Skip to main content

Knowledge Base

Description

Use the Knowledge Base step to retrieve your proprietary organization specific information by connecting to the AI Studio Knowledge Management (KM) service. This step analyzes the user's prompt to locate matching content within your configured documents. By restricting the scope to your uploaded data, the step ensures that responses remain grounded and isolated from external web sources. Refer to AI Studio Documentation for more details about AI Studio Knowledge Management (KM) service. Example using Knowledge base step: The process enables a customer service chatbot to answer policy-related questions quickly and accurately using internal knowledge sources. The process performs the following actions:

  1. The chatbot receives a customer query, such as What is my coverage for water damage under policy #789?
  2. It searches internal insurance documents using the Knowledge Base plugin step.
  3. The chatbot identifies relevant policy clauses and provides a clear, accurate answer to the customer or branch representative.
  4. All responses are generated from internal knowledge only, without using external web data. The process effectively handled customer queries by retrieving accurate information from internal policy documents, ensuring precise and reliable responses without relying on external sources.

Configurations

No.Field NameDescription
1Step NameSpecify a unique name for the step.
The name helps you identify the step in the workflow and makes it easier to debug or link it with other steps.

Project tab

Use the tab to configure the connection details and search parameters for the AI Studio Knowledge Management service.

No.Field NameDescription
1Project IDSpecify the unique identification string for the project created in AI Studio. This value must exactly match the ID assigned in the AI Studio platform.
The field accepts variable or static value.
2Project SecretSpecify the secret key associated with the Project ID. This key authenticates the connection between the workflow and the AI Studio project.
The field accepts variable or static value.
3Test ConnectionClick Test Connection to verify that the entered Project ID and Project Key are valid and that a successful connection is established with the AI Studio KM project. This action provides confirmation before proceeding and helps ensure the system is ready, including validating connectivity prior to loading tags
4Search TypeSelect how the step searches for relevant documents in AI Studio before sending data to the LLM.
Vectors — Searches based on meaning. The system converts text into embeddings and finds documents with similar meaning, even if the words are different.
Hybrid — Searches using both meaning and exact keywords. This option improves accuracy when the query includes specific terms like error codes or unique keywords.
Example:
Query: How to fix login error?
Keyword search finds documents with the exact words fix, login, and error.
Vector search finds documents related to sign-in issues using similar meaning.
Hybrid search combines both to return the best results.
5Top KDefine the maximum number of matching results to retrieve. This setting determines how many of the highest-scoring document chunks are passed to the model. For example, a value of 5 returns the top five matches.
The field accepts variable or static value.
6ThresholdSpecify the score threshold.
The threshold is the level above which the prediction of final intent(s) will be returned.
Threshold is the minimum similarity score a document must have to be considered relevant.
If the score is below the threshold, the retriever ignores the matched data (reference).
If the score is above the threshold, the retriever returns the matched data (reference).
For example, if the Threshold is 0.25, then it returns the references for vector of the score is above 0.25.
The field accepts variable or static value.
Note: Threshold is only applicable for Search Type - Vector
Default value: 0.25
7TagSelect a tag to restrict the search scope. When a tag is applied, the step searches only within the documents associated with that specific tag.
Click Load Tags to populate the list with available options from AI Studio.

Prompt tab

No.Field NameDescription
1PromptSpecify the input string used to query the Knowledge Base. This content serves as the search criteria for locating and retrieving matching document chunks. You can enter a static value or use the variable picker to map data from a previous step, allowing the step to dynamically process specific user requests. This field is required.
The field accepts variable/static values and field from previous step. The field is mandatory.

Output tab

No.Field NameDescription
1Reference Output FieldSpecify the name of the variable that will store the search results. This variable captures the retrieved document references, allowing you to access and use this data as input in subsequent workflow steps.
Default value: references