Ten capabilities, each real and shipped, running on one shared pipeline: a dashboard for the account owner, an embeddable widget for visitors, and a server-to-server API for developers.
Hybrid dense + full-text retrieval (RRF-ranked) over the account's own library, with a semantic answer cache so repeat questions cost nothing.
Why it matters: Answers stay accurate because they're pulled from your material, not the model's general knowledge — and the cache keeps common questions cheap.
PDF, DOCX, TXT, and images (PNG/JPG/WEBP/TIFF/BMP). Scanned and image-only pages fall back to OCR; diagrams, charts and photos are described by a vision model so they're searchable too.
Why it matters: Real business libraries are messy — scanned policies, screenshotted tables, photographed forms. Nothing needs to be re-typed first.
A draggable chat bubble via one script tag, or a plain iframe for CSP-locked sites. Branded per widget: display name, accent color, watermark logo.
Why it matters: Works whether your site is WordPress, a strict CSP-locked enterprise portal, or something custom-built.
Secret API keys, session tokens for browsers/mobile apps, idempotent message sends, streaming responses — for teams building their own chat UI.
Why it matters: If you already have a support UI or a mobile app, you don't have to bolt on a widget — call the same pipeline directly.
A drag-and-drop canvas for the pre-chat sequence — greeting, qualifying questions, branching logic, lead capture — before handing off to free-form AI chat. Includes a completion funnel and drop-off analytics per node.
Why it matters: A chat box alone doesn't qualify leads. This turns the first thirty seconds of a conversation into a structured, measurable funnel.
Contact form always available in the widget footer, custom qualifying-answer fields attached automatically, CSV export, email sharing, and a live in-app notification bell — this is a lightweight CRM inbox, not just a table.
Why it matters: Leads show up ready to act on, with the context of what the visitor already answered — no digging through chat logs.
Rules that shape what the bot will and won't say, scoped globally or per collection — plus an assistant that reads the account's own documents and proposes rules with quoted evidence, never auto-applied.
Why it matters: You stay in control of tone and boundaries, and get a head start from suggestions grounded in your own content instead of guessing at rules cold.
Group documents into named subsets so one account can publish multiple bots — a public FAQ widget and an internal-only one — each seeing only what it should.
Why it matters: One account, multiple audiences — a public support widget doesn't need to see your internal runbooks.
Per-widget privacy policy link, a required consent checkbox, and a permanent snapshot of exactly what each visitor agreed to — aimed at India's DPDP Act.
Why it matters: Consent handling is native to the product, not something you have to bolt on separately to stay compliant.
A prepaid wallet with a live balance, a per-model/per-widget/per-API-key cost breakdown, and a monthly subscription with a visible document-upload cap — no surprise invoice.
Why it matters: You can watch the balance drain in real time instead of waiting for a metered bill at the end of the month.
Upload a policy PDF and try the pre-chat flow builder yourself.