Show HN: Rstructor – Painless structured data from LLMs in Rust

I built rstructor, a Rust library for structured LLM outputs. It’s like Python’s Instructor/Pydantic, but with the power and type-safety of Rust.

You define your data models as Rust structs/enums, and rstructor handles the rest: JSON Schema generation, LLM communication (OpenAI, Anthropic, etc.), parsing, and validation.

Example – triaging a bug report from messy text:

    use rstructor::{Instructor, LLMClient, OpenAIClient, OpenAIModel};
    use serde::{Serialize, Deserialize};

    // Define enums for classification
    #[derive(Instructor, Serialize, Deserialize, Debug)]
    enum Severity { Low, Medium, High, Critical }

    #[derive(Instructor, Serialize, Deserialize, Debug)]
    enum IssueType {
        Bug { severity: Severity },
        FeatureRequest,
        Performance { ms_impact: u32 },
    }

    // Define the main structured output
    #[derive(Instructor, Serialize, Deserialize, Debug)]
    struct TriagedIssue {
        title: String,
        issue_type: IssueType,
    }

    let client = OpenAIClient::new(env::var("OPENAI_API_KEY")?)?
        .model(OpenAIModel::Gpt4OMini);

    // Go from messy text to a type-safe Rust struct in one call
    let issue: TriagedIssue = client.materialize(
        "The login page is super slow, takes like 500ms to load, and it crashes if I type my email wrong."
    ).await?;

    // Leverage Rust's powerful pattern matching
    match issue.issue_type {
        IssueType::Performance { ms_impact } => {
            println!("High-priority performance issue: {}ms impact", ms_impact);
        }
        IssueType::Bug { severity } => {
            println!("Bug of severity {:?} found", severity);
        }
        _ => {}
    }

    Key features:
    - Support for OpenAI, Anthropic, Grok (xAI), and Gemini
    - Custom validation rules with automatic detection
    - Nested structures, arrays, and enums with associated data
    - Automatic retry with validation error feedback

Crate: https://crates.io/crates/rstructor

GitHub: https://github.com/clifton/rstructor

Docs: https://docs.rs/rstructor

Would love feedback from anyone building LLM-powered tools in Rust!

Comments URL: https://news.ycombinator.com/item?id=45801441

Points: 1

# Comments: 0

Source: news.ycombinator.com