Crate cvxtract 

cvxtract

LLM-powered structured extraction from CVs and resumes.

cvxtract loads a CV/resume in any common format (PDF, DOCX, HTML, plain text), sends the text to an LLM, and deserialises the response directly into typed Rust structs — no regex, no hand-written parsers.

Quick start

use cvxtract::{Extractor, Model};

#[tokio::main]
async fn main() {
    // Use any provider — here we use a local quantised model (no API key required).
    let mut extractor = Extractor::new(Some(Model::from_local()));

    match extractor.extract_resume("resume.pdf".into()).await {
        Ok(resume) => println!("{:#?}", resume),
        Err(e) => eprintln!("Extraction failed: {e}"),
    }
}

Providers

Constructor	Backend	Requires
Model::from_local()	llama-cpp-2 on-device (Qwen3.5-2B)	nothing — model auto-downloaded
Model::from_openai()	OpenAI API	OPENAI_API_KEY env var
Model::from_openrouter()	OpenRouter	OPENROUTER_API_KEY env var
Model::from_ollama()	Local Ollama	Ollama running on localhost:11434
Model::from_openai_compatible()	Any OpenAI-compatible endpoint	explicit key + URL
Model::from_copilot()	GitHub Copilot	COPILOT_TOKEN env var

GPU acceleration

Compile with a feature flag to offload the local model to your GPU:

# NVIDIA CUDA
cargo build --release --features cuda
# Apple Silicon (Metal)
cargo build --release --features metal
# AMD / Intel / Vulkan
cargo build --release --features vulkan

Custom types

Implement serde::Deserialize and schemars::JsonSchema on any struct to extract arbitrary shapes from a CV:

use cvxtract::{Extractor, Model};
use schemars::JsonSchema;
use serde::Deserialize;

#[derive(Debug, Deserialize, JsonSchema)]
struct ContactInfo {
    name: String,
    email: Option<String>,
    phone: Option<String>,
}

#[tokio::main]
async fn main() {
    let mut extractor = Extractor::new(Some(Model::from_local()));
    let info: ContactInfo = extractor.extract("resume.pdf".into()).await.unwrap();
    println!("{:#?}", info);
}

Structs

Award

Certification

DateRange

Half-open date range. start and end are both null when dates are unknown.

Document

Represents extracted document content with metadata

DocumentElement

Document elements (structured content)

DocumentMetadata

Document metadata

Education

Experience

Extractor

Orchestrates document loading and LLM-powered structured extraction.

Language

Model

An LLM provider that can generate text from a prompt.

PartialDate

Year-always-present, month/day optional.

Project

Resume

Top-level resume — covers the vast majority of real-world CVs.

SkillGroup

Skills can be flat or grouped by category.

UnstructuredLoader

Loads CV/resume documents in any supported format, automatically detecting the type.

Enums

ExtractionError

Errors that can occur during structured extraction.

FileType

Supported file types

LoaderError

Error types for document loading