LLM chains and RAG for Rust developers who miss LangChain

What it does

langchain-rust is a Rust reimplementation of the LangChain framework. It wraps providers like OpenAI, Azure, Anthropic, and Ollama into a common interface, then lets you chain prompts, vector searches, and tools into sequential or conversational workflows. The crate also handles document ingestion from PDFs, HTML, CSV, and even Git history, feeding them into backends like Postgres, Qdrant, or SQLite for RAG. Everything is async and built around macros and builder patterns that will feel familiar if you’ve used the Python original.

The interesting bit

The most notable choice is the heavy use of macros like message_formatter! and fmt_placeholder! to construct prompts at compile time, leaning into Rust’s metaprogramming rather than string templating. It also bundles a surprising breadth of vector stores and document loaders for a young community port, suggesting the maintainers are aiming for parity rather than a minimal subset.

Key highlights

• Supports OpenAI, Azure, Anthropic Claude, and Ollama for both chat and embeddings
• Vector store integrations include Qdrant, Postgres, OpenSearch, SQLite, and SurrealDB
• Document loaders cover PDF, HTML, CSV, Pandoc conversions, Git commits, and source code
• Built-in chains for Q&A, SQL generation, conversation, and retrieval-augmented generation
• Optional feature flags (postgres, qdrant, sqlite-vss, etc.) keep compile times selective

Verdict

Worth a look if you are already committed to Rust and need standard LLM orchestration patterns like RAG and tool-calling agents under one crate. If you are in Python or JavaScript, the original LangChain ecosystems are still far larger and better documented.