v0.8.8 Semantic search — ONNX embeddings + OpenAI API support

Your AI knowledge
assistant, built for
the desktop.

Chat with your documents, build a private knowledge base, and learn anything with eight AI‑powered formats. Knovex runs entirely on your machine — your files never leave it. Now with semantic search via local ONNX or OpenAI embeddings.

Download for your OS — detecting… All releases press D to jump to downloads
v0.8.8 Free & open source — MIT license No Python, Node or Docker needed SHA‑256 verified releases
Semantic search — ONNX + OpenAI embeddings 100% offline — no cloud required RAG over PDF · DOCX · XLSX · TXT · MD 8 learn formats + quiz & flashcards API keys encrypted at rest (Fernet AES) OpenAI · Anthropic · Groq · Gemini · Ollama
knovex — Knowledge / Q4 Analysis
⌘K search anything
Knowledge / Q4 Analysis · Semantic search enabled

What drove the user retention improvement?

📎 5 sources 🔍 semantic · FTS5 ⏱ 0.9s 🧠 gpt-4o-mini
GT
What were the three biggest drivers of retention improvement in Q4, and how did they compare to Q3?
Retention improved 23 % YoY in Q4, driven by three changes: onboarding redesign (+9 pp), push‑notification tuning (+8 pp), and the new "continue learning" feature (+6 pp). Compared to Q3, all three metrics accelerated — onboarding alone was flat at +2 pp in Q3.
Citations: Q4-exec-summary.pdf · p.7 retention-dashboard.xlsx product-review.docx · §3
› Ask a follow‑up, or type /learn to generate a quiz from these docs… ↵ send
1 ms
Avg. query latency
8
Learn formats
6+
LLM providers
100%
Offline capable
0
Telemetry / analytics
01Download

One installer per platform.
Everything bundled inside.

No Python, no Node, no command‑line gymnastics. Ships as a single native binary that bundles the backend, database and UI — auto‑updates and stays out of your way.

Detected
Windows
Windows 10 / 11 — 64‑bit
~85 MB .exe NSIS
Knovex‑Setup‑0.8.8.exe x64

Run the installer → Next → Next → Install → Finish. Knovex appears in your Start Menu and Desktop.

Windows says "Don't run" or "Unrecognised app"?

Knovex is open-source and unsigned. Windows SmartScreen flags all new unsigned apps. Two quick steps to proceed:

  1. Browser download bar — Chrome / Edge may show a warning. Click the ⋯ or ^ next to the file → Keep → Keep anyway.
  2. SmartScreen popup — "Windows protected your PC" appears. Click More info (small link, bottom-left) → then the blue Run anyway button.
  3. The installer opens normally. Knovex is installed to your user folder — no admin password required.
Detected
macOS
macOS 12 Monterey and later
~92 MB .dmg · x64
Knovex‑0.8.8.dmg Intel x64

Open the .dmg, drag Knovex to Applications, launch. On first open Gatekeeper may flag it — right‑click → Open → Open to allow.

Looking for older versions? Browse all releases → Secure · Fast · Reliable · Cost-Effective
02Features

Everything you need,
nothing you don't.

Built for privacy‑conscious professionals. All compute runs on‑device. Cloud is optional — and only ever for the LLM call.

🔍
Semantic Search New
Dense vector embeddings via local ONNX (~45 MB, one‑time download) or the OpenAI Embeddings API. Hybrid FTS5 + ANN + RRF fusion for best recall.
💬
Document Chat
Ask questions across PDFs, Word, Excel and Markdown via RAG. Every answer cites the exact page and section it came from.
🧠
Knowledge Bases
Organise files into named KBs. Index once, query at ~1 ms with SQLite FTS5. Stale‑file detection keeps the index in sync automatically.
🎓
Learn Mode
Turn any document or topic into a quiz, flashcard set, mind map, timeline, story, ELI5, speed‑learn or brainstorm session.
🔧
Any LLM Provider
OpenAