← 返回 Skills 市场
jerv

Edgebric

作者 Jeremy Venegas · GitHub ↗ · v0.9.6 · MIT-0
cross-platform ✓ 安全检测通过
141
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install edgebric
功能描述
Search and manage your private knowledge base. Find documents, query knowledge, upload files, and manage data sources in Edgebric.
使用说明 (SKILL.md)

Edgebric Knowledge Base

Use this skill when the user asks about documents, files, knowledge, policies, notes, records, or any information that might be stored in their private knowledge base. Also use it when they want to save, upload, or organize documents.

Setup

Two environment variables are required:

  • EDGEBRIC_URL: The base URL of the Edgebric instance (e.g. http://localhost:3001 or https://edgebric.company.com:3001)
  • EDGEBRIC_API_KEY: An API key starting with eb_, created in Edgebric Settings > API Keys

All requests use Authorization: Bearer $EDGEBRIC_API_KEY header.

When to Use

  • User asks about documents, files, knowledge, policies, procedures, notes, records
  • User says "check my docs", "what do we know about...", "find the policy for..."
  • User wants to save/upload a document ("save this to my Edgebric", "upload this file")
  • User asks to create a new knowledge source or manage their data

Endpoints

Discovery

GET $EDGEBRIC_URL/api/v1/discover

Returns API version, available sources, capabilities, and endpoint map. Call this first if unsure what sources exist.

List Sources

GET $EDGEBRIC_URL/api/v1/sources

Returns all data sources the API key has access to, with document counts.

List Documents in a Source

GET $EDGEBRIC_URL/api/v1/sources/{sourceId}/documents

Returns documents with name, type, size, upload date, and processing status.

Search (Preferred)

POST $EDGEBRIC_URL/api/v1/search
Content-Type: application/json

{
  "query": "what is the vacation policy?",
  "sourceIds": ["optional-source-id"],
  "topK": 5
}

Returns ranked chunks with citations. Prefer this over /query -- it returns raw search results without LLM synthesis, letting you (the smart model) do the synthesis with full context of the conversation.

Response:

{
  "results": [
    {
      "content": "The vacation policy allows...",
      "relevanceScore": 0.92,
      "citation": {
        "documentName": "HR Handbook.pdf",
        "page": 12,
        "section": "Benefits > Time Off",
        "sourceId": "abc-123",
        "sourceName": "HR Documents"
      }
    }
  ]
}

Query (Full RAG)

POST $EDGEBRIC_URL/api/v1/query
Content-Type: application/json

{
  "query": "what is the vacation policy?",
  "sourceIds": ["optional-source-id"],
  "stream": false
}

Returns a synthesized answer from the local LLM with citations. Use this only when you specifically need the local model's interpretation, or when /search returns results and you want a pre-formatted answer.

Response:

{
  "answer": "According to the HR Handbook...",
  "citations": [
    {
      "documentName": "HR Handbook.pdf",
      "page": 12,
      "section": "Benefits > Time Off",
      "sourceId": "abc-123",
      "sourceName": "HR Documents"
    }
  ]
}

Set stream: true for SSE streaming (Server-Sent Events).

Create a Source

POST $EDGEBRIC_URL/api/v1/sources
Content-Type: application/json

{
  "name": "Project Alpha Docs",
  "description": "Documentation for Project Alpha"
}

Requires read-write or admin permission.

Upload a Document

POST $EDGEBRIC_URL/api/v1/sources/{sourceId}/upload
Content-Type: multipart/form-data

file: \x3Cbinary file data>

Supported types: PDF, DOCX, TXT, MD (max 50MB). Returns document ID and job ID. The document will be processed asynchronously (text extraction, chunking, embedding, PII detection).

Requires read-write or admin permission.

Check Job Status

GET $EDGEBRIC_URL/api/v1/jobs/{jobId}

Check if a document upload/ingestion job is complete.

Delete a Document

DELETE $EDGEBRIC_URL/api/v1/documents/{documentId}

Requires read-write or admin permission. Always confirm with the user before deleting. Never delete documents without explicit user approval.

Delete a Source

DELETE $EDGEBRIC_URL/api/v1/sources/{sourceId}

Deletes the source and ALL its documents. Requires admin permission. Always confirm with the user before deleting a source — this is a destructive, irreversible operation. Never delete sources without explicit user approval.

Formatting Results

When presenting search results or query answers to the user, always include citations:

According to HR Handbook.pdf (p. 12, Benefits > Time Off), the vacation policy allows...

