibility key. Callers must not synthesize or edit payloads.
LlmFailure: Model-specific failure that can retain usage returned before output validation failed.
Message: Provider-neutral conversation message.
ModelConstraints: Capabilities required by an invocation, without naming a provider or model.
ModelMode: Logical model mode selected by the product, independent of provider and model identifiers.
ModelProfile: Provider-neutral limits for one frozen logical model route.
TokenUsage: Normalized token accounting reported by an implementation.
ToolCall: Tool call returned by an LLM.
ToolDefinition: Tool schema exposed to an LLM.
ToolPolicy: Runtime policy declared by a tool implementation.
UsageSummary: Aggregated model and tool usage for one Agent run.
UseCase: Logical model use case resolved by the deployment’s LLM implementation.

ApprovalPolicy: Whether approval is fixed by the schema or strengthened after canonical preparation.
ApprovalRequirement: When a prepared tool operation requires a user decision.
ContentPart: One typed part of a model message.
FinishReason: Why a model invocation stopped.
MessageRole: Role of a message in a model conversation.
ToolEffect: Side-effect class used by Runtime access policy.