A KnowledgeCollection groups related documents for the AI to search. The `embed_docs` command generates vector embeddings for each document chunk. Semantic search finds relevant content based on meaning — not just exact word matches. Example: the query 'my login is broken' semantically matches 'authentication failure troubleshooting' even though none of the words overlap.
Your Challenge
You're building an internal help desk bot for a 50-person software company. Design a KnowledgeCollection:
1. List at least 4 document types you would include
2. Explain how you would organize them
3. Explain why semantic search (bge-m3 embeddings) is better than keyword search for this use case. Give one concrete example of a query that would work with semantic search but fail with keyword search.