Intelligent Search

Our documentation site features a powerful, AI-driven intelligent search engine. It goes beyond simple keyword matching to understand the semantic meaning of your queries, providing highly relevant results from across our entire knowledge base.

How It Works

The intelligent search system is a combination of a frontend React component and a backend API endpoint.

1. Frontend: The Search Interface (VectorSearch.tsx)

When you type in the search bar, you are interacting with the VectorSearch React component.

• Debounced Input: It waits for you to finish typing before sending a request to the backend, ensuring a smooth user experience.
• Contextual Filtering: You can refine your search by userRole, businessContext, and technicalLevel to get results tailored to your needs.
• Rich Results: The search results display not just the title and a snippet, but also rich metadata like relevance score, source, and tags.

2. Backend: The Brains (intelligent-search.ts)

Your search query is sent to the /api/intelligent-search endpoint, which orchestrates the search process:

1. Query Reception: The API receives your query and any contextual filters.
2. Initialization: It initializes our core docsCrewImprover library, which contains the logic for interacting with the vector database.
3. Intelligent Search Execution: It calls docsCrewImprover.intelligentSearch(), passing along your query and context.
4. Vector Similarity Search: The library converts your query into a vector embedding and performs a similarity search against the millions of vectors in our ChromaDB instance.
5. Claude.ai Reranking & Enhancement (Future): The raw results are then reranked and enhanced by a Claude.ai model to improve relevance based on business logic and contextual understanding.
6. Response Transformation: The final, ranked results are formatted into a user-friendly structure and sent back to the frontend for display.

3. The DocsCrewImprover Library

This is the core library that bridges the gap between our documentation and the AI.

• Vector Store Connection: It manages the connection to the ChromaDB vector store.
• Embedding Generation: It uses OpenAI's text-embedding-3-large model to convert both documents and user queries into vectors.
• Search Logic: It contains the core functions for performing semantic similarity searches and retrieving documents.

Using the Search

To get the most out of our intelligent search, consider the following tips:

• Use Natural Language: Ask questions just as you would to a person. Instead of "API rate limit", try "What is the rate limit for API requests?".
• Leverage Filters: If you're a developer, set your technicalLevel to 'expert' to get more technical results. If you're on the business team, use the businessContext filter.
• Explore Metadata: The metadata provided with each search result can help you understand why a particular document was returned and how relevant it might be.

This system embodies our Retrieval-First Architecture, a core principle of our Claude.ai integration, ensuring that you always get the most accurate and relevant information from our existing knowledge base before any new content is generated.