GitHub ↔ Vertex AI Sync Skill
Purpose
Synchronize knowledge between GitHub docs/brain/ and Vertex AI RAG (BigQuery + GCS)
Sync Directions
1. GitHub → Vertex AI
When files in docs/brain/ change:
- •Generate embeddings using Vertex AI Embeddings API
- •Store in BigQuery for similarity search
- •Backup to GCS
2. Vertex AI → GitHub
When AI debates conclude:
- •Update
docs/brain/DECISIONS.mdwith results - •Commit changes automatically
- •Create PR if configured
Usage
bash
# Bidirectional sync (default) /github-sync # GitHub to Vertex only /github-sync --to-vertex # Vertex to GitHub only /github-sync --from-vertex # Initial full sync /github-sync --initial-sync
What Gets Synced
To Vertex AI:
- •
docs/brain/CONTEXT.md - •
docs/brain/DECISIONS.md - •
docs/brain/DEBATES.md - •All
.mdfiles indocs/brain/
From Vertex AI:
- •Debate results →
docs/brain/DECISIONS.md - •Session logs →
docs/brain/DEBATES.md - •Knowledge updates →
docs/brain/CONTEXT.md
Conflict Resolution
- •Vertex AI is source of truth for decisions
- •GitHub is source of truth for manual edits
- •Timestamp-based merge for conflicts
Implementation
Python script: .claude/skills/github-sync/sync_manager.py
- •Git diff detection for changed files
- •Vertex AI Embeddings API for vectorization
- •BigQuery upload with deduplication
- •GCS backup with versioning