Knowledge Search

Developer

What is Knowledge Search?

The search_knowledge tool performs semantic search (RAG) on your workspace’s knowledge base. Unlike keyword search, it understands the meaning behind your query and returns the most relevant documents, even if the exact words don’t match.

Parameters

Parameter	Required	Description
`query`	Yes	Describe what you are looking for — be specific and descriptive
`top_k`	No	Maximum number of results to return (default: 5)

When to use it

Before implementing a feature — check if there’s existing documentation or decisions
When investigating a bug — search for related procedures or known issues
Before making architectural decisions — find prior discussions or ADRs
When writing documentation — check what already exists to avoid duplication

Tips for better results

Be descriptive, not keyword-based

Semantic search works best with natural language descriptions:

Instead of…	Try…
“auth"	"How does user authentication work in this workspace?"
"deploy"	"What is the deployment process for production releases?"
"API rate limit"	"What are the rate limiting rules for the REST API?”

Adjust top_k for your needs

Use top_k: 3 for focused, high-confidence results
Use top_k: 10 when exploring a broad topic or gathering multiple perspectives

Combine with project context

For best results, load the project context first (it happens automatically), then search the knowledge base. This gives your AI assistant both the high-level product overview and specific detailed information.

What gets indexed?

The knowledge base includes all documents imported into Prodgy:

Product specifications and requirements
Architecture decision records (ADRs)
Playbooks and runbooks
Meeting notes and decisions (captured by Prodgy Assistant)
Technical documentation
Any other documents uploaded to the knowledge base