Azure
The Azure
provider includes processors that correspond to models from Azure's OpenAI service.
Azure ChatGPT
Input
prompt
: The prompt to ask the ChatGPT model.search_filters
: (Optional) The search filters to use to retrieve data from a vector store as a string. It is of the formatkey1 == value1 || key2 == value2
orkey1 == value1 && key2 == value2
.
Configuration
datasources
: (Optional) List of datasource UUIDs to use to retrieve data from the vector store for the asked question. If not provided, it will not provide any context to the ChatGPT model.
-
model
: (Optional) ChatGPT model to use for the chat. Default:ChatCompletionsModel.GPT_4
. -
system_message_prefix
: (Optional) Prefix to use for system message to the ChatGPT. Default:None
. -
instructions
: (Optional) Instructions to pass in the messages to the ChatGPT. Default:None
. -
documents_count
: (Optional) Maximum number of chunks of data to retrieve from the vector store for the asked question. Default:10
. -
chat_history_limit
: (Optional) Maximum number of chat history messages to save in a session and pass to the ChatGPT in the next prompt. Default:100
. -
temperature
: (Optional) Temperature to use for the ChatGPT. Default:0.7
. -
use_azure_if_available
: (Optional) Use Azure's ChatGPT models if a key is configured in the settings or in the organization that the user is part of. Default:True
. -
chat_history_in_doc_search
: (Optional) Include chat history in the search query to the vector store. Default:False
. -
show_citations
: (Optional) Cites the sources used to generate the answer. Default:False
. -
citation_instructions
: (Optional) Instructions to pass in the messages to the ChatGPT for citations. This can be used to control how the citations are generated and presented. Default:None
.
Output
answer
: The answer from the ChatGPT model.citations
: (Optional) The list citations for the answer.