Skip to main content

​
The problem

The answer is spread across dozens of long PDFs, papers, or reports. Reading them all to answer one question doesn’t scale.

​
How Ragrails solves it

Ingest the documents from local files, URLs, or bytes, then ask questions that pull from across the whole set.
  1. Parse papers and reports (mix local paths, file URLs, and bytes).
  2. Chat to synthesize, with sources for every claim.
from ragrails import RagRails

rag = RagRails()

rag.ingest(
    docs=[
        "files/reports/q1.pdf",
        "https://example.com/research/whitepaper.pdf",   # remote file URL
    ],
    embedding={"provider": "voyage", "model": "voyage-3"},
    storage={"vector_db": "qdrant", "collection": "research", "url": "http://localhost:6333"},
)

llm = rag.llm(provider="openai", model="gpt-4o")   # stronger model for synthesis
embedder = rag.embedder(provider="voyage", model="voyage-3", input_type="query")

result = rag.chat(
    "What do these reports agree on about retention?",
    llm=llm, embedder=embedder,
    vector_db="qdrant", collection="research", url="http://localhost:6333",
    history=[],
)
print(result.answer)

​
Features used

Document ingestion · Chat · Reranking