← 返回 Skills 市场
152
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install turboquant
功能描述
Compress conversation context into a TurboQuant vector store inside OpenClaw memory, then retrieve the most relevant entries on demand to stay within token b...
安全使用建议
This skill is coherent with its stated purpose, but check a few practical things before installing: (1) ensure the 'uv' CLI you provide is a trusted implementation (inspect the upstream repo https://github.com/openclaw/openclaw-turboquant if you want to review code), (2) be aware the skill will create and read files under ~/.openclaw/memory/, so avoid placing highly sensitive data there unless you accept local persistence, (3) embeddings must be produced by you and saved as .npy files the tool will read—confirm your workflow for generating and securing those files, and (4) if you need stricter privacy, consider encrypting the memory directory or using ephemeral session IDs so the store does not persist across restarts.
功能分析
Type: OpenClaw Skill
Name: turboquant
Version: 0.1.0
The turboquant skill is a utility for managing conversation context through vector embeddings and local storage. It provides standard commands (ingest, assemble, compact) for RAG-style context management and uses the 'uv' package manager to execute its operations. All data storage is confined to the local directory (~/.openclaw/memory/), and the instructions in SKILL.md are purely functional without any signs of malicious prompt injection, data exfiltration, or unauthorized execution.
能力评估
Purpose & Capability
Name/description (compress conversation context into a compressed vector store) match the instructions: ingest, assemble, compact, and store-info commands all operate on a local store path. The declared dependency ('uv') is the only external requirement and matches the CLI invocations shown.
Instruction Scope
SKILL.md only instructs the agent to run the uv CLI against a local store and local .npy/.npz files, and to persist data under ~/.openclaw/memory/turboquant-$OPENCLAW_SESSION_ID. It does not ask the agent to read unrelated system files, call external web endpoints, or exfiltrate data. It does assume the caller produces embeddings and places them in files the tool will read.
Install Mechanism
There is no install spec (instruction-only), so nothing will be downloaded or written by the skill itself. The only runtime requirement is that a trusted 'uv' binary be available on PATH; the skill also mentions running 'uv sync' once — this is an operational note but not an installer.
Credentials
No environment variables, credentials, or config paths are required by the skill. It references $OPENCLAW_SESSION_ID and $TURN_NUMBER as context-derived names for store paths/ids, which is proportionate to maintaining per-session memory.
Persistence & Privilege
The skill persists a store under the user's home (~/.openclaw/memory/...), which is expected for a memory compression tool. always is false and autonomous invocation is allowed (platform default). Persistence across restarts is a design choice appropriate for the stated purpose but users should be aware data will be stored on disk.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install turboquant - 安装完成后，直接呼叫该 Skill 的名称或使用
/turboquant触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.1.0
Initial release:
- MSE quantization for vector reconstruction
- Product quantization for inner-product estimation
- Context store with ingest/assemble/compact lifecycle
- CLI tools: benchmark, compress, retrieve
元数据
常见问题
TurboQuant Context 是什么？
Compress conversation context into a TurboQuant vector store inside OpenClaw memory, then retrieve the most relevant entries on demand to stay within token b... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 152 次。
如何安装 TurboQuant Context？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install turboquant」即可一键安装，无需额外配置。
TurboQuant Context 是免费的吗？
是的，TurboQuant Context 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
TurboQuant Context 支持哪些平台？
TurboQuant Context 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 TurboQuant Context？
由 Twacqwq（@twacqwq）开发并维护，当前版本 v0.1.0。
推荐 Skills