← 返回 Skills 市场
chair4ce

BrainDB

作者 ChairForce · GitHub ↗ · v0.5.2
cross-platform ⚠ suspicious
942
总下载
0
收藏
1
当前安装
5
版本数
在 OpenClaw 中安装
/install braindb
功能描述
Persistent, semantic memory for AI agents. Gives your AI long-term recall that survives compaction and session resets — 98% accuracy, 20ms latency.
使用说明 (SKILL.md)

BrainDB

Persistent, semantic memory for AI agents. Built for OpenClaw.


What It Does

Your AI forgets everything between sessions. BrainDB fixes that.

It gives your assistant a memory system that automatically captures important context from conversations and recalls it when relevant — who you are, what you're working on, what you've told it before. Memories persist across compaction, session resets, and restarts.

How it works:

You say something → OpenClaw captures important facts → BrainDB stores them
You ask something → OpenClaw recalls relevant memories → AI has context

No commands. No manual saving. It just works.


Install

Requires Docker and ~4 GB RAM.

openclaw plugin install braindb

Or manually:

git clone https://github.com/Chair4ce/braindb.git ~/.openclaw/plugins/braindb
cd ~/.openclaw/plugins/braindb
bash install.sh

First run: 3–5 minutes (downloads embedding model). After that: ~10 seconds.

What the installer does:

  1. Backs up your existing memory files to ~/.openclaw/braindb-backup/
  2. Builds and starts 3 Docker containers (Neo4j, embedder, gateway)
  3. Patches your OpenClaw config (~/.openclaw/openclaw.json) to enable the plugin
  4. Optionally offers to migrate existing workspace files into BrainDB

Review install.sh before running if you want to understand each step.


What You Get

  • 768-dim semantic search — finds conceptually related memories, not just keyword matches
  • 4 memory types — episodic (events), semantic (facts), procedural (skills), association (links)
  • Tiered ranking — semantic similarity always beats keyword match
  • Auto-dedup — won't store near-duplicate memories
  • Hebbian reinforcement — memories strengthen with use, decay without it
  • Query expansion — understands colloquial phrases
  • 98% recall accuracy on a 50-test benchmark suite
  • 12–20 ms average query latency

Security & Privacy

Core operation is fully local:

  • Gateway binds to localhost only — not exposed to your network
  • Neo4j and embedder are not accessible from the host (isolated Docker network)
  • Neo4j password is auto-generated (24-char random)
  • Optional API key authentication via BRAINDB_API_KEY
  • Containers run as non-root users
  • All embedding, search, and storage runs locally — no external API calls during normal operation

What the installer reads/writes:

  • Reads your OpenClaw config (~/.openclaw/openclaw.json) to add the plugin entry
  • Reads workspace files during optional migration (preview with --scan first)
  • Writes .env with generated Neo4j credentials
  • Creates Docker volumes for persistent storage

Migration privacy notice:

  • Default migration (--no-swarm): Fully local. File contents never leave your machine.
  • Migration with swarm: Sends file contents to Google's Gemini API for intelligent fact extraction. This is opt-in only — you must have swarm installed and explicitly allow it. Use --no-swarm to guarantee local-only processing.
  • Always run node migrate.cjs --scan or --dry-run first to see exactly what would be processed.

Migrating Existing Memories

Already have MEMORY.md, daily notes, or other workspace files? Import them:

node migrate.cjs --scan /path/to/workspace   # Preview files (no data sent anywhere)
node migrate.cjs --dry-run /path/to/workspace # Extract facts locally, don't encode
node migrate.cjs --no-swarm /path/to/workspace # Import, fully local
node migrate.cjs /path/to/workspace           # Import (uses swarm if available)

Your files are never modified. BrainDB copies facts from them — it doesn't replace anything.


Failover

BrainDB fails gracefully:

  1. Gateway down: OpenClaw works normally — the memory block is simply absent from prompts. Your AI still has MEMORY.md and workspace files.
  2. Neo4j down: Gateway returns empty results. No errors, just no memories.
  3. Embedder down: Falls back to text-only search (less accurate but functional).

Your workspace files are the safety net. BrainDB is additive — remove it and you're back to defaults with zero data loss.


Uninstall

openclaw plugin remove braindb

The uninstaller exports all memories (JSON + readable markdown), stops containers, removes the plugin config from OpenClaw, and leaves your workspace files untouched. Docker volumes are preserved until you explicitly delete them.


Performance

Metric Value
Recall accuracy 98% (50-test suite)
Avg latency 12–20 ms
Cold query ~60 ms
Capacity 10K+ memories
Storage ~3 GB
RAM ~2.5 GB

