drin-vector-database-sync

Vector Database Sync

Purpose

To ensure the AI has access to the most high-fidelity and relevant project knowledge at all times.

Workflow

1. Ingest: Pull new content from source (GitHub, Docs, Internal Wikis).
2. Chunk: Break down content into semantically meaningful pieces.
3. Embed: Generate vectors using the latest embedding model.
4. Index: Update the vector index (Pinecone/Milvus/Chroma).
5. Optimize: Prune old/irrelevant data to keep the context window clean.

Rules

• Every chunk must carry its source metadata (URL/FilePath).
• Use overlapping chunks to preserve context at boundaries.
• Re-embed if the embedding model version changes.

---

subagents:

• data-engineer
• prompt-engineer