root 1a394ddd2b fix: _init_graph 更新 _db_path 修复用户隔离
问题:
- _init_graph 正确解析了用户 DB 路径,但未更新 self._db_path
- web_api.py 的 create_server() 用 backend_server._db_path 创建 graph_db
- 导致所有用户共用一个 CWD/graph_memory.db(回忆图谱页面数据混在一起)
- 同时 jianf 的 config.json 首次创建时空 apikey,未正确写回

修复:
- _init_graph 解析用户路径后,同步更新 self._db_path
- 现在各用户 graph_db 和 config.json 完全隔离
2026-04-28 20:56:01 +08:00
2026-04-14 23:54:46 +08:00

TrulyMEM - TrueHumanMEM

TrulyMEM Logo

📜 License: GNU General Public License v3.0 (GPLv3)

中文: README.md

Give AI self-awareness, plasticity, and a sense of proportion in long-term memoryThe More Human Choice.

License: GPL v3 Python 3.8+


In a Nutshell

TrulyMEM gives memory authority back to the LLM. Using graph database (triplets) instead of the traditional messages array, the LLM autonomously decides what to remember and what to forget.


Quick Start

python trulymem_entry.py    # from source
./dist/TrulyMEM             # packaged binary

First run → TUI login screen → create/sign in → press F2 for API Key → start chatting

📖 Full guide: docs/en/quick_start.md


Features

Feature Description
🧠 Graph Memory Triplet storage, LLM autonomous navigation
🔐 Multi-User Isolated profiles + Admin/User roles
🌐 Web UI Embedded Flask server (thread mode) for knowledge graph browsing
🎮 TUI Textual-based terminal UI with F2 config panel
📦 Single Binary PyInstaller build, Web server embedded

Documentation

Document Content
docs/en/quick_start.md 🔥 Full setup guide (Web, multi-user, building)
docs/en/architecture.md System architecture and design
docs/en/memory.md Memory working mechanism
docs/en/persona.md Persona Graph mechanism
docs/en/api.md Backend API
docs/en/prompts.md Prompt management

Special Thanks

  • Prof. Meiting Wang — Academic guidance
  • 逝水秋生白 — Architecture support
  • anzhitinglan — Testing resource support
  • 崔莉萍老师 — Theoretical guidance
  • Annie — Professional guidance
  • 王梓沣、马悦华、隆梦婷 — Neuroscience theory support

License

GNU General Public License v3.0 (GPLv3)

Description
No description provided
Readme 25 MiB
Languages
Python 63.5%
HTML 33.5%
Shell 1.6%
CSS 0.8%
C++ 0.3%
Other 0.3%