RAG Configuration Resource

The RAG Configuration Connection Shared Resource specifies the configuration required to connect with an external AI model service and manage Retrieval-Augmented Generation (RAG) workflows. Through this resource, users can select the language model service (such as, OpenAI or Ollama), choose an embedding model to generate vector representations, and configure the vector database used for storing and retrieving embeddings.

General

The General panel contains the following fields.

Property Module Property? Yes/No Description
Package No The name of the package in which you want to create a shared resource.
Name No The name to be displayed as the label for the shared resource in the process.
Description No Provide a short description for this shared resource.

Configuration

The Configuration panel contains the following fields.

Property Module Property? Yes/No Description
LLM Selection  

The following are the supported LLMs:

  • Ollama

  • OpenAI
Ollama Connection
Ollama Endpoint Yes The field indicates the endpoint URL of the Ollama service used to process model requests.
Model Name Yes The field indicates the name of the Ollama model used to generate responses.
Open AI Connection
OpenAI LLM Model Selection Yes The field is to identify the OpenAI Language Model to generate responses from retrieved data.
OpenAI API Key Yes The field indicates the API key used to authenticate with the OpenAI service.
Temperature Yes The field indicates the level of randomness incorporated when generating responses.
Max Tokens Yes The field indicates the maximum number of tokens permitted in the generated response.
Top P Yes The field indicates the cumulative probability threshold that restricts token selection.
Frequency Penality Yes The field indicates the penalty applied to reduce repetition of words or phrases in generated response.
Presence Penality Yes The field indicates the penalty applied to encourage the introduction of new topics in the generated response.
Test Connection   Click the Test Connection button to verify that the configuration can establish a successful connection to the OpenAI service.
Vector Database Configuration
Embedding Model No

The field indicates the embedding model provider for generating vector representations. Select the OpenAI for OpenAI embedding model, or Ollama for Ollama embedding model.
OpenAI Embedding Endpoint Yes The field indicates the URL for a custom OpenAI embedding endpoint. If OpenAI is chosen as the LLM, it automatically applies the default end point when the field is left blank.
OpenAI Embedding Model Name No

The field indicates the OpenAI embedding model used to generate vector representations.

The following are the supported embedding models:

  • text-embedding-3-large

  • text-embedding-3-small

  • text-embedding-ada-002

Embedding Ollama Endpoint Yes The field indicates the endpoint URL of the Ollama service running locally to generate embedding.
Embedding Ollama Model Name No

The field indicates the Ollama embedding model used to generate vector representations.

The following are the supported embedding models:

  • all-mililm

  • nomic-embed-text
Vector Database Storage No

The field indicates the vector database storage that manages the storage and retrieval of embeddings.

The following are the supported storage options:

  • local

  • Weaviate

local

Configure the following field when you select the local database storage.
Storage File Yes

The field indicates the local file system directory for storing vectors.
Note: The above field is enabled only when the Local is selected as the vector database storage.

Weaviate

Configure the following fields when you select the Weaviate database storage.
API Key Yes

The field indicates the API key required for authentication with the Weaviate database.
Host Yes The field indicates the host name or IP address of the Weaviate server.
Port Yes The field indicates the port used to connect to the Weaviate server.
Scheme No

The dropdown indicates the communication scheme for Weaviate connection.

The following are the supported values:

  • HTTP

  • HTTPS
Allow Duplicates No The checkbox indicates if the Weaviate database permits duplicate vectors.
Objectclass Yes The field indicates the object class used to store vectors in Weaviate.
Consistency No

The dropdown indicates the consistency level configured for Weaviate operations. Following are the supported values:

  • QUORUM

  • ONE

  • ALL
Textfield Name Yes The field indicates the name of the text field to store text content in Weaviate.
Metadata Field Name Yes The field indicates the name of the metadata field in Weaviate.
Metadata Keys Yes The field indicates the keys of metadata fields to be stored or retrieved in Weaviate.