synwire 0.1.0

Convenience re-exports and reference implementations for Synwire
Documentation

synwire

Convenience umbrella crate for the Synwire AI framework. One dependency gives you chat models, embeddings, graph orchestration, agent runtime, and utilities.

What this crate provides

Re-exports from all Synwire crates plus:

  • agent::prelude::* — Agent, AgentNode, Runner, Directive, DirectiveResult, AgentError, AgentEvent, Session, SessionManager, Usage, OutputMode, HookRegistry
  • cache::CacheBackedEmbeddings — moka-backed embedding cache wrapping any Embeddings impl
  • chat_history::InMemoryChatMessageHistory + RunnableWithMessageHistory — stateful conversation management
  • prompts — few-shot prompt templates, example selectors
  • text_splitters — RecursiveCharacterTextSplitter, character splitter
  • output_parsers — CommaSeparatedListOutputParser, RetryOutputParser, RegexParser, EnumOutputParser, XmlOutputParser, CombiningOutputParser

Quick start

[dependencies]
synwire = { version = "0.1", features = ["openai"] }
tokio = { version = "1", features = ["full"] }

First chat:

use synwire::prelude::*;
use synwire::llm_openai::ChatOpenAI;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let model = ChatOpenAI::builder()
        .model("gpt-4o")
        .api_key_env("OPENAI_API_KEY")
        .build()?;

    let result = model.invoke("Hello, Synwire!").await?;
    println!("{}", result.content);
    Ok(())
}

Cached embeddings:

use synwire::cache::CacheBackedEmbeddings;
use synwire::llm_openai::OpenAIEmbeddings;

let base = OpenAIEmbeddings::builder().api_key_env("OPENAI_API_KEY").build()?;
let cached = CacheBackedEmbeddings::new(base, 1000); // LRU cache, 1000 entries

Stateful conversation:

use synwire::chat_history::{InMemoryChatMessageHistory, RunnableWithMessageHistory};
use synwire::llm_openai::ChatOpenAI;

let model = ChatOpenAI::builder().model("gpt-4o").api_key_env("OPENAI_API_KEY").build()?;
let history = InMemoryChatMessageHistory::new();
let chain = RunnableWithMessageHistory::new(model, history);

Feature flags

Flag Enables
openai synwire-llm-openai (ChatOpenAI, OpenAIEmbeddings)
ollama synwire-llm-ollama (ChatOllama, OllamaEmbeddings)

When to use this crate vs individual crates

  • Use synwire for application development and prototypes
  • Use synwire-core directly when publishing a third-party extension crate — it carries no provider dependencies

Documentation