Search AI Assistant
An AI-powered search router that understands a request, selects the most suitable approved website, generates an optimized query, and opens the search automatically.
Tech Stack
Overview
Search AI Assistant is an intelligent search-routing tool that removes the need to manually decide which website is best for a particular request.
The user describes what they want to find in natural language. The assistant analyzes the request, selects the most suitable platform from a list of approved websites, creates an optimized search query, and opens the results automatically.
For example, if the user enters:
I want to build a cardboard car. Show me how.
The assistant recognizes that the user needs a visual tutorial, selects YouTube, generates a more useful search such as “how to build a cardboard car,” and opens the search results.
Why I Built It
Search AI Assistant is being developed as a future component of Alpha, my personal AI-assistant project.
Alpha is eventually intended to recognize when a user wants to search for something, pass the request to this system, and allow it to select the best source and formulate the search automatically.
Building the search system separately makes it easier to test and improve before integrating it into Alpha.
How It Is Different
A conventional search engine requires the user to select a platform and formulate the query manually.
Search AI Assistant makes that decision for the user. It considers the intention behind the request and routes it to an appropriate approved source.
Examples include:
- YouTube for tutorials and demonstrations
- Wikipedia for general reference information
- Major search engines for broader or less specific requests
- Other approved platforms suited to particular types of content
General-purpose search engines are used when a more specialized source would not provide a better result.
Current Version
Version 1.0 is complete and supports basic AI-directed searches.
The current system can:
- Accept a natural-language request
- Analyze the user’s search intent
- Select an appropriate approved website
- Generate a relevant search query
- Open the search in the user’s browser
It currently uses Ollama for request analysis, JSON for configuration and stored information, and Python’s webbrowser module to open the selected destination.
Current Limitations
Version 1.0 routes the user to a generated search-results page. It does not yet inspect the individual results or select the best result from the page.
For example, it can generate and open a YouTube search, but it cannot yet choose the most relevant video.
Planned Improvements
The next version will move beyond generating searches and begin selecting individual results.
Planned improvements include:
- Selecting the most relevant YouTube video
- Finding the most relevant Wikipedia article directly
- Evaluating results before opening them
- Supporting additional approved websites
- Improving intent recognition
- Integrating the system into Alpha
Long-Term Role in Alpha
When integrated into Alpha, this project will act as its search-routing system.
Alpha will recognize that the user needs information, generate a clear search request, send it to the search assistant, and direct the user to the most relevant source. Eventually, the system should retrieve and evaluate information without requiring the user to manually navigate between platforms.