Skip to main content

Dynamic Tool Descriptions for MCP Server

Overview​

The RAG-lite MCP server features dynamic tool descriptions that automatically detect and advertise the actual capabilities of each database instance. This enables AI assistants to intelligently route queries to the appropriate database when running multiple MCP server instances.

What It Does​

Dynamic tool descriptions provide self-documenting MCP servers where each instance automatically:

  1. Detects its database mode (text-only or multimodal) from stored configuration
  2. Analyzes content to determine what types of documents are indexed
  3. Generates descriptive labels that clearly indicate capabilities
  4. Advertises specific features available for that database

This creates a seamless experience where AI assistants can make informed decisions about which database to query based on the user's request.

How Tool Descriptions Adapt​

Text-Only Database​

When connected to a text-only database, the tool description clearly indicates:

[TEXT MODE] Search indexed documents using semantic similarity. 
This database contains 150 text documents. Supports .md and .txt files only.

Multimodal Database with Images​

When connected to a multimodal database, the description advertises additional capabilities:

[MULTIMODAL MODE] Search indexed documents using semantic similarity. 
This database contains 200 documents. Contains both text and image content. 
Image results include base64-encoded data for display. 
Supports cross-modal search (text queries can find images).

Uninitialized Database​

For databases that haven't been populated yet:

Search indexed documents using semantic similarity. 
Database not initialized - ingest documents first.

Key Benefits​

Intelligent Query Routing​

AI assistants can automatically select the appropriate database based on the query:

  • Image searches route to multimodal databases
  • Text searches route to text-only databases
  • No manual database selection needed

Self-Documenting Servers​

Each MCP server instance clearly communicates:

  • What mode it's running in (text/multimodal)
  • How many documents are indexed
  • What content types are available
  • What features are supported

Seamless Multi-Instance Support​

Run multiple databases simultaneously without confusion:

  • Each server advertises its unique capabilities
  • AI assistants understand what each database contains
  • Users get relevant results without specifying which database to use

Usage Example​

Setting Up Multiple Instances​

Configure multiple MCP server instances in your MCP client (e.g., Claude Desktop):

{
  "mcpServers": {
    "rag-lite-text-docs": {
      "command": "npx",
      "args": ["rag-lite-mcp"],
      "env": {
        "RAG_DB_FILE": "./text-docs/db.sqlite",
        "RAG_INDEX_FILE": "./text-docs/index.bin"
      }
    },
    "rag-lite-multimodal-images": {
      "command": "npx",
      "args": ["rag-lite-mcp"],
      "env": {
        "RAG_DB_FILE": "./mixed-content/db.sqlite",
        "RAG_INDEX_FILE": "./mixed-content/index.bin"
      }
    }
  }
}

Intelligent Routing in Action​

Scenario 1: Image Search

User: "Find images of architecture diagrams"

What Happens:

  1. AI assistant sees both servers' tool descriptions
  2. Recognizes rag-lite-multimodal-images has [MULTIMODAL MODE] with image support
  3. Automatically routes query to the multimodal database
  4. Returns relevant architecture diagram images

Scenario 2: Text Search

User: "Search the API documentation"

What Happens:

  1. AI assistant sees both servers' tool descriptions
  2. Recognizes rag-lite-text-docs has [TEXT MODE] with text documents
  3. Automatically routes query to the text-only database
  4. Returns relevant API documentation

Use Cases​

Separating Content by Type​

  • Text database: Documentation, articles, code files
  • Multimodal database: Technical diagrams, screenshots, visual content
  • AI automatically chooses the right database based on query intent

Organizing by Project​

  • Project A database: All content related to Project A
  • Project B database: All content related to Project B
  • Shared docs database: Common documentation
  • AI routes queries to the appropriate project database

Domain-Specific Knowledge Bases​

  • Engineering docs: Technical specifications and diagrams
  • Marketing content: Product images and descriptions
  • Support articles: Help documentation and tutorials
  • AI selects the relevant domain based on the question

Getting Started​

To enable dynamic tool descriptions:

  1. Set up your databases - Ingest content into separate databases with appropriate modes
  2. Configure multiple MCP servers - Point each instance to a different database
  3. Start using - AI assistants will automatically route queries based on tool descriptions

No additional configuration needed - the feature works automatically once you have multiple MCP server instances configured.

Summary​

Dynamic tool descriptions enable intelligent, automatic routing of queries across multiple RAG-lite databases. By clearly advertising each database's capabilities, AI assistants can seamlessly select the right knowledge base for each query, creating a powerful and intuitive multi-database search experience.