Quickstart

This is a skeleton quickstart. Exact package names, method signatures, and endpoints are finalized alongside the SDK and API reference issues — treat the snippets below as the intended shape.

1. Install

npm install @vectorless/sdk

pip install vectorless

go get github.com/hallelx2/vectorless-go

2. Configure

Set your engine endpoint and an LLM provider key.

export VECTORLESS_API_URL="https://your-engine.example.com"
export VECTORLESS_API_KEY="vl_..."

3. Ingest a document

import { Vectorless } from '@vectorless/sdk';

const vl = new Vectorless();

const doc = await vl.documents.ingest({
  source: './annual-report.pdf',
});

from vectorless import Vectorless

vl = Vectorless()

doc = vl.documents.ingest(source="./annual-report.pdf")

4. Ask a question

The agent navigates the document tree with treewalk and returns an answer with citations.

const result = await vl.ask({
  document: doc.id,
  question: 'What changed in Q3 revenue, and why?',
  strategy: 'treewalk',
});

console.log(result.answer);
console.log(result.citations); // -> exact tree nodes

result = vl.ask(
    document=doc.id,
    question="What changed in Q3 revenue, and why?",
    strategy="treewalk",
)

print(result.answer)
print(result.citations)  # -> exact tree nodes

What you get back

Every response carries the answer and the nodes it came from:

{
  "answer": "Q3 revenue rose 12% QoQ, driven by ...",
  "citations": [
    { "node": "3.2.1", "title": "Q3 Results", "path": ["Financials", "Q3"] }
  ],
  "strategy": "treewalk"
}

Next, read Core Concepts to understand how the tree and treewalk produce these citations.

1. Install

2. Configure

3. Ingest a document

4. Ask a question

What you get back

On this page