stellarion_memory_search

Finds relevant memories from the store using the same hybrid search engine as symbol search — combining keyword matching (BM25) with semantic similarity. Retrieve past debugging discoveries, architectural decisions, and project knowledge by describing what you're looking for.

When to Use

• You're debugging an issue and want to check if you've seen it before
• You need to recall an architectural decision or its rationale
• You want to find all memories related to a topic or component
• You're onboarding and want to surface accumulated project knowledge

Parameters

Examples

Search for past debugging knowledge

Natural language:

Have we seen CORS errors before? What was the fix?

MCP call:

{
  "tool": "stellarion_memory_search",
  "arguments": {
    "query": "CORS error fix"
  }
}

Search within a category

Natural language:

What architectural decisions have we made about the database?

MCP call:

{
  "tool": "stellarion_memory_search",
  "arguments": {
    "query": "database",
    "category": "architecture"
  }
}

Search by tag and topic

Natural language:

Find all debugging memories related to authentication

MCP call:

{
  "tool": "stellarion_memory_search",
  "arguments": {
    "query": "authentication issues",
    "tags": ["debugging"]
  }
}

Output Format

Returns ranked memory results:

• id — Memory ID
• content — Memory content
• score — Relevance score
• tags — Associated tags
• category — Category label
• problem / solution — If stored as a problem/solution pair
• createdAt — When the memory was stored

Tips

• Describe the problem you're facing in the query — semantic search will match it against stored solutions even if the wording is different.
• Use tags to narrow results when a broad query returns too much. For example, query: "error handling" with tags: ["api"].
• Use minScore to filter out low-relevance matches when you want only strong hits.
• This is a good first step when you encounter a new issue — check if past you already solved it.