bosun-ai/swiftide ? reverse-engineered prompt

Build me a Rust library and a few working examples for an LLM app framework that can do three things really well, run an agent that can talk to a model and use tools, orchestrate typed task flows, and stream documents through an indexing and retrieval pipeline for RAG.

I want the basic setup to feel easy to use, with clear examples I can run right away, like a simple assistant, a task workflow that passes structured data between steps, and a search over local documents with embeddings and retrieval. Please make the agent support tool calls, message history, streaming responses, and a simple stop condition. For the task side, I want typed nodes, branching, joining, and the ability to pause and resume. For the indexing side, I want loaders, transformers, embeddings, storage, and querying that can handle bigger workloads without loading everything at once.

If you need current details for model or provider setup, look up the latest docs online.