shield_lock
金大哥
← 返回 Skill 列表
extension
分类: 其它需要 API Key

super-brain-skill

Use this skill when WorkBuddy or another host needs to query the Super Intelligent Brain knowledge base, ingest files or URLs into the knowledge base, check ingest job progress, or inspect accessible knowledge base documents through the backend HTTP APIs under /skill/brain/*.

person作者: user_e3654e34hubcommunity
download下载资源包

Super Brain Skill

Use this skill as a thin entry layer for the super_intelligent_brain backend service.

Keep the division of responsibility stable:

  1. Use this skill to decide which user action is needed.
  2. Call the backend /skill/brain/* APIs through scripts/super_brain_client.py.
  3. Let the backend own retrieval, ingest, job orchestration, auth, and storage details.

Default workflow

  1. If the user asks a question about existing enterprise knowledge, call ask.
  2. If the user wants to upload or register new material, call ingest.
  3. If ingest is asynchronous, return job_id and then call get_job when asked for progress.
  4. If the user wants to inspect accessible libraries or documents, call list_kbs, list_docs, or get_doc.

Scope selection

Prefer shared when the knowledge is team-level and already exists in the shared Milvus collection.

Prefer private when the data belongs only to one user or when the user explicitly asks for a personal library.

When using shared, send team_id if the host can provide it, so the request lands on the correct team-scoped library.

Answering rules

If the backend returns a knowledge-grounded answer with citations, present that answer directly.

If the backend returns a fallback LLM answer and the notice says the knowledge base had no hits, make that clear to the user instead of pretending it came from the knowledge base.

If the host requires knowledge-only answers, set strict_kb=true in ask.

Configuration

The Python client reads these environment variables:

  • SUPER_BRAIN_BASE_URL
  • SUPER_BRAIN_SKILL_AUTH_HEADER
  • SUPER_BRAIN_SKILL_AUTH_CODE
  • SUPER_BRAIN_TIMEOUT_SECS

See references/install-config.md for install-time config, references/api-usage.md for the request mapping, and references/examples.md for concrete examples.