AI Chat Assistant
AI Chat Assistant can be used for Q&A over web or document sources, accomplish tasks by utilizing a set of tools or a general conversation with the AI assistant to achieve a valuable result.
There are currently 3 types of chat interactions, Agent, Retrieval, Default, these types can share a Chat History allowing you combine them into a single conversation by providing each chat a copy of the past messages as context to influence its future responses.
Default Chat
The image above shows various components of the interface that are common between all of the different types of chats.
New Session: Creates a new chat session in history.
Model Selection: Choose from supported foundational and open-source models.
Temperature: Range of tonality from with 0 being generic to more creative responses.
System Message: The most important of the components is used to create a concise and direct set of instructions or personas you would like the agent to exhibit during for its responses.
Chat Input: Where the user will input their queries and queries.
Quick Actions: Coming Soon
Chat Selection: Select from 3 chat types Default, Retrieval, and Agent.
Default: What you would expect from normal Chat GPT
Retrieval: Load documents and hyperlink resource to provide context that influence responses.
Agent: Select a set of tools the agent will utilize to accomplish tasks.
Retrieval Chat
Combine various sources of from pdf, csv, website, or web pages into a single index that can be used as context for a AI assisted conversations over the sources that are indexed.
The image above shows the 2 main components of Retrieval Chat. The ability to choose from a isolated indexes for question and answering chat sessions and the ability to create indexes from hyperlink and document sources.
The image above shows how to create indexes from various source types. You can create a new or choose from an existing index to add sources to. Select from a loader type which will conditionally render the beneath for the selected loader. Depending on the type you can choose from files, or a hyperlink based source.
Agent Chat
The image above shows how to select a set of tools to be used in Agent Chat to utilized to accomplish tasks on behalf of the user. Inputs logs will appear in the top right of each response to inform the user of the actions and parameters the agent uses to call functions from the tool kits.
Last updated