# Dive Coder RAG Enterprise

> Enterprise RAG v2 (offline-first): incremental indexing, BM25 + hybrid hooks, GraphRAG/RAPTOR/CRAG adapters, eval + EvidencePack.

- Skill: `duclm1x1/dive-coder-rag-enterprise` (Agent Skill)
- Install (CLI): `npx skillmds@latest add duclm1x1/dive-coder-rag-enterprise`
- Raw SKILL.md: https://api.skillmd.com/api/skills/duclm1x1/dive-coder-rag-enterprise/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: duclm1x1 (https://skillmd.com/u/duclm1x1)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/duclm1x1/dive-coder-rag-enterprise

---


## What this skill does

- RAG v2 core: incremental ingest + hash cache, BM25 lexical retrieval, heuristic multi-query + step-back, overlap rerank, max context cap
- Advanced adapters (skeletons): GraphRAG / RAPTOR / CRAG pipelines (offline-first defaults)
- Governance outputs: report JSON, claims ledger (E2), EvidencePack (E3 bundle)

## CLI usage (repo root)

```bash
export PYTHONPATH="$PWD/.shared/vibe-coder-v13:${PYTHONPATH:-}"

# Build / update KB (skips unchanged docs)
python3 .shared/vibe-coder-v13/vibe.py rag ingest \
  --spec .vibe/inputs/v13/rag_spec.yml \
  --kb .vibe/kb

# Query
python3 .shared/vibe-coder-v13/vibe.py rag query \
  --kb .vibe/kb \
  --q "your question" \
  --max-context-chars 24000

# Eval (report + claims + evidencepack)
python3 .shared/vibe-coder-v13/vibe.py rag eval \
  --kb .vibe/kb \
  --eval .vibe/inputs/v13/rag_eval.yml \
  --out .vibe/reports/rag-eval.json \
  --evidencepack-out .vibe/evidencepacks
```

## Where to extend

- Dense embeddings + hybrid fusion: `.shared/vibe-coder-v13/rag/adapters/embedding/*`
- Cross-encoder / LLM rerank: `.shared/vibe-coder-v13/rag/adapters/rerank/*`
- GraphRAG/RAPTOR/CRAG: `.shared/vibe-coder-v13/rag/advanced/*`