GenAIScope v0.4.0

GenAIScope v0.4.0 - Universal Memory Access

Local-first AI memory now speaks every protocol: MCP, REST, OpenAI, Anthropic, Gemini, and vector search - all zero-dependency by default.

---

What's new

Semantic & Hybrid Search

• Three search modes: keyword (v0.3 behaviour), vector (cosine similarity), hybrid (fused score - default)
• Pluggable embedding backends: LocalHashEmbedder (zero-dep, pure Python), SentenceTransformerEmbedder, OpenAIEmbedder
• LocalVectorStore (SQLite-backed) and RedisVectorStore - drop-in swappable
• MemorySearchResult now carries vector_score, keyword_score, fused_score, embedder_name
• Semantic cache upgraded to embedding cosine similarity (threshold 0.92, deterministic fallback retained)

memory.context() - Injectable Context Block

ctx = store.context("coding style preferences", max_chars=2000)
system_prompt = ctx.text  # ready to inject into any LLM call