Rusty Common Sense Guide to AI Engineering

Rust implementations of the book A Common Sense Guide to AI Engineering published by Pragmatic Programmers.

All examples expect the OPENAI_API_KEY to be exported as an environment variable, e.g.,

$ OPENAI_API_KEY=<api-key> cargo run -p hello-world -- ...

As this repo uses Cargo workspace to organize the demonstrates, each demo's source directory name has the book's chapter prefix, e.g., 01-hello-world is the demo from chapter 1. However, when running the demo, drop the chapter prefix when the package -p (as shown above).

Ports by Chapter

• Chapter 1: HeLLMo, World!
  • Creating Our First App
• Chapter 5: Building a Chatbot
  • Augmenting the Prompt
    • Pirate tone
    • Spanish translation
    • Grammar checker
  • Multi-turn dialogue
  • Managing state with memory systems
  • Treating the prompt as an array
• Chapter 6: Augmenting a Prompt with Knowledge
  • Preparing the data
  • Implementing the knowledge chatbot
• Chapter 7: Efficiently Adding Knowledge with RAG
  • Implementing a RAG chatbot
• Chapter 8: Measuring Quality with Evals
  • Setting Up Our App
  • Generating Traces: reuses implementation from [Setting Up Our App][eval-chatbot] for simplification
• Chapter 10: Reducing Hallucinations
  • [Reducing Hallucinations][10-reduce_hallucinations]: reuses implementation from Setting Up Our App for simplification
• Chapter 11: Evaluating and Optimizing RAG
  • Expanding the Query: reuses implementation from [Reducing Hallucinations][10-reduce_hallucinations] for simplication

Notable Differences

Other than the language and the libraries used, this port:

• Uses embedded LanceDB instead of Pinecone for simplicity, i.e., no need to sign up Pinecone; also uses fastembed for text embedding.
• Supports interfacing with local LLM servers, such as LM Studio and Ollama.

Wheel-equivalent (almost) Crates

The following table shows the crates used in place of the wheels:

Wheel	Crate	Remarks
Docling	Transmutation	Augmented with reqwest to retrieve resources from the web
openai	async-openai
-	lancedb	Replaced Pinecone as vector database
-	fastembed	Generate text embeddings when populating vector database

As LanceDB uses ProtoBuf, you may need to install protobuf using Homebrew:

brew install protobuf

Or using asdf-vm:

asdf plugin add protoc https://github.com/paxosglobal/asdf-protoc.git

Resources

For convenience, all the files in resources directory are copied from the book's source code; the copyright of those files belong to the book's author and the publisher.