DocSearch Node - Automate Info Search and Retrieval

The DocSearch Node improves how information is found and processed. It uses advanced searching called Retrieval Augmented Generation (RAG).

By connecting to your Search AI App, the DocSearch Node provides high-quality search results that are relevant to the current context. These results are delivered right inside the tool workflow.

Key Features

• RAG Searching Framework:

  RAG blends traditional retrieval systems (e.g., search, databases) with generative LLM capabilities, ensuring more accurate, relevant, and up-to-date results tailored to your needs.

  RAG enhances generative AI outputs through these key steps:

  • Retrieval & Pre-processing: Powerful search algorithms query external data (e.g., web pages, databases). Retrieved data is pre-processed through tokenization, stemming, and stop-word removal.
  • Grounded Generation: Pre-processed data integrates with the LLM, enriching its context for more accurate, informative, and engaging responses.

• Integration with SearchAI
  • The DocSearch Node connects to a configured Search AI App via Agent Platform’s integration page.
  • It accesses indexed resources from the Sources repository within Search AI to retrieve relevant information.
• Dynamic Input Queries
• Accepts search input either as a variable or static text.
• Facilitates seamless query processing by Search AI, which identifies and extracts/retrieves contextually relevant chunks from the available resources using SearchAI’s Answer Generation, corresponding to the input query and the meta filters provided via the Advanced Search API. Learn more.

• Configurable Meta filters

  • You can configure optional meta filters to narrow the search field with the query.
  • If meta filters are not provided, the query is applied to all the documents uploaded to that connection. Learn more.

• Redirection Capability

  • Includes an option to redirect users to the Search AI App directly from the node.

• Enhanced Search Relevance

  • By prioritizing results using RAG criteria, the node provides precise, complete, and context-aware answers.
  • Improves search relevance by focusing on delivering responses tailored to user intent.

• Contextual Personalization

  • Ensures results are personalized to the query context, enhancing user experience and satisfaction.

• Connectivity with Other Nodes

  • Connect the DocSearch Node inputs and outputs to other nodes for seamless integration and data flow within the Tool Flow Canvas.

Configuration Overview

Configuring the DocSearch Node consists of the following steps:

1. Set up a Search AI application and the information source for Agent Platform integration.
2. Link the Search AI application in Agent Platform.
3. Add and configure the DocSearch node.
4. Test the Flow you have built.

Step 1: Set up Search AI App

The integration of Search AI with Agent Platform involves setting up a Search AI application, configuring it for integration, and modifying Agent Platform to interact with Search AI in response to specific conditions or events. Search AI provides REST APIs that enable seamless interaction with any application.

Follow the detailed steps here to complete the configuration.

Note

To receive answers from Search AI, you must enable the Answer Generation option under the API scopes section.

Step 2: Link the Search AI App in Agent Platform

After fetching the Search AI application credentials, configuring the source, and enabling the channel communication via API, you must link the app in Agent Platform.

Steps to integrate Search AI in Agent Platform

1. Sign in to your Agent Platform account.
2. Navigate to the Settings console.
3. Click Integrations on the left navigation menu.

4. Scroll down to Search AI and click Link an App.

5. In the Search AI window, provide the required information that you copied from the SearchAI application in Step 1.

6. Click Test to test the connection.
7. If the connection is successful, click Confirm.

A success message is displayed and the connection is listed for SearchAI.

If the connection fails with the following message, check and re-enter the correct Search AI app credentials.

Note

Currently, we support connections through Search AI. You must provide "https://platform.kore.ai" for the Search AI URL field.

Step 3: Add and Configure a DocSearch Node

Setting up a DocSearch node in an tool flow involves adding the node at the appropriate location in the flow and configuring various node properties.

Steps to add and configure the node

1. On the Tools tab, click the name of the tool to which you want to add the node. The Tool Flow page is displayed.

2. Click Go to flow to edit the in-development version of the flow.

3. In the flow builder, click DocSearch -> + New DocSearch on the Assets panel. Alternatively, click DocSearch in the bottom panel.

4. To provide a unique name, right-click the node and click Rename since the node is provided a default name.

   

Alternatively, click the node and change the value for Node Name in the configuration panel.

5. Add the input for the Query field to capture the user's search query.

   To accept dynamic inputs, either configure an input variable in the Start node or use the Manage Input Variables feature.

   You can also add plain text for hard-coded inputs like "What does the Mckinsey report forecast for Q3 in 2025 based on historic data taken in the past 3 years."

   
6. Select the Search AI connection you set up in Step 2 on the Agent Platform integration page.

Note

Use search to look up and select the required connection.

To set up a new connection, click + New Connection. This will redirect you to the Agent Platform Integrations page. Follow the steps mentioned here to complete the integration.

7. (Optional) Set Meta filters (click the expansion arrow to access the editor) to define rules that will narrow down the search results. For example, if the sources have multiple files, you can define the specific file names to look up in the meta filters code. Learn more.

Note

• If you do not provide any meta filters, the query is applied to all the documents uploaded to that connection.
• The filters can be a context variable in the flow depending on the builder’s requirement.

Managing Errors

Appropriate canvas-level errors are displayed (in the error log) when you add a new node. You can view and fix the errors by clicking the warning icon.

Step 4: Test the Flow

To run and test the flow, follow the steps below:

1. Click the Run Flow button at the top-right corner of the flow builder.

2. (Optional) Add the value for Input Variable if you have configured it to test the flow in Step 5 of this section. Otherwise, go directly to the next step.

3. Click Generate Output.

The Debug window displays the flow log and execution status of each node on the canvas, starting from the Start node to the End node, along with the result shown in the Output window. Learn more about running the tool flow.

Accessing the Output

The answer (extracted chunks) from the node can be accessed via the context variable (key) in the output path. This key is dynamic and depends on the API response. The format is {{context.steps.<<nodename>>.dynamic path}}. For example, it could be context.steps.DocSearch.response.response.answer in one case or context.steps.DocSearch.response.response.response.answer in another. Check the sample response from SearchAI to find the correct key that holds the content. Use that key in your tool flow. You may also need to extract multiple chunks from different keys.

Related Links

• Manage Nodes in Tool Flow - Add and manage nodes on the canvas.
• Run Tool Flow - Run and test an tool flow in real time.