Current configured baseUrl = / (default value)\n We suggest trying baseUrl = \n\n',document.body.prepend(n);var e=document.getElementById("__docusaurus-base-url-issue-banner-suggestion-container"),s=window.location.pathname;e.innerHTML="/"===s.substr(-1)?s:s+"/"}
Deep and reliable. Vectorless plays nicely with your documents. Ask questions in plain language; get answers by reasoning with Vectorless. Compile a document, then ask in plain language — every answer carries its source. Same engine, same answers — index and query your documents without writing any code. From raw file to grounded answer — compile to a semantic tree, then reason over it. No vectors anywhere in the pipeline. Point Vectorless at your documents and ask in plain language. Every answer comes back with a traceable source.Knowing by reasoning,
not vectors.pip install vectorlessUse it from Python
import asynciofrom vectorless import Engine
async def main(): async with Engine(api_key="sk-...", model="gpt-4o") as engine: doc = await engine.compile(path="./report.pdf")
res = await engine.ask( "What is the total revenue?", doc_ids=[doc.doc_id], ) print(res.answer) # grounded answer for ev in res.evidence: # ...with sources print(ev.node_title, ev.source_path)
asyncio.run(main())
Or straight from your terminal
# index a document (or a folder with -r)$ vectorless add ./report.pdf
# ask a one-off question$ vectorless query "What is the total revenue?"
# interactive REPL over your documents$ vectorless ask
# inspect the compiled tree$ vectorless tree <doc_id>$ vectorless list
Inside the engine
Stop embedding.
Start reasoning.