All plugins

excel_line

Excel-backed hierarchical long-term memory tree (10-row cap per workbook, dynamic branching, leaf pointers). Classifier defaults to the host model; third-party keyless rotation is opt-in.

kalice-viMemoryUpdated 27 days ago

excel_line

Excel-backed long-term memory with a hierarchical tree store. Each .xlsx
holds at most 10 rows (ID | Title | Content | Tags | Branch | Updated); a row
whose Branch ends in .xlsx is a node pointing to a child workbook, a
non-.xlsx branch is a leaf asset, and an empty branch is a plain memory.

Harness-agnostic core

excel_line_core/ is pure Python + openpyxl with zero agent-runtime
imports. It can be pip-installed and used from any harness:

pip install excel-line   # → import excel_line_core
from excel_line_core import BrainStore
store = BrainStore("./brain")
rid = store.add("brain.xlsx", title="note", content="...", tags="x")
print(store.tree())

Hermes plugin

The Hermes entry point in __init__.py wraps the core; store.py,
brain_store.py and worker.py are thin shims kept for backward
compatibility and still re-export the core API.

Privacy & model egress

The optional memory classifier can call an LLM to summarize agent I/O logs
into knowledge rows. By default it uses the host's own configured model
(preferred = "host") — no transcript excerpts leave your machine or account.

Rotation through third-party keyless models is strictly opt-in: run
/excel-line model auto (or pick a specific free model by number) to enable
it. Revert any time with /excel-line model host. Until you opt in, nothing
is sent to any external endpoint.

Brain viewer server

The bundled brain-map HTTP server (scripts/brain_server.py, port 8766) is
not started automatically. Start it manually when needed:

python scripts/brain_server.py   # serves http://127.0.0.1:8766

README from the reviewed commit a48cc5a; it updates when the author re-pins.