Crate senno 

senno

senno — a provider-agnostic LLM client and streaming tool-calling agent engine, Vertex AI first, with built-in history compaction.

• Core (no default features): canonical LLM types, the LlmProvider, EmbeddingProvider, BatchGenerationProvider, and BatchEmbeddingProvider traits, one_shot, and the agent engine.
• vertex feature: providers::vertex — Gemini and Claude on Vertex AI with API-key or service-account auth, plus Gemini text embeddings and Gemini Batch API bulk generation and bulk embedding at half price.
• batch feature (implied by vertex): run_batch and run_embedding_batch — submit a batch job and poll it to completion.
• axum feature: agent::to_sse_event — adapt agent events to axum::response::sse.
• mcp feature: mcp — register foreign MCP servers and add their tools to any agent::AgentFlow.

Re-exports

pub use error::Error;

Modules

agent

error

mcp

providers

Provider implementations of crate::LlmProvider, each behind its own feature flag.

Structs

BatchJob

BatchPolling

EmbedRequest

EmbedResponse

Embedding

GenerateRequest

GenerateResponse

GenerationParams

Message

ParameterSchema

ToolAnnotations

ToolDefinition

Usage

Enums

BatchState

ContentPart

EmbedApi

Vertex serves embeddings through two disjoint APIs, and a model speaks exactly one of them.

EmbedTaskType

Role

SchemaType

StreamChunk

Traits

BatchEmbeddingProvider

BatchGenerationProvider

EmbeddingProvider

LlmProvider

Functions

one_shot

run_batch

run_embedding_batch

wait_for_batch

wait_for_embedding_batch

Type Aliases

LlmStream