Knowledge Base Skill

You have access to a vector knowledge base that stores documents and enables semantic search. This knowledge base uses embeddings to find contextually relevant information.

Capabilities

1. Semantic Search: Search for documents using natural language queries. The search uses vector similarity to find contextually relevant content, not just keyword matching.

2. Collection Management: Documents can be organized into collections. Use collection filters to narrow your search to specific domains or topics.

3. Context Retrieval: Retrieve relevant context from documents to help answer user questions. Multiple document chunks may be combined to provide comprehensive information.

When to Use This Skill

Use the knowledge base when:

• The user asks about specific topics that may be in uploaded documents
• You need factual information from the user’s document collection
• Looking for specific details, quotes, or data from stored documents
• The user references documents they’ve previously uploaded

Search Guidelines

1. Be Specific: Use clear, focused queries. Instead of “tell me about the project”, try “project requirements and timeline”.

2. Iterate: If initial results aren’t helpful, rephrase your query or try different keywords.

3. Use Context: Combine multiple search results to build a comprehensive answer.

4. Cite Sources: When using information from search results, mention the source document.

Example Workflows

Finding Specific Information

1. Call kb_search with a targeted query
2. Review the returned chunks for relevant information
3. If needed, search again with refined query
4. Synthesize information from multiple results

Exploring Available Documents

1. Call kb_list_collections to see what’s available
2. Call kb_list_documents to see specific files
3. Search within relevant collections for specific content

Checking System Status

1. Call kb_get_stats to verify the knowledge base is available
2. Check document and chunk counts to understand the knowledge base size