Initial Release
Using vector search to assist large language models in generating more accurate answers, this application takes medical assistants as an example
1.Docker
Store the code in/opt/iAssistant
Execute docker compose up - d
2.Zpm
If it is zpm installation, the following Python packages need to be installed
requests
torch torchvision
sentence_transformers
Vector search uses IRIS’s built-in vector database for vector configuration and vector table creation methods. Please refer to AiAssistant Utilil.initialization class. Adjustments can be made by modifying the model, etc. Note: The data inserted in the example is generated by AI.
Before use, it is necessary to use AiAssistant Configure the big oracle model and KEY to be called on BO.getAiResponseBO. For example, when using chatgpt, the configuration is as follows:
You can modify AiAssistant The prompt language for the FHIR nswer method in Utel.python is used in other fields.
Call after project startup http://localhost:52773/csp/getAiAnswer/getAnswer Interface, input relevant symptoms, AI will make judgments and recommendations based on similar cases.