Home / Projects / NiPlex-Harness / Glossary

Pictures

Glossary

Terms used in these pages, explained in plain words.

Kaizen Manas

Terms, one at a time

agent

The model plus the harness, acting through tools.

Allow Session

Approve similar non-HIGH calls until reset or restart.

allow-list

The list of user ids that may use the bot.

approval gate

The check that stops risky actions until the owner presses a button.

embeddings

Numbers that capture the meaning of text.

fallback

The next model tried when one fails.

gateway

The part that talks to Telegram or Discord.

handoff

The package given to the power model.

hard block

A rule that refuses even when the owner would allow it.

harness

The code around a language model: chat adapters, tools, safety checks and memory.

kaizen

A Japanese word for steady, continuous improvement.

manas

Sanskrit for mind or consciousness. Also the agent's short name.

MCP

Model Context Protocol, a way to connect tools. Not wired in here.

owner

The person on the allow-list who chats with the bot.

placeholder

A name like {=NAME=} standing in for a secret.

process group

A set of related processes that can be stopped together.

provider

The service that serves the model.

rollback

Returning to the last good commit.

sandbox

A limited place to run code.

session

One chat's history and state.

skill

A Markdown file of instructions the agent can load.

SSRF

Tricking a server into requesting private addresses.

supervisor

The process that runs the bot and watches for updates.

tier

A model level: small, medium or power.

tool

A function the model can ask the harness to run.

tool group

A set of tools switched on or off together.

vault folder

The folder of Markdown notes you can open in Obsidian.

venv

A Python virtual environment.

Checked against the source at commit f4fd89f on 29 Oct 2026. Everything here is merged on the main branch; no release has been tagged yet. The source is private, so these pages are the public description.