AgentSkillsCN

vertex-search

从 Vertex AI RAG（物理项目）中搜索知识。使用“/vertex-search <查询>”或直接询问“在 Vertex 中搜索”。

SKILL.md
--- frontmatter
name: vertex-search
description: Search knowledge from Vertex AI RAG (phsysics project). Use "/vertex-search <query>" or ask "Search in Vertex"
user-invocable: true

Vertex AI RAG Search Skill

Purpose

Search related knowledge from phsysics project's BigQuery (4,362 embeddings) + GCS (context/, decisions/, session_logs/)

Search Method

  1. •Analyze query and extract keywords
  2. •BigQuery vector similarity search (COSINE_SIMILARITY)
  3. •GCS metadata search
  4. •Integrate and sort results by relevance

Usage Examples

bash
/vertex-search NoiseComputer multiplication rules
/vertex-search previous debate results
/vertex-search RTL optimization patterns

Output Format

  • •Relevance score (0.0-1.0)
  • •Content snippet
  • •Source (BigQuery/GCS)
  • •GitHub link (if available)
  • •Timestamp

Configuration

Edit config/vertex_config.yaml to adjust:

  • •similarity_threshold: Minimum relevance (default: 0.7)
  • •max_results: Maximum results to return (default: 10)

Implementation

Python script: .claude/skills/vertex-search/vertex_search.py

  • •Uses BigQuery client for embedding similarity search
  • •Uses GCS client for metadata search
  • •Combines results with relevance ranking