# Mind Mem Development > MIND-Mem Python development guide Use when this capability is needed. - Skill: `tomevault-io/mind-mem-development` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/mind-mem-development` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/mind-mem-development/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/mind-mem-development --- # MIND-Mem Development ## Package - PyPI: `pip install mind-mem` - Source layout: `src/mind_mem/` - Tests: pytest, 3610 passing - CI: GitHub Actions, 3 OS × 4 Python versions (16 matrix jobs) ## Architecture - BM25F retrieval with Porter stemming + RM3 query expansion - Hybrid BM25+Vector search with RRF fusion (sqlite-vec) - A-MEM block metadata evolution - 9-type intent router with adaptive confidence weights - ConnectionManager: thread-safe SQLite pool with WAL - BlockStore protocol: decoupled block access - Delta-based snapshot rollback ## MCP Server (81 tools) Grouped surfaces (full list in `docs/api-reference.md` and `src/mind_mem/mcp_server.py`): recall, hybrid_search, prefetch, propose_update, approve_apply, rollback_proposal, scan, list_contradictions, reindex, index_stats, create_snapshot, list_snapshots, restore_snapshot, briefing, category_summary, cross_encoder_rerank, find_similar, memory_evolution, delete_memory, export_memory, import_memory, intent_classify, retrieval_diagnostics, get_mind_kernel, list_mind_kernels, verify_chain, audit_replay, tier_decay_apply, encrypt_status, alerts_subscribe, and more (57 total). ## Key Files - `src/mind_mem/mcp_server.py` — MCP server (57 @mcp.tool entries) - `src/mind_mem/retrieval/` — search engine (bm25, vector, hybrid) - `src/mind_mem/governance/` — contradiction detection, drift analysis - `src/mind_mem/blocks/` — block store, parser, evolution - `tests/` — comprehensive test suite ## Conventions - Python 3.10+, type hints everywhere - Docstrings: Google style - No dynamic allocation in hot paths - All SQL queries parameterized - Config: mind-mem.json (not mem-os.json) - Auth header: X-MindMem-Token --- > Source: [star-ga/mind-mem](https://github.com/star-ga/mind-mem) — distributed by [TomeVault](https://tomevault.io).