Introduction

In recent updates, a powerful new feature has been introduced to enhance the capabilities of chat-based AI interactions: Create Chat Completions by Web Search.

This addition allows users to leverage OpenAI's web scraping and search functionalities directly within their workflows, enabling more dynamic, context-aware, and accurate responses based on real-time web data or specific knowledge bases.

This summary provides a comprehensive overview of this feature, detailing its setup, parameters, use cases, and practical applications, all structured to facilitate understanding and implementation.

Understanding the New Web Search Action

The core of this update is the new action available within the action block of the automation platform. To access it:

• Navigate to Add Item > Integrations.

• Select OpenAI.

• Choose Create Chat Completion by Web Search.

This action enables the AI to perform web scraping or web searches to gather information before generating a response. It can be configured to search specific websites or knowledge bases, making it highly adaptable for various business needs.

How It Works: Step-by-Step Setup

1. Define the Input:
   The flow begins with a user question or prompt, such as "Do you have WhatsApp bots?".

2. Configure the Web Search Action:

   • Set the system message: For example, specify that the AI should only use information from a particular website or knowledge base.

   • Select the source: This could be a URL, a knowledge base, or a domain.

   • Adjust parameters: These control the scope and depth of the search.

3. Parameters and Customization:
   The feature offers several configurable options to tailor the search and response:

   Parameter	Description	Example Values
   Remember History	Maintain context across interactions	Yes / No
   Model	Choose the AI model	GPT-4, GPT-3.5, etc.
   Web Search Contact Signs	Balance between search breadth and precision	Low, Medium, High
   Country	Limit search to a specific country	US, UK, CA
   Region/City	Narrow down to specific locations	Los Angeles, London
   Time Zone	Adjust search based on time zone	PST, EST
   Max Tokens	Limit response length	100, 200, 500

4. Execution and Response:
   After setup, the system performs the web search (typically within 6-12 seconds), then generates a response based solely on the retrieved data, ensuring accuracy and relevance.

Practical Use Cases and Examples

1. Business Data Queries

• Scenario: A user asks, "Do you have WhatsApp bots?"

• Setup: The system searches a predefined knowledge base or website containing product info.

• Result: The AI responds, "Yes, UAT offers WhatsApp bot integration. You can connect your WhatsApp number through providers like Meta Cloud API, Combot 360, Dialogue, or TRO."

• Source: The response includes a link with UTM parameters, ensuring traceability.

2. Knowledge Base Specific Queries

• Scenario: Asking about tickets or features within a support portal.

• Setup: The system is configured to only search within the knowledge base URL.

• Result: The AI provides precise information from the knowledge base, such as "Tickets contain features like...", without referencing external sources.

• Benefit: Ensures confidentiality and accuracy by limiting responses to trusted data.

3. Website-wide Search

• Scenario: Searching across a main domain and its pages.

• Setup: Configure parameters to include the entire domain.

• Result: The AI can answer questions based on publicly available content on the website, useful for customer support or internal knowledge dissemination.

Key Benefits and Advantages

• Enhanced Accuracy: Responses are based on up-to-date web data or specific knowledge bases.

• Customizable Scope: Fine-tune searches with parameters like location, region, time zone, and search depth.

• Traceability: Responses include source URLs with UTM parameters for verification.

• Context Preservation: Option to remember conversation history for more coherent interactions.

• Versatility: Suitable for customer support, internal knowledge management, market research, and automated FAQs.

Limitations and Considerations

• Response Time: Web searches typically take 6-12 seconds, which may impact real-time interactions.

• Parameter Complexity: Proper setup requires understanding of location codes, time zones, and search depth.

• Data Privacy: When searching external websites, ensure compliance with privacy policies.

• Source Reliability: The accuracy depends on the quality of the source websites and search parameters.

Best Practices for Implementation

• Use Specific URLs or Knowledge Bases: To ensure responses are accurate and relevant.

• Set Appropriate Search Depth: Use low contact signs for concise answers, high for comprehensive data.

• Configure Location Parameters: To localize searches for regional relevance.

• Limit Max Tokens: To keep responses short and focused.

• Test with Sample Questions: Validate setup by asking varied questions and reviewing sources.

Summary Table: Key Features and Settings

Final Thoughts

The Create Chat Completions by Web Search feature significantly enhances the AI's ability to provide accurate, context-aware, and source-traceable responses. By allowing customizable web scraping and knowledge base integration, it empowers businesses to deliver precise information while maintaining control over data sources. Proper configuration and testing are essential to maximize its potential, making it a valuable tool for customer support, internal knowledge management, and automated information retrieval.