A knowledge base assistant for organizations

Answers from your documents — enterprise RAG

The first job of a knowledge base assistant is to speak only from sources you approve. RAQS crawls your site via its sitemap, processes JavaScript-rendered corporate portals and SPA interfaces as they truly render using a headless browser, cleans away menu and footer noise with trafilatura to keep only meaningful text, and respects robots.txt. Crawl scope is bounded by a per-plan page quota, so your content stays auditable.

Content is chunked, embedded with Azure text-embedding-3-small and stored in pgvector. At query time a hybrid retrieval runs: semantic vector search is fused with Postgres full-text keyword search via RRF, and irrelevant results below a similarity threshold are dropped. That captures enterprise jargon, product codes and acronyms both by meaning and as exact terms — queries that need a precise match, like a policy number or part code, don't get lost.

Beyond crawling, you can feed content directly: product catalogs, procedure documents, HR policies. Upload PDF, Word, TXT, MD and CSV files, or paste FAQ text. Without being tied to a single source, you gather scattered enterprise knowledge into one answer layer.

In your brand's voice, with a consistent tone

In corporate communication, consistency is the foundation of trust. RAQS's voice, tone, persona and name are set from the dashboard, so the assistant produces a single, uniform voice that follows your brand style guide — formal or warm, your call. Turkish and English are standard, so a multilingual corporate site serves both audiences at the same quality and from the same knowledge base.

Visitors tap the tap-to-talk orb and speak, or you pick a 3D mascot to be a face for your brand. A low-latency LiveKit realtime transport layer, Azure Speech for speech-to-text and text-to-speech, barge-in (interrupting the assistant mid-sentence) and natural pauses make the exchange feel like a real dialogue, not a form. On the brain side, Azure OpenAI gpt-4o-mini inside a Pipecat pipeline keeps replies limited strictly to the retrieved content.

Security and compliance — RLS, KVKK, GDPR

For larger organizations, security isn't a feature — it's a precondition. RAQS runs on a multi-tenant architecture, and every customer's data is isolated at the database level with row-level security (RLS): one organization's knowledge base is never mixed into another tenant's query. That isolation rests on database policies directly, not on application logic.

The crawler is SSRF-guarded: it visits only the domains you declare and refuses redirects that try to reach internal-network addresses or metadata endpoints. The widget runs only on permitted origins (an allowed-domains allowlist), and Cloudflare Turnstile blocks abuse and bot traffic. Infrastructure is hosted in Azure's EU region — a foundation built for data residency and for KVKK and GDPR compliance.

Because you can review conversations and analytics from the dashboard, it stays auditable what the assistant answered and on which content — leaving a concrete trail for compliance and quality teams.

Not just answers — guiding the visitor

On a corporate site, most questions end with a next step: “where do I fill out this form”, “where's the datasheet for that product”, “take me to the dealership application page”. RAQS sees the page live; it guides the visitor to the right page or product, shows an item from a list, and triggers an on-page action when needed. Sensitive operations (payment, account actions, deletion) require explicit user confirmation; the smart safety layer never runs them without that consent.

This turns a knowledge base assistant from a passive FAQ box into a corporate guide that carries the visitor to the goal — cutting the repetitive routing burden on your support team.

Dashboard, multi-tenant management & one-line setup

RAQS is a managed, hosted cloud service — you don't stand up servers or run models. From the dashboard you add sites, manage the knowledge base (crawl depth, re-crawl, file upload), set the allowed-domains allowlist, choose a mascot/persona, and review conversations and analytics. Keeping content fresh is one click: re-crawl the knowledge base and the assistant updates as your corporate documents change.

Going live is a single <script> line. The backend is Python/FastAPI, the voice pipeline is Pipecat, the site is statically built with Next.js and fronted by Cloudflare — all managed under one domain (raqs.ai). Pricing is pay-as-you-go usage plus a base plan, with a free trial.

<script async src="https://raqs.ai/v1/raqs.js"
  data-raqs="YOUR_SITE_KEY"></script>