v0.2.4

GroundMemory gives AI agents structured, searchable memory that persists across sessions - long-term facts, a user profile, agent instructions, an entity graph, and daily logs- all managed automatically via MCP or the Python API.

Highlights

• Zero-setup by default - BM25-only mode runs entirely on SQLite with no API key, no GPU, and no extra dependencies
• Hybrid search - BM25 keyword scoring and vector cosine similarity combined and re-ranked in a single query
• Pluggable embedding providers - local sentence-transformers, any OpenAI-compatible endpoint (OpenAI, Ollama, LM Studio, LiteLLM), or BM25-only
• Relation graph - typed entity triples (Alice → works_at → Acme Corp) stored in SQLite and mirrored to a human-readable RELATIONS.md, with semantic deduplication
• Bootstrap injection - all memory tiers assembled into a compact system-prompt block at session start
• Compaction hooks - structured prompts flush important facts to storage before the context window rolls over
• Workspace isolation - each project, user, or agent gets its own directory-backed workspace
• MCP server - exposes 9 tools and 1 prompt over HTTP; works with Cursor, Cline, Claude Desktop, Windsurf, and any MCP-compatible client
• Python API - drop-in integration for OpenAI and Anthropic agents via MemorySession and ALL_TOOLS
• 316 passing tests