Skip to main content
Vectorless

No vector database. No embeddings. No similarity search.
Retrieve by reasoning, not by math.

Get StartedGitHub
Python
import asyncio
from vectorless import Engine, IndexContext


async def main():
    engine = Engine(
        api_key="sk-...",
        model="gpt-4o",
    )


    # Index a document
    result = await engine.index(
        IndexContext.from_path("./report.pdf")
    )
    doc_id = result.doc_id


    # Query — LLM navigates the tree
    result = await engine.query(
        doc_id, "What is the total revenue?"
    )
    print(result.single().content)


asyncio.run(main())

Why Vectorless?

RAG without the baggage.

🧠

Reasoning-Native

LLMs navigate hierarchical document trees with semantic understanding — not vector proximity.

🗂️

No Vector Database

Eliminate embedding pipelines, vector stores, and similarity search entirely. Trees are the index.

⚡

Rust-Powered

Core engine in Rust with Python bindings. Arena-based trees, async I/O, and zero-copy traversal.

🔍

Multi-Algorithm Search

Beam search, MCTS, and greedy algorithms with LLM-guided Pilot at key decision points.

📊

Explainable Results

Full reasoning chain traces every navigation decision. Audit how and why content was retrieved.

📄

PDF & Markdown

Index PDFs and Markdown out of the box. Hierarchical structure extracted automatically.

How It Works

01

Index

Parse documents into hierarchical semantic trees with LLM-generated summaries.

02

Navigate

Pilot uses LLM to navigate the tree at key forks — beam search explores multiple paths in parallel.

03

Retrieve

Evaluate sufficiency and backtrack if needed. Aggregate only the most relevant content within budget.

Start building in minutes

pip install vectorless

Read the DocsView on GitHub