Synaptic Four · Data and research

BioResearch Assistant

Search, collect and question the literature, with a language model on your own computer.

Discuss putting it to use

What it is about

Research produces and reads a lot: publications, protocols, reports. What a group has already read and collected is often scattered and hard to find again. General AI services help with summaries, but they send the content to another provider.

The BioResearch Assistant is a working environment that runs on your institution’s computer or server.
You collect literature in it and ask questions of your collection. A language model writes an answer from the matching stored texts.

What stays on your computer and what goes outside

What is stored stays with you, and the language model runs on the same computer by default. Two paths lead outside. Both are deliberate choices.

Stays with you

  • stored texts in your institution’s database
  • stored files in a folder the operator sets
  • the summary with the model on the same computer, by default

Goes outside

  • every literature search as a query to PubMed
  • the summary, only when a key for a cloud model is set. That call then goes to the chosen provider
The operator chooses the provider, the model and the address. Without a cloud key, the summary stays on the computer. A recording of the network traffic that shows this for every setting does not exist yet.

How you work with it

  1. Find a paper.
    You enter a literature search. The search queries PubMed.
  2. Add it to the library.
    A paper you want to keep, you save in your library. It sits in your institution’s database.
  3. Ask the collection.
    You ask the library a question in ordinary language.
  4. See the papers the answer uses.
    You see the stored papers behind each answer, with the title and how close they are to the question. The screen does not show the passage itself.
  5. Open the original source.
    You open the paper on PubMed.

What it is good for:

  • gaining an overview of a larger collection
  • bringing together statements from several papers
  • getting into a new topic
  • finding again what you have read
Working in a team. Each person first sees only their own entries. A group can see its entries together when the operator allows that. In production, sign-in through the institution’s sign-in system is required.

What you need

  • A language model that fits the machine. The guide names small models for laptops with 8 to 16 GB, medium models for desk computers with 24 to 64 GB, and large models for institute servers with a graphics processor. A graphics card is not required, but it speeds up the answers.
  • A sign-in system for production.
  • A person who takes on installation and updates. The guides describe this for a computer without internet as well.

Status and limits

Early use
Whether every statement in the answer can be traced to a stored source has not been checked yet. Read the sources yourself.
Implemented
  • Search, library, question and answer
  • Sign-in and a person’s own entries
  • Guides for installation and operation, including without internet
Not checked yet
  • a full run from the search to the answer
  • a recording of the network traffic
Not included
  • clinical decisions
No Gaia-X certification and no GDPR certification.

What it is not

It does not make clinical decisions.

An AI answer is not scientific evidence. It is a start for the search, not a substitute for checking the sources yourself.

Cost and licence

The BioResearch Assistant itself is free (Apache 2.0).
  • €3,600

    Team package, up to 10 people, per year

  • €480

    Setup without a subscription, half a day remotely

Included are all updates during the term, one remote setup of up to half a day, and e-mail support on working days with a reply within two working days. Not included are operation and troubleshooting of the server infrastructure, and custom changes.

All offers on this path

Do you want to work through your literature and records with AI, without handing them over?

Describe your sources, the size of your team and your computers.

Discuss putting it to use