Upload your documents, configure your bot once, and get a production-ready assistant — embedded on any website and locked inside your own infrastructure. Runs entirely free on local AI models out of the box, and upgrades to premium cloud LLMs & embedding models with a single key whenever accuracy and speed demand it.
Everything needed to run client-ready AI assistants as a service — ingestion, retrieval, guardrails, theming, and one-line deployment.
Configure persona, tone, welcome message, theme colors and do-not-answer rules once. Template variables like {{company_name}} flow through every prompt automatically.
Drop in PDF, DOCX, XLSX, CSV, HTML or web URLs. Scanned files are OCR'd, tables become self-describing rows, and every document can carry custom context the bot respects.
Deterministic row-level reranking puts exact matching table lines in front of the model — precise price lookups instead of plausible guesses, even on dense spreadsheets.
Every bot starts free on local Ollama models. Premium tier: paste a key for OpenAI, Groq, OpenRouter, Gemini or premium embedding models — instant accuracy and speed gains, encrypted keys, zero code changes.
Floating widget, inline iframe, or standalone page — themed with client brand colors, icons and bubble text. One script tag deploys it anywhere.
The bot collects visitor details (name, city, email…) naturally in conversation or via a pre-chat form, remembers them across visits, personalizes every answer — and hands you a leads list with CSV export.
Topic blacklists, fallback messages, and per-bot source visibility (show / filenames-only / hidden) keep every deployment compliant and on-brand.
Drag-and-drop documents or point at web pages. The pipeline parses, OCRs where needed, chunks intelligently, and embeds everything into a local vector store — nothing uploaded anywhere.
The setup wizard captures the bot's identity, tone, boundaries and template variables. The LLM even drafts document descriptions and interpretation tips for you to review.
Chat with the bot instantly on localhost. Ask pricing questions, edge cases and restricted topics — refine until answers are exactly right, then preview it on a mock client site.
A Docker bundle carries app, vectors and config to any client server or cloud VM. Swap the host in one script tag and go live — same behavior, their infrastructure.
Every deployment starts on free local models — private, offline-capable, zero running cost. The moment a workload demands more accuracy or speed, upgrade to premium models without changing a single line of your setup.
Same platform, same features core — only capacity and model power change. Mix and match per client.
No proprietary runtime, no license servers, no vendor lock-in. Every layer is inspectable and replaceable.
Spin up the studio, drop in your first document, and talk to your data before your coffee cools.
Create Free Account Preview the Widget