Format: Document Name (p. Page, Section Path)

If multiple sources contribute to an answer, cite each one.

Error Handling

All errors return JSON:

{
  "error": "Human-readable message",
  "code": "MACHINE_CODE",
  "status": 401
}

Common codes:

  • AUTH_REQUIRED (401): Missing or invalid API key
  • INVALID_KEY (401): Key is revoked or malformed
  • INSUFFICIENT_PERMISSION (403): Key doesn't have required permission level
  • NOT_FOUND (404): Resource doesn't exist or is outside key's scope
  • RATE_LIMITED (429): Too many requests, check Retry-After header
  • INFERENCE_UNAVAILABLE (503): Local LLM not running

Tips

  • /search is almost always better than /query -- you can synthesize better answers with more context
  • Check /discover first to see what sources are available
  • Source IDs filter searches to specific collections -- omit to search everything
  • The API key's scope may limit which sources are visible
  • Documents take a few seconds to process after upload (extraction + embedding)
  • Works with both localhost and network URLs
安全使用建议
This skill appears internally consistent with a client for an Edgebric instance. Before installing: 1) Verify EDGEBRIC_URL points to a trusted host (your company or localhost) — an API key sent to an untrusted URL would expose your data. 2) Create and use the least-privileged EDGEBRIC_API_KEY possible (read-only if you only need searches; read-write/admin only for uploads/deletes). 3) Be cautious when granting upload/delete permissions and confirm destructive actions as recommended. 4) Note the registry metadata rendering showed 'Required env vars: [object Object]' — confirm the two env vars listed in SKILL.md (EDGEBRIC_URL and EDGEBRIC_API_KEY) are the actual required settings in your environment. If the skill had required unrelated credentials, accessed other files/paths, or included an installer that downloaded code from an unknown URL, the assessment would become suspicious; none of those issues are present here.
功能分析
Type: OpenClaw Skill Name: edgebric Version: 0.9.6 The 'edgebric' skill is a standard API integration for a private knowledge base system. It defines RESTful endpoints for searching, uploading, and managing documents via user-provided credentials (EDGEBRIC_API_KEY) and a user-defined instance URL (EDGEBRIC_URL). The SKILL.md file contains appropriate safety warnings regarding destructive actions like deletion and follows standard patterns for AI agent tool definitions without any signs of malicious intent, obfuscation, or data exfiltration.
能力评估
Purpose & Capability
Name/description match the declared env vars and endpoints in SKILL.md. The two required environment variables (EDGEBRIC_URL and EDGEBRIC_API_KEY) are exactly what a client for a self-hosted Edgebric API would legitimately need.
Instruction Scope
SKILL.md only instructs the agent to call the Edgebric HTTP API endpoints (discover, sources, search, query, upload, delete, jobs). It does not ask the agent to read unrelated files, system state, or other environment variables, and it explicitly requires user confirmation for destructive actions (deletes).
Install Mechanism
This is an instruction-only skill with no install spec and no code files, so nothing is written to disk or downloaded during install.
Credentials
The skill requires only EDGEBRIC_URL and EDGEBRIC_API_KEY. Both are directly relevant: the URL targets the service and the API key is used for Authorization. The primary_credential declaration matches the SKILL.md.
Persistence & Privilege
The skill is not marked always:true and does not request any special persistent privileges. model-invocable is true (normal for skills that can be used autonomously) and there is no evidence the skill modifies other skills or system-wide settings.
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 edgebric
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /edgebric 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
v0.9.6
- Added metadata fields: version, author, homepage, repository, and license to SKILL.md. - Updated environment variable definitions to include field descriptions, required status, and scope in SKILL.md. - No code or functional changes—documentation metadata only.
v0.9.54
- Added primary_credential: EDGEBRIC_API_KEY to metadata. - Document and source deletion sections now explicitly require user confirmation before deleting, with strong warnings to never delete without explicit user approval. - No code changes; documentation update only.
v0.9.1
Initial release — search, query, and manage private knowledge bases
元数据
Slug edgebric
版本 0.9.6
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

Edgebric 是什么?

Search and manage your private knowledge base. Find documents, query knowledge, upload files, and manage data sources in Edgebric. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 141 次。

如何安装 Edgebric?

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

Edgebric 是免费的吗?

是的,Edgebric 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Edgebric 支持哪些平台?

Edgebric 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 Edgebric?

由 Jeremy Venegas(@jerv)开发并维护,当前版本 v0.9.6。

💬 留言讨论