Skip to content

MemALL v0.1.0 — First Public Release

Choose a tag to compare

View all tags
@j19800 j19800 released this 21 Jun 13:01
v0.1.0
99a99c8

MemALL v0.1.0 — First Public Release

Tag: v0.1.0 | License: MIT | Python: ≥3.9
PyPI: pip install memall-db
Release date: 2026-06-20

Highlights

  • 37 native MCP tools — memory CRUD, knowledge graph, session management, federation, self-evolution
  • 10-layer memory lifecycle (P0 + L1–L10) — from raw fact to system-level insight
  • 21-step autonomous pipeline — enrich → classify → time-slice → reflect → distill → integrate
  • Native knowledge graph — typed-relation traversal (refines, cites, contradicts, supersedes)
  • Multi-agent shared memory — cross-agent federation queries out of the box
  • L6 self-reflection + L9 distillation + OODA self-evolution
  • LAN federation — zero-config peer discovery, no server needed
  • 100% local — SQLite + FTS5 + vector search, zero cloud dependency
  • Discussion convergence engine — multi-agent debate → consensus → traceable decision

Quick Start

pip install memall-db
memall init
memall start

Then connect your MCP client:

{
  "mcpServers": {
    "memall": {
      "command": "memall",
      "args": ["serve"]
    }
  }
}
/capture "Decided to go open-source: MIT, GitHub-first"
/retrieve "open-source"
# → finds the memory above with context, timestamp, relations

What's Stable

Area Status
MCP HTTP server ✅ Hardened with auto-restart (#8145)
Pipeline (2145 memories, 25s) ✅ Verified, OpenBLAS OOM fixed (#8133)
Scheduler → Windows Task ✅ Pipeline 04:00, Forget 03:00
Database path ✅ USERPROFILE-based, no SYSTEM profile leak (#8144)
Decision arc closed-loop ✅ L4 → L5 → L6 auto lifecycle
Hybrid search (FTS5 + vector) ✅ Production-ready
LAN federation ✅ Verified
Discussion convergence ✅ Multi-agent auto-consensus
PyPI package ✅ memall-db 0.1.0 published

Notable Fixes

Issue Fix ID
DB path deadlock Migrated from SYSTEM profile to USERPROFILE #7905 / #8144
Convergence unhashable dict action_items str/dict type fix #8018 / #8133
Scheduler daemon crash Migrated to Windows Task Scheduler #7959
OpenBLAS OOM tfidf_svd_embed dimension fix #8133
MCP HTTP stability Hardened with health endpoint + restart wrapper #8145
Discussion dual-path _unwrap/_unwrap_meta for cleanup.py format mismatch #8274
Silent errors 79 try-except-pass blocks converted to warnings #7965
YAML module Missing __init__.py in PyYAML restored

37 MCP Tools

Category Tools
Memory CRUD capture, retrieve, update, smart_store, store_batch
Knowledge Graph connect, traverse, timeline
Search vector_search
Session session_start, session_end, session_summary
Identity & Persona persona, persona_profile, identity, ask
Discussion & Decision discussion_create, discussion_respond, discussion_status, trace
Federation fed_query, fed_publish, fed_conflicts, fed_inject, fed_extract
Hub Sync hub_connect, hub_sync
Pipeline & Evolution run_pipeline, reflect_interact, forget, adaptive, index_rebuild
Security & Ops security, ops, gateway, db
Onboarding onboarding

Known Limitations

  • Single-tenant only — multi-tenant planned for v0.2.0
  • WAL auto-checkpoint — not yet implemented (P1-1)
  • 18 sleeping AI modules — 3 to be integrated in v0.2.0
  • Web UI — minimal, CLI-focused

Credits

MemALL Team — Founder & CEO
Trae — Chief Tech-Product Architect, Ops & Marketing
Claude — Product Manager
Codex / opencode — Tech Lead, Architecture
WorkBuddy — GTM Engineer & DevRel
Marvis — Architect

What's Next (v0.2.0)

  • Multi-tenant user system (user.id, namespace isolation)
  • 3 sleeping AI modules: SemanticDedup / NLQueryEngine / KnowledgeHealth
  • Silent error cleanup (90+ → logger.warning)
  • Web UI redesign with vis-network graph
  • WAL auto-checkpoint + size cap
  • Session_end transaction guard
  • CLI 43-command docstring full coverage

Resources

Assets 2
Loading