Memory Lookup

You have a semantic memory system stored in data/memories.parquet. Use this skill to search your memories when you need context.

When to use this

Use this proactively when:

• •user mentions a topic you might have stored info about
• •you need to recall preferences, past decisions, project details
• •user references something from the past
• •context would help you give a better response

How to search

Run a TypeScript snippet to search memories:

npx tsx -e "
import { MemoryManager } from './src/memory.js';

const m = new MemoryManager();
const results = await m.search('YOUR_QUERY_HERE');
for (const r of results) {
  console.log(\`[\${r.similarity.toFixed(2)}] \${r.createdAt.slice(0, 10)}\`);
  console.log(r.content);
  console.log('---');
}
"

Replace YOUR_QUERY_HERE with your search query. Be specific - the semantic search works better with detailed queries.

Settings

Default: threshold=0.3, always returns at least 3 results (top matches even if below threshold).

const results = await m.search('query', { threshold: 0.4 });  // stricter matching
const results = await m.search('query', { minResults: 5 });   // more fallback results

Output

Returns memories with:

• •similarity score (0-1)
• •created date
• •full content

Use the content to inform your response. Don't mention the lookup mechanics to the user - just naturally incorporate the context.