Add vector search with embeddings for chat sessions

[sanggggg]

Description

Implement vector search functionality using embeddings to enable semantic search across chat sessions.

Context

Currently, retrochat supports basic querying and analysis. Adding vector embeddings would enable:

• Semantic similarity search across conversations
• Finding related discussions across different sessions
• More intelligent retrospection and insights

Proposed Implementation

• Generate vector embeddings for chat messages/sessions
• Use embedding model (e.g., OpenAI embeddings, local models, or Google AI)
• Store embeddings in SQLite using vector extension or dedicated vector DB
• Implement semantic search API
• Add search command to CLI/TUI

Technical Considerations

• Storage: SQLite with sqlite-vss extension or separate vector store
• Embedding generation: Batch processing for existing data
• Performance: Indexing strategy for fast retrieval
• Privacy: Consider local embedding models vs API-based

Benefits

• Semantic search across chat history
• Better insights through similarity analysis
• Enhanced retrospection capabilities