RAG as a service

Retrieval,
minus the plumbing.

Upload files or raw text from the dashboard. Rx handles chunking, embedding and indexing — then your app retrieves with one SDK call.

Get started $ pip install rx-sdk
quickstart.py
from rx import Rx

rv = Rx()  # reads RX_API_KEY

# index a file — or raw text
rv.add("annual-report.pdf")

# query it
for hit in rv.search("revenue growth"):
    print(hit)

HOW IT WORKS

Three steps. No pipeline to build.

01

Upload

Drop files or paste text in the dashboard — or push from code with rv.add().

02

Index

Rx chunks, embeds and indexes automatically. Gemini embedding models — on your own API key.

03

Query

Semantic, BM25 or hybrid search from the SDK — with optional cross-encoder reranking.

THE SDK

Full control when you want it.

Hybrid search

Vector and BM25 scores, merged. Inspect both on every hit.

Cross-encoder reranker

One flag to rerank top candidates for sharper precision.

Filters & scoping

Restrict by document ids or metadata. Page numbers come back with every result.

Multilingual

Arabic and English out of the box — index "القاهرة عاصمة مصر", query it either way.

search.py python
hits = rv.search(
    "payment terms",
    top_k=3,
    search_type="hybrid",      # semantic | bm25 | hybrid
    use_reranker=True,        # cross-encoder rerank
    model="Space-v1",
    metadata_filter={"topic": "contracts"},
)

for r in hits:
    print(r.score, r.scores.vector, r.scores.bm25)
    print(r.document, r.page_number, r.metadata)
    print(r.text[:70])

PRICING

Simple plans, in EGP.

Starter

250EGP / month
500 MB storage 10 documents 1,000 API calls / month 1 member Hybrid search included
Start with Starter

Custom

Talk to us
Custom storage & limits Unlimited members Dedicated support We're flexible — tell us what you need
Contact on Telegram

FAQ

Questions

Which embedding models are supported?

Gemini embedding models — you bring your own Gemini API key, so embedding runs on your quota. Space-v1 is available for query embedding.

What can I upload?

Files or raw text, from the dashboard or the SDK. Chunking and embedding are handled for you.

How does hybrid search work?

Every hit is scored by vector similarity and BM25, then merged. Optionally, a cross-encoder reranks the top candidates.

Does Rx support Arabic?

Yes — index and query in Arabic, English, or both. The dashboard ships with an Arabic (RTL) interface.

What if I outgrow the limits?

Upgrade to Pro, or message us on Telegram — custom plans are easy.