Crate ailoop_core 

Foundation crate for the ailoop workspace: shared vocabulary used by the ailoop façade and by every adapter / middleware crate.

Most application code does not depend on ailoop-core directly — it depends on ailoop, which re-exports the types you need to build a Conversation and write a ChatMiddleware. Reach for ailoop-core when:

• implementing a new CompletionModel / CompletionClient adapter for a provider HTTP API;
• writing a generic middleware that should be reusable across façades, examples, or test harnesses;
• building reliability or observability infrastructure on top of the shared event vocabulary.

Layout

• Message, UserBlock, AssistantBlock, ToolResultContent, SystemPrompt — the wire model exchanged between the engine and the provider.
• ChatRequest, ToolDefinition, ToolChoice, ToolTag — the per-turn request shape adapters lower to the provider.
• StreamChunk, FinishReason, Usage — the event stream the engine emits for every run.
• ChatMiddleware, HookAction, ToolDecision — the extension point for transforms, gating, and observation.
• CompletionClient, CompletionModel — the contract every provider adapter implements.
• RetryingModel, Retryable, RetryConfig — backoff decorator that wraps any CompletionModel.

Re-exports

pub use config::RunConfig;

pub use ids::RunId;

pub use ids::StepId;

pub use message::AssistantBlock;

pub use message::CacheControl;

pub use message::Message;

pub use message::Source;

pub use message::SystemBlock;

pub use message::SystemPrompt;

pub use message::ToolResultBlock;

pub use message::ToolResultContent;

pub use message::UserBlock;

pub use middleware::ChatMiddleware;

pub use middleware::HookAction;

pub use middleware::ToolDecision;

pub use request::ChatRequest;

pub use request::ToolChoice;

pub use request::ToolDefinition;

pub use request::ToolTag;

pub use retry::RetryClassification;

pub use retry::RetryConfig;

pub use retry::Retryable;

pub use retry::RetryingModel;

pub use stream::FinishReason;

pub use stream::StreamChunk;

pub use stream::Usage;

Modules

config

Per-run configuration: see RunConfig.

ids

Run and step identifiers used by the engine to correlate crate::StreamChunks and crate::ChatMiddleware hooks.

message

Conversation message model: Message, its block enums, and the SystemPrompt / CacheControl support types.

middleware

ChatMiddleware extension point and its decision enums (HookAction, ToolDecision).

request

Per-turn request shape: ChatRequest plus ToolChoice, ToolDefinition, and ToolTag.

retry

Reliability decorator: RetryingModel wraps any CompletionModel and retries setup-time failures with exponential backoff that honours Retry-After hints surfaced by the adapter.

stream

Engine event vocabulary: StreamChunk, FinishReason, and Usage.

Structs

CancellationToken

A token which can be used to signal a cancellation request to one or more tasks.

CharTokenizer

Fallback Tokenizer that approximates tokens as text.len() / 4.

Traits

CompletionClient

Factory for CompletionModel handles bound to a provider account / endpoint.

CompletionModel

Provider-agnostic streaming completion contract.

Tokenizer

Counts tokens in text and full messages.