AI Search Engine - Ragrails

The problem

Keyword search misses results that use different words for the same idea. “How do I get my money back” may not match a page titled “Refund policy.”

Why Ragrails fits

Ragrails can run retrieval without answer generation. That means you can build semantic search with no LLM generation cost, then add reranking when result order matters.

Need	Ragrails capability
Meaning-based search	Embedding and Retrieval
Better ordering	Reranking
No generated answer	Use `retrieve()` or `query()` instead of `chat()`
Structured result cards	Use `items[].text`, `items[].metadata`, `score`, and `rerank_score`

Sample

result = rag.retrieve(
    "how do I get my money back",
    top_k=20,
    use_rerank=True,
    rerank_top_k=5,
)

for hit in result.items:
    print(hit.rerank_score, hit.metadata.get("title"), hit.text[:120])

ragrails retrieve "how do I get my money back" \
  --vector-db qdrant \
  --collection docs \
  --url http://localhost:6333 \
  --provider voyage \
  --model voyage-3 \
  --top-k 20 \
  --rerank \
  --rerank-top-k 5

curl -X POST http://127.0.0.1:8000/v1/retrieve \
  -H "Content-Type: application/json" \
  -d '{
    "query": "how do I get my money back",
    "provider": "voyage",
    "model": "voyage-3",
    "vector_db": "qdrant",
    "collection": "docs",
    "url": "http://localhost:6333",
    "top_k": 20,
    "use_rerank": true,
    "rerank_top_k": 5
  }'

Result card shape

{
  "title": "Refund policy",
  "snippet": "Refunds are available within 30 days...",
  "score": 0.76,
  "rerank_score": 0.93,
  "source": "policy/refunds.md"
}

Add Chat later when you want generated answers over the same index.