Upload your documents. Maya answers from your content in under 50ms. Semantic search that understands what users mean, not just what they typed.
Retrieval-Augmented Generation (RAG) means the AI looks up the answer inside your actual documents before responding. Your policies, your scripts, your product list — that is what gets searched. Nothing invented. Nothing hallucinated.
RAG search loads your documents and data once — products, policies, listings, records, anything — and returns meaning-matched results on every query. Not by keywords. By meaning.
Keyword search works at 100 records. It breaks at 10,000. At 100,000 — users give up and call your support team instead. Your support team is answering search questions. That is not a people problem. It is a search problem.
| Query | Keyword search returns | RAG search returns |
|---|---|---|
| "quiet dishwasher" | Every product containing either word. 200+ results, most irrelevant. | The 3 models rated lowest for operating noise in your catalog. |
| "maternity leave" | Every document containing those words. Staff reads 40 pages to find the answer. | The exact policy clause with number of weeks and eligibility rules. |
| "force majeure construction contract 2023" | 400 documents. Manual review to find what is legally relevant. | The 5 clauses directly legally relevant to the query. |
| "software engineer Bangkok 3 years" | Every job with any of those words. Misses "Developer" or Thai-language listings. | Software engineering roles in Bangkok, ranked by relevance to 3 years' experience — in any language. |
Features
Why it matters
Features
Why it matters
Features
Why it matters
| Metric | Value |
|---|---|
| Records indexed (live) | 4,190 jobs across 13 countries |
| Languages | 30 — query in any, return from any |
| Embedding dimensions | 1,536 |
| Search response time | <50ms — no degradation at millions of records |
| Cost per query | Negligible — embedding is a one-time cost at ingestion |
Staff spends 20 minutes manually looking up answers during live calls. The person is on hold. The staff member is searching. RAG search in co-pilot mode returns the answer on screen in under 2 seconds. The hold time disappears.
People leave your page when search returns the wrong items. They do not call to complain. They go to a competitor whose search works. You never see this in your analytics — it shows as a bounce. The lost opportunity is invisible.
If you are already on any paid plan, the RAG Search Engine on your data is already included. No extra line item. No separate quote.
Extra cost: $0 — infrastructure cost per client is under $20 total. Absorbed.
For organisations that only need document search — no voice, no screening tools.
Customisation is scoped per project. Basic — name, logo, colours, result template. Larger customisations (multi-language UI, custom scoring layers, API integrations) quoted separately.
The technology is industry-agnostic. Only the data changes.
| Industry | What gets searched |
|---|---|
| Healthcare | Appointment protocols, intake procedures, patient FAQs |
| Real Estate | Property listings, area guides, qualification scripts |
| Legal | Contracts, compliance documents, case precedent |
| Financial Services | Tax deadlines, document checklists, compliance obligations |
| Logistics | Driver handbooks, route guides, customs documentation |
| HR / Staffing | Job descriptions, hiring SOPs, policy libraries |
| Retail / E-Commerce | Product catalogs, pricing sheets, return policies |
| Schools | Enrollment guides, course catalogs, training manuals |
RAG search is trained on your data, not general knowledge. Your product catalog, your documents — that is what gets searched. Your terminology is what gets indexed. If it is in your data, it gets found. If it is not, nothing is invented.
Keyword search is a feature. RAG search is a different technology. The test is simple: type a natural language query. Count how many of the top 10 results are actually relevant. Then send us 100 of your own records and we will run real queries against them. The difference is the answer.
License RAG Search as part of Maya's capability for your own organisation. Upload your documents, Maya answers from your content across all configured channels.
Deploy RAG Search as part of configured Maya workflows for multiple client organisations. Each client's documents are tagged to their API key — multi-tenant by design, impossible to query without specifying whose data.
Any format. Any industry. We index your sample, run real queries, and send back results within 48 hours. You see exactly what your users would see — before committing to anything.
Send 100 Records Try It Free First Full Product Page