KnowFlow leverages GROQ Cloud’s hosted models under the hood, giving you access to high‑performance LLMs tuned for technical and code-centric queries. We automatically pick the best model for your workload, and you can upgrade to the most capable models on paid plans.

Out of the box KnowFlow works with any language that appears in your documentation: English, Spanish, French, German, Hindi, Chinese, and more. We index and search Unicode text, so multi‑lingual docs are fully supported.

Yes. In your KnowFlow dashboard you can create multiple “Organizations” (formerly Projects), each pointing to its own set of sources—GitHub repos, Markdown files, Notion pages, PDFs, etc.—and deploy each assistant independently.

Instead of a generic model trained on the open web, KnowFlow trains directly on your docs and APIs, so answers are always accurate and context‑aware. It integrates semantic search + custom fine‑tuning via GROQ models and offers developer‑friendly features like code snippet support, versioned retraining, and self‑serve deployment.

Embedding is a one‑line install: copy our JavaScript snippet into your site or install the Slack/Discord bot from our dashboard. No backend changes required—just authorize the integration and your assistant goes live in minutes.

Retraining frequency depends on your plan: Free tier retrains once every 24 hours, Starter tier retrains daily, and Professional tier retrains automatically every 6 hours. You can also trigger a manual retrain in the dashboard at any time.

Absolutely. KnowFlow’s GROQ‑powered models are optimized for code: it can parse, highlight, and return fully formatted snippets in languages like JavaScript, Python, Java, Go, and more.

Yes. We currently support Markdown, HTML, PDF, and plain text. Support for image‑based docs (via OCR) is coming soon. All files are indexed and made searchable in context.

We offer a self‑hosted enterprise edition upon request. You can run our Docker images in your AWS, GCP, or Azure account and manage data storage entirely on‑premise or in your VPC.

All uploads are encrypted in transit (TLS 1.2+) and at rest (AES‑256). We store only vector embeddings by default, and raw files are optionally purged after indexing. You have full control