Add PRD for Vector Similarity Search Feature

[murrayju]

Summary

Added a comprehensive Product Requirements Document (PRD) for the next major feature: Vector Similarity Search. This feature will enable semantic search and retrieval of memories, transforming Tiger Memory from a simple key-value store into an intelligent memory system.

What's Included

• Problem Statement: Current limitations with scope-only queries (no semantic search)
• Vision & Goals: Transform into intelligent memory system supporting 10k+ memories per scope
• Technical Specification: New search API tool with pgvector and OpenAI embeddings
• Implementation Plan: Phased rollout over 3+ months with clear deliverables
• User Stories: 4 core scenarios for the feature
• Testing & Metrics: Success criteria and monitoring strategy
• Risk Mitigation: 5 key risks with mitigation strategies
• Future Enhancements: Phase 2-4 improvements (local embeddings, re-ranking, clustering)

Expected Impact

• 10-100x improvement in query efficiency for large memory sets
• Scope scalability: 100-1000 → 10k+ memories per scope
• Context efficiency: 50% reduction in context tokens per query
• Enterprise-ready: Production-grade semantic memory system

Architecture Highlights

• Async embedding: Non-blocking memory creation with background embedding
• pgvector integration: Native PostgreSQL vector similarity search
• OpenAI embeddings: text-embedding-3-small (MVP), extensible to other providers
• Backward compatible: Existing APIs unchanged, new search tool added
• Configurable: Can be disabled, supports cost control

Key Components

1. EmbeddingService: Abstract interface for embedding providers
2. SearchFactory: New tool for semantic search
3. EmbeddingQueue: Async background processing
4. Database Schema: New memory_embeddings table with pgvector
5. Migrations: Automated schema updates with rollback support

See PRD.md for complete details, timelines, and success metrics.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Hermes Agent seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}