RAG Chatbot Integration

This section will cover how to connect your Knowledge Base with a Retrieval-Augmented Generation (RAG) based chatbot for interactive AI-powered Q&A and automation.

Create DI Application #

Click the Create DI App button on Document Intelligence page.

Application Name: Provide a unique name to identify your application.

Tags: Add relevant key words to categorize and organize your application.

Click OK to proceed.

Data Source (Step 1) #

Connect Content Source: Configure source to collect the input data.

Select Data Source Type #

Select the type of data source to collect the input data. Available options are: Knowledge Base and Vector DB.

Provide below details upon selecting Knowledge Base as Source Type.

Knowledge Base: Select the knowledge base from the drop down list.

Description #

Option to Auto Generate the description is available.

Add More Knowledge Sources #

You can also add more knowledge sources. You can also edit or delete the description.

Provide below details upon selecting Vector DB as Source Type.

Source Meta Fields #

Select the metadata fields from the source table to be used for generating embeddings.

Provider: Select Pinecone as provider. Select or Add connection. Provide Index and Namespace.

Embedding Provider: Select OpenAI or Voyage AI provider. Select or Add connection. Provide Model and Dimension.

Instructions (Step 2) #

Configure source to collect the input data.

Tone #

The distance metric to be used for similarity search. You can use euclidean, cosine or dotproduct. Aditionally, specify the dimensions of the vectors to be inserted into the index.

Format #

Defines the similarity function (euclidean, cosine, or dot product) used in Pinecone for vector comparison. Must correspond with the embedding model’s vector dimension.

Purpose #

Controls how retrieved documents are ranked based on their closeness to the user query in vector space.

Out of Scope #

Define the boundaries of the application here, so the system can understand which queries to handle and classify others as out of scope.

AI Providers (Step 3) #

Configure source to collect the input data.

AI Provider #

Advance Configuration: Select the type of Vector DB to be used for storing and retrieving vector data.

ReRanking: If checked, then provide the below fields:

Knowledge Base Name: The name of Knowledge Base selected.

Reranker Provider: Select OpenAI or Anthropic.

Connection: Select or Add a new connection.

Model: Select Model.

TopK: The TopK values to be inserted in the index.

Configuration: Click Configuration to provide further configuration details.

Additional Prompt: Add extra instructions to customize the summary’s style, tone, or focus. Use this to guide the AI on what aspects to emphasize or avoid. Example: “For technical readers”, “Make it brief”, or “Highlight key benifits.”

Max Token: The value used to module the next token probabilities.

Click OK to change settings.

Large Chunk Handling: Select the type of Vector DB to be used for storing and retrieving vector data.

Large Chunk Handling: Select either Summarize or Truncate option.

Text Summarization Provider: Select OpenAI, Anthropic, AzureOpenAI.

Connection: Select connection or Add new connection.

Model: Select the model.

Configuration: Click Configuration to provide further configuration details.

Additional Prompt: Add extra instructions to customize the summary’s style, tone, or focus. Use this to guide the AI on what aspects to emphasize or avoid. Example: “For technical readers”, “Make it brief”, or “Highlight key benifits.”

Max Token: The value used to module the next token probabilities.

Click OK to change settings.