---

name: rag-ingestion-v1 description: Document ingestion pipeline - docs to chunks to metadata for RAG

RAG Ingestion Skill v1.0

Purpose: Ingest documents into the knowledge base for AI retrieval.

1) Non-Negotiables (Policy)

Document Sources

Copycontent/                    # Markdown content files
docs/                       # Documentation
ANDARA_*.md                 # System documentation
AI_LEARNING_*.md            # Learning sessions

Database Tables

• knowledge_base - Main knowledge entries
• rag_memory_objects - Chunked content for RAG
• documents - Ingested documents

---

2) Inputs & Outputs

Ingestion Input

Copyinterface IngestRequest {
  source: string;           // File path
  type: 'markdown' | 'pdf' | 'chat';
  category: string;
  metadata?: Record<string, any>;
}

Chunk Output

Copyinterface RAGChunk {
  id: string;
  content: string;
  embedding?: number[];
  metadata: {
    source: string;
    category: string;
    timestamp: Date;
  };
}

---

3) Quality Gates

Before Ingestion

• Source file exists and readable
• Category defined
• No duplicate content

After Ingestion

• Chunks created in database
• Searchable via RAG API
• Metadata accurate

---

Ingestion Scripts

---

Ingestion Workflow

Single Document

Copytsx scripts/ingest-individual-sources.ts path/to/doc.md

Batch Ingestion

Copy./scripts/master-ingest.sh

Hub Pages to RAG

Copynpm run sync:hubs

---

RAG Query API

Endpoint

POST /api/ai/rag/query

Request

Copy{
  "query": "What is structured water?",
  "limit": 5
}

Response

Copy{
  "results": [
    {
      "content": "Structured water is...",
      "source": "water-science.md",
      "score": 0.95
    }
  ]
}