Implement Content-based Search in SQLiteLTMStore

[csmangum]

Currently, the SQLiteLTMStore lacks implementation for content-based search functionality, which causes failures when trying to use search_by_content() method in the memory system. This causes the memory retrieval demo to fail or require error handling workarounds.

Current Behavior

When memory_system.search_by_content() is called with a SQLite backend, an AttributeError is thrown because the method is not implemented in the SQLiteLTMStore class.

Expected Behavior

The SQLiteLTMStore should implement content-based search capability to allow searching through stored memories based on text content, with proper embedding support.

Proposed Solution

Implement a search_by_content method in the SQLiteLTMStore class that:

1. Accepts a search query and optional parameters (k, threshold)
2. Leverages text embeddings to find semantically similar content
3. Returns memories ranked by relevance

Implementation Details

Suggested approach:

• Add a new column for embeddings in the SQLite schema
• When storing memories, calculate and store embeddings for the content
• Implement vector similarity search (cosine similarity) within SQLite
• Consider using SQLite extensions like sqlite-vss or implement a basic vector similarity calculation
• Return results sorted by similarity score

Related Components

• memory/stores/sqlite_ltm_store.py - Needs to be modified
• memory/memory_system.py - May need updates for consistent interface

Acceptance Criteria

• search_by_content() works with SQLite backend without errors
• Relevant test cases are added
• The memory retrieval demo runs successfully with content-based search
• Performance is acceptable for reasonably sized memory stores

Additional Notes

This would complete the full feature parity between different storage backends and make the memory system more robust.