scbkr-memory-index
我把這個專案做成一個可下載、可執行、可擴充的 SCBKR 開放式記憶索引層。
I built this project as a downloadable, runnable, and extensible open SCBKR memory indexing layer.
我不把它包裝成完整治理引擎。
它的工作很明確：先把記憶整理好，再做責任鏈截斷，最後只把可進決策的內容交給下一層使用。
I do not package this as a full governance engine.
Its role is very specific: structure memory first, apply responsibility-chain cutoff, and only pass decision-ready memory to the next layer.
我用 SCBKR（S / C / B / K / R） 把跨來源記憶整理成可：
- 查詢
- 重播
- 稽核
- 路由
的索引結構。
I use SCBKR (S / C / B / K / R) to structure cross-source memory into indexes that are:
- queryable
- replayable
- auditable
- routable
我提供一個 starter-package，讓你下載 ZIP 後就能直接跑起來，不需要先理解整套封閉治理核心。
I provide a starter package so you can run it directly after downloading the ZIP, without needing access to the closed governance core.
我把核心規則寫得很死：
沒有責任鏈，不進決策。
I enforce one hard rule:
No responsibility chain, no decision path.
我沒有把封閉治理核心開源。
I do not open-source the closed governance core.
我沒有把它包裝成完整的企業合規產品。
I do not present this as a full enterprise compliance product.
我沒有做成自動責任判定器。
I do not provide automatic responsibility adjudication.
它不是萬能 AI 殼，也不是聊天玩具。
它是一個開放式索引層。
It is not a universal AI shell or a chatbot toy.
It is an open indexing layer.
cd main-root/starter-package
./run_open_layer.sh
python3 services/scbkr_api_server.py --index ./memory-index/index.scbkr.decision-ready.json --port 9000
這個流程會做三件事：
1. 建立索引
index.scbkr.generated.json
2. 套用責任鏈截斷
index.scbkr.decision-ready.json
3. 啟動本地 API
(/health, /query)
This flow does three things:
1. Build the index
index.scbkr.generated.json
2. Apply responsibility-chain cutoff
index.scbkr.decision-ready.json
3. Start the local API
(/health, /query)
---
開源層工具清單 / Open-layer tools
目前公開 repo 內的工具包括：
tools/auto_index.py
tools/scbkr_human_gate.py
services/scbkr_api_server.py
tools/build_optimized_index.py (experimental)
tools/r_field_recommender.py (experimental)
services/scbkr_llm_bridge.py (experimental)
The current public repo includes:
tools/auto_index.py
tools/scbkr_human_gate.py
services/scbkr_api_server.py
tools/build_optimized_index.py (experimental)
tools/r_field_recommender.py (experimental)
services/scbkr_llm_bridge.py (experimental)
也就是說，現在已經能做的不是空展示，而是：
自動索引
責任鏈截斷
decision-ready 輸出
本地查詢 API
查詢優化骨架
R 欄位候選建議
LLM prompt payload 橋接骨架
In other words, this is not just a concept page.
It already provides:
auto indexing
responsibility-chain cutoff
decision-ready output
local query API
an optimization skeleton
R-field candidate suggestions
an LLM prompt-payload bridge skeleton
---
商業層邊界 / Commercial layer boundary
商業層不放在這個公開 repo 裡。
它包含：
治理權重與參數引擎
模型上層白盒規則
企業審計與導入支援
The commercial layer is not inside this public repo.
It includes:
governance weights and parameter engine
white-box rules above model outputs
enterprise audit and onboarding support
請看：
COMMERCIAL_LAYER_OVERVIEW.md
See:
COMMERCIAL_LAYER_OVERVIEW.md
---
前端展示 / Frontend showcase
cd main-root
python3 -m http.server 8080
首頁的重點不是行銷包裝，而是：
可執行命令
SCBKR 結構
責任鏈邏輯
開源層 / 商業層邊界
The homepage is not designed as marketing fluff.
It is designed to show:
executable commands
the SCBKR structure
responsibility-chain logic
the boundary between the open layer and the commercial layer
---
相關文件 / Related docs
GENERAL_AUDIENCE_GUIDE.md
DEVELOPER_OVERVIEW.md
main-root/starter-package/GETTING_STARTED_IMPROVEMENTS.md
---
一句話總結 / One-line summary
這不是完整治理引擎。
這是一個讓記憶在進入決策前，先被整理、先被切邊界、先被接上責任鏈的開放式索引層。
This is not a full governance engine.
It is an open indexing layer that structures memory, applies boundary control, and connects responsibility before memory enters any decision path.