Skip to main content
SupaCortex
Sign in Book a Demo
Feature

Drop in the documents people keep asking about

The policy PDF, the product manual, the pricing sheet, the onboarding guide. Upload them in bulk and each one is parsed, split into passages and indexed - ready to be cited in an answer.

Book a demo See pricing
  • PDF, Word, Excel, Markdown, HTML
  • Whole .zip files
  • Up to 500 files at once

How do document uploads work in SupaCortex?

You choose a project and upload files - one at a time, many at once, or as a .zip. Each supported file becomes a document that is parsed, split into passages and indexed for both meaning-based and keyword search. While it is being processed it shows as queued, then indexed or failed with the reason. Every document can carry a category, tags, a customer ID and a trust tier, and can be reindexed or deleted at any time.

  • Supported: PDF, DOCX, XLSX, Markdown, TXT and HTML.
  • A .zip becomes one document per supported file inside it.
  • Up to 50 MB per file and 500 files per upload.
  • Category, tags, customer ID and trust tier on every document.
  • Delete removes the document and every indexed passage from it.
  • An ingest API for internal tools, updating records by your own reference.
How it works

From file to cited answer

  1. 01

    Upload

    Pick the project, drop the files, and optionally set a category, tags, a customer ID and a trust tier for the batch.

  2. 02

    Parse and split

    Text is extracted with its headings, so a passage knows which section it came from - the section a citation will name.

  3. 03

    Index

    Each passage is indexed for meaning and for exact words, so both "how do I claim travel" and "form TR-7" find it.

  4. 04

    Answer and cite

    When a question matches, the passage is used as evidence and cited with the document and section it came from.

In detail

Small controls that make answers better

Tell search what to trust

Not every document deserves the same weight. The signed policy should beat a forum thread that paraphrases it.

  • Tier 1, official: ranked higher than everything else.
  • Tier 2, standard: the default for most documents.
  • Tier 3, community and past tickets: ranked lower, and dropped when enough official evidence exists.
  • Categories narrow a search; tags help you find and manage documents later.

Keep the library tidy

Knowledge goes stale. The useful thing is to make replacing it as easy as adding it.

  • A status for every document: queued, indexed, failed (with the reason) or deleting.
  • Reindex a document with one click; delete it and its passages go with it.
  • Search the document list by title or file name, and filter by status or source.
  • Documents from feeds are managed by the feed, so a crawl and a hand upload never fight over one page.

For files that change often, a feed is usually better than re-uploading: it re-reads the source on a schedule and only fetches what changed.

Related

Preparing Documents for AI Answers

Ten habits that make the same documents answer twice as well. Most take minutes.

Website & Help-Centre Crawling

Point it at a help centre, Google Site or docs site and it stays in step on a schedule.

Google Drive, SharePoint & Confluence

Connect once with OAuth, pick folders, sites or spaces, and only changed files are fetched again.

Answers With Citations

Hybrid search, a confidence gate and checked citations - it says "I don't know" instead of guessing.

Knowledge Projects

Keep each product, team or audience in its own project - its own sources, widget and API keys.

Search & Answer API

The same brain behind your own tools: scoped API keys, /v1/search and /v1/answer.
Frequently asked

Questions people ask before they start

PDF, Word (DOCX), Excel (XLSX), Markdown, plain text and HTML. A .zip is expanded and each supported file inside it becomes its own document; anything else is listed back to you as rejected, with the reason.

Up to 50 MB per file and 500 files in one upload. Storage per plan is 1 GB on Starter, 10 GB on Growth, 100 GB on Business and unlimited on Enterprise, and the number of documents is 500, 5,000, 50,000 and unlimited respectively.

Only if the PDF contains text. A scanned image with no text layer has nothing to index, and the document is marked as failed with the reason, so you know to replace it with a text version.

It tells search how much to rely on a document. Tier 1 is official material and is ranked higher; tier 2 is standard; tier 3 is community content and past tickets, ranked lower and dropped altogether when enough official evidence answers the question.

No. Files are split into passages and indexed for search; the model reads the relevant passages at the moment it answers. Nothing is trained on your documents, and they are processed on open models on infrastructure we operate rather than sent to a third-party AI API.

Yes. The ingest API accepts content with your own reference for each record and updates it in place when you send it again - useful for content that lives in an internal tool rather than in files.

Bring the ten documents your team is asked about most.

Book a 30-minute demo on your own documents. We set up a 14-day trial workspace for you - no credit card, no setup fee.

Book a demo See pricing

Questions first? Email info@greatdevelopers.com or call +91 8585953085.

Still curious? Ask your favorite AI about SupaCortex.

ChatGPT Claude Perplexity Grok