This custom step allows users to do the following:
- Submit single or multiple queries to a large language model;
- Test prompts (prompt engineering) and view results in the output table node;
- Save prompt to the prompt history and prompt catalog tables;
- Promote prompt history and prompt catalog tables.
The following large language models are supported:
- Offline Models (e.g. Llama-2-7b)
- Online Models (e.g. OpenAI API)
Output table options can be specified to WORK.RESPONSE.
- A SAS Viya 4 environment version 2023.12 or later.
Follow the instructions on Using a Custom Step from this GitHub repository in SAS Studio to download custom steps from this repository and make them available for use in your SAS Studio environment.
- Uses hardcoded SAS libraries (anemoi) and CAS libraries (public) to store results and save history information.
To understand more about Prompt Engineering, see this Prompt Engineering Guide.
- Xin Ru Lee ([email protected])
- Version 1.0 (02FEB2024)
- Initial version