Skip to content

Assistants

Assistants are Large Language Model (LLM) integrations with a focus on a particular task, e. g., for translating text or answering questions from integrated knowledge.

Access to Assistants

Find assistants at https://assistant.llm.scads.ai. Click on the assistant you want to use and submit your query. If you are interested, click on the link below to find out how the particular assistant works.

How to Create your own Assistant?

You can also create your own assistant, i. e., a system that uses texts from your research project or documentation of your system in order to answer questions and queries from users. To build your own assistant/question answering system, please send an email to scadscat@tu-dresden.de with the following information:

Subject

TUD:AI Assistant

Body

Hi team,

Could you please give me an API key to build a question answering system?

I work on [put_your_topic_or_interest] in [put_name_of_your_group_leader]'s group
and want to create an assistant that answers questions about [put_the_assistant_purpose].

Best regards,

When we receive your request, we send you credentials for our assistant builder, so that you can set up your assistant via your web browser.

Hints:

  • You can create several assistants, each for a specific purpose. Send an individual email for each assistant.
  • You can create public or hidden assistants. You can switch visibility at any time.
  • Several features can be enabled for an assistant and the "How does it work" page reflects the chosen configuration.
  • You can use the REST API to manage your assistant or automatically update your assistant. Updating assistants automatically currently requires an MkDocs-based documentation with the search feature enabled.

REST API

Automatically generated documentation is available on https://assistant.llm.scads.ai/docs.

HPC integration

On our high performance clusters, you can use the script ask_assistant.sh to send a query to the HPC Buddy:

$ ask_assistant.sh
Your request to HPC-Buddy: What compiler flag switches optimization on?

Asking HPC-Buddy...

Answer:

To switch optimization on, you can use the **-O** flag, followed by a number that specifies the level of optimization. The levels of optimization are:
* **-O0**: no optimization
* **-O1**: low optimization
* **-O2**: moderate optimization, where most optimization algorithms are applied
* **-O3**: aggressive optimization

As stated in the [compiler flags section](https://compendium.hpc.tu-dresden.de/software/compilers/#compiler-flags) and the [software development overview section](https://compendium.hpc.tu-dresden.de/software/software_development_overview/#software-development-and-tools), using the **-O2** flag will turn on a moderate level of optimization.

Additionally, other flags like **-Ofast** and **-fast** can also be used to maximize performance, but they imply a couple of other flags and may have specific effects on the behavior of floating-point exceptions.

It's also worth noting that the specific effects and available optimization levels may vary depending on the compiler being used, so it's a good idea to refer to the [specific documentation of your compiler of choice](https://compendium.hpc.tu-dresden.de/software/compilers/#compiler-flags) for more detailed information.

Sources:
https://compendium.hpc.tu-dresden.de/software/compilers/#compiler-flags
https://compendium.hpc.tu-dresden.de/software/software_development_overview/#software-development-and-tools

Your request to HPC-Buddy: