Skip to content
CorpusKit

CorpusKit, the open source research portal for Progress Agentic RAG

Search a research collection, ask it questions and follow the connections. Every answer cites the passage it came from. Apache 2.0, built and maintained by Noice.

Visit corpuskit.orgCorpusKit Cloud
01 / Inside a portal

One portal for a whole research collection

A portal is built around one organisation's research collection. People search it, ask it questions, browse the library and follow the links between topics. Administrators add and maintain the material.

Explore

Suggested questions and topics drawn from your own collection, so first-time visitors know where to start.

Search

Match by meaning, exact words or both. Filter by topic and document type, and go straight to the passage.

Ask

Questions in plain language, answered from the collection with numbered citations and a confidence rating.

Library

Every document with a readable title, summary and key takeaways written from its contents.

Knowledge graph

The names, places and topics in the research and how they connect, with the passage behind each link.

Investigations

Long-running questions that collect evidence and notes over weeks or months, and export to Word.

Generate

Briefings, comparisons, timelines and FAQs from the collection, each with numbered references.

Manage

Uploads, scheduled website re-syncs, labels, and a check that flags any document to reload.

Access

Public, single sign-on or invite-only, with four roles and an exportable audit log.

02 / Checked answers

Answers you can check against the research

Figures, populations, named studies and years are tested against the cited text, sentence by sentence. A sentence that fails the check is removed, and a note says so.

  1. Step 01

    Choose a claim

    Every claim in an answer is linked to the passage that supports it.

  2. Step 02

    Open its citation

    The numbered citation opens the source document.

  3. Step 03

    Read the passage

    The supporting passage is highlighted in the source.

  4. Step 04

    Review the rating

    Each answer is rated high, moderate or low confidence.

03 / Run it

Run it yourself, or let us run it

Noice builds and maintains CorpusKit. We also stand portals up: loading the collection, including scanned back catalogues, tuning the topics and graph, and measuring answer quality. The licence and knowledge box are in your organisation's name, so your documents, index, labels and graph stay yours.

Open source · Apache 2.0

Run CorpusKit yourself

The application code is public under the Apache 2.0 licence. Clone the repository, connect your Progress Agentic RAG account and run the portal as a container or on common cloud platforms.

corpuskit.org →GitHub repository →

Hosted by Noice

CorpusKit Cloud

We host the portal at your own web address, apply upgrades and handle support. You choose where documents live: Australia, Europe or the United States. A managed plan adds care of the collection, from loading new material to tuning and quality reviews.

corpuskit.cloud →Try the demo →
04 / Who it is for

Organisations with years of research behind them

Rural research and development corporations, research funders, government agencies, universities and research institutes, corporate R&D teams and industry bodies. The reports are published, and most were funded so that people would use them. CorpusKit makes them answerable.

05 / Questions

CorpusKit questions

CorpusKit is the open source research portal for Progress Agentic RAG. It turns a research collection into a portal where people can search, ask questions and get cited answers, browse a library and follow a knowledge graph. Noice builds and maintains it.

Yes. The application code is published under the Apache 2.0 licence on GitHub. You can run it, change it and use it commercially. Content storage, retrieval and generation run on a Progress Agentic RAG account, which has its own terms.

A Progress Agentic RAG account with a knowledge box for your collection. Some features, such as generated briefings, need a Pro platform licence. We can supply licences through BlueChip IT.

You can run it yourself on infrastructure you manage, or use CorpusKit Cloud, where Noice hosts the portal at your own web address and handles upgrades and support.

You choose Australia, Europe or the United States. The index and knowledge graph stay with the documents, and the licence and knowledge box are in your organisation's name.

Figures, populations, named studies and years are tested against the cited passages, sentence by sentence. A sentence that fails is removed and a note says so, and each answer is rated high, moderate or low confidence.

A portal can be public, open to everyone who signs in with your organisation's single sign-on, or limited to people you invite. There are four roles, from viewer to portal administrator, and an exportable audit log.

Yes. The demo at demo.corpuskit.org is a working portal you can question now. To see CorpusKit on your own reports, talk to us about a portal on your collection.