Links


MIT — Oaiken LLC

安全使用建议
Things to check before installing: - Review install.sh and the patch step that updates ~/.openclaw/openclaw.json. The installer will enable autoCapture/autoRecall automatically; if you prefer manual enablement, don't let the installer patch config or back up the config first. - Understand auto-capture: the plugin will automatically record tool execution results, discovered scripts, and workspace file metadata into the DB. That can include sensitive outputs (tokens, paths, command output). If you have secrets in outputs or files, either disable auto-capture or run the plugin in an isolated environment first. - Migration privacy: importing with --swarm will send file contents to Google's Gemini API. This is opt-in, but double-check you never run migrate.cjs without --no-swarm or without understanding the flag. Use --scan or --dry-run to preview what would be ingested. - Confirm local exposure: gateway is configured to bind to 127.0.0.1 only, and Neo4j/embedder are on an internal Docker network. Still verify port mappings and .env (NEO4J_PASSWORD) after install. - If you lack full trust in the release: run the installer in a disposable VM/container first, inspect generated .env and Docker images, and examine stored memories/exported backups before connecting to any production data. - If you want least privilege: decline the automatic config patch, keep BrainDB disabled until you audit encode/auto-capture behavior, and never use --swarm unless you explicitly need and approve external processing.
功能分析
Type: OpenClaw Skill Name: braindb Version: 0.5.2 The skill bundle is classified as suspicious due to a significant lack of transparency regarding external network calls. While the `SKILL.md` and `README.md` claim 'No external API calls during normal operation' and only explicitly mention Google Gemini API usage for the `--swarm` migration option, the `gateway.js` implementation reveals that several core features (`/memory/auto-encode`, `/memory/smart-recall`, and predictive warming) *also* send user conversation data and queries to `https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent` if a `GEMINI_KEY` is configured. This contradiction and under-disclosure of data transmission for 'normal operation' features, even to a legitimate service, constitutes a critical privacy and transparency vulnerability, making the skill suspicious rather than benign. There is no evidence of intentional malicious exfiltration to an unauthorized third party.
能力评估
Purpose & Capability
The name/description (persistent semantic memory) align with the included files: a gateway, embedder, Neo4j DB, migration and auto-capture code. Required binaries (docker, node) are reasonable for the bundled installer. However there are internal inconsistencies: the registry metadata said 'no install spec / no required config paths' while SKILL.md metadata and installer expect to write/patch ~/.openclaw/openclaw.json and create ~/.openclaw/plugins/braindb. The included code also reads many workspace and user files (~/bin/, workspace scripts, skills), which is plausible for a memory/migration tool but broader than the registry's earlier declarations.
Instruction Scope
Runtime instructions and code do more than simple encode/recall. install.sh will (by default) back up workspace files, start Docker containers, and patch OpenClaw's config to enable autoCapture/autoRecall. execution-awareness.js actively enumerates system/tooling (which CLI tools are installed, ~/bin scripts, workspace scripts, other installed skills) and auto-capture middleware (auto-capture.js) will ingest tool execution outputs (errors, command output) and encode them into the DB. That means command outputs, script contents, and discovered metadata can be stored automatically. The migration tool can optionally send file contents to Google's Gemini API if the user opts into '--swarm'. These behaviors are within a memory-plugin's remit but are high-risk for sensitive data exfiltration if not understood/controlled.
Install Mechanism
SKILL.md declares a download/install from a GitHub Releases URL (braindb-v0.5.0.zip) and postInstall runs the included install.sh; GitHub releases is a standard host. The installer builds and runs Docker Compose (building local images, downloading the embedding model). That is expected for this workload but will write files, create containers, and download a ~420MB model. The registry-level summary incorrectly indicated 'instruction-only' (no install spec) which is inconsistent with the packaged install scripts and SKILL.md metadata.
Credentials
The skill requests no external API keys by default, and the gateway binds to localhost. It generates and stores NEO4J_PASSWORD in a .env and will patch your OpenClaw config. However the code reads numerous local paths and tool outputs (workspace files, ~/bin scripts, other skills' SKILL.md, /tmp/fleet-nodes.json), and auto-capture may store outputs of arbitrary tool executions (including errors and command output). Those actions can capture secrets or sensitive data. Migration can optionally transmit file contents to Google Gemini (opt-in via --swarm), which is explicitly documented but must be chosen by the user to expose data externally.
Persistence & Privilege
The installer patches OpenClaw config to enable the plugin and autoCapture by default (writes to ~/.openclaw/openclaw.json), creates Docker volumes and persists data. always:false so it's not force-included globally, but it does gain persistent presence in the agent config and autonomous invocation (autoCapture/autoRecall). That persistence is normal for a plugin but increases blast radius because the skill will run automatically and record data unless you disable it in config.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install braindb
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /braindb 触发
  4. 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
v0.5.2
Fix display name
v0.5.1
v0.5.1: Session-context endpoint for compaction-proof conversational memory
v0.5.0
Execution Awareness: tool/capability memory for AI agents. New encoder introspects environment, new /memory/execution-focus endpoint for targeted tool recall.
v0.4.1
Address ClaWHub security scan: migration defaults local-only, metadata declares permissions, explicit privacy disclosure.
v0.4.0
Initial public release. 98% recall accuracy, 20ms latency, fully local Docker deployment.
元数据
Slug braindb
版本 0.5.2
许可证
累计安装 2
当前安装数 1
历史版本数 5
常见问题

BrainDB 是什么?

Persistent, semantic memory for AI agents. Gives your AI long-term recall that survives compaction and session resets — 98% accuracy, 20ms latency. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 942 次。

如何安装 BrainDB?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install braindb」即可一键安装,无需额外配置。

BrainDB 是免费的吗?

是的,BrainDB 完全免费(开源免费),可自由下载、安装和使用。

BrainDB 支持哪些平台?

BrainDB 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 BrainDB?

由 ChairForce(@chair4ce)开发并维护,当前版本 v0.5.2。

💬 留言讨论