Vectorless

A hierarchical, reasoning-native document intelligence engine.

Features

• Tree-based indexing — Documents are organized as hierarchical trees, not flat vectors
• LLM-driven retrieval — Uses reasoning to navigate document structure
• Multi-format support — Markdown, PDF, HTML, DOCX (planned)
• Workspace persistence — LRU-cached storage with lazy loading
• Configurable retrieval — Pluggable retriever strategies (LLM navigate, beam search, MCTS)

Quick Start

use vectorless::client::{Vectorless, VectorlessBuilder};

#[tokio::main]
async fn main() -> vectorless::core::Result<()> {
    // Create client
    let mut client = VectorlessBuilder::new()
        .with_workspace("./workspace")
        .build()?;

    // Index a document
    let doc_id = client.index("./document.md").await?;

    // Query
    let result = client.query(&doc_id, "What is this about?").await?;
    println!("{}", result.content);

    Ok(())
}

Configuration

Create config.toml in your project root:

[summary]
model = "gpt-4o-mini"
endpoint = "https://api.openai.com/v1"
api_key = "sk-..."

[retrieval]
model = "gpt-4o"
retriever_type = "llm_navigate"
top_k = 3

[storage]
workspace_dir = "./workspace"

Status

Early development. Core functionality works:

• ✅ Markdown indexing with LLM summaries
• ✅ Tree-based retrieval via LLM navigation
• ✅ Workspace persistence with LRU cache
• 🚧 PDF/HTML/DOCX parsing
• 🚧 Additional retriever strategies

License

Apache-2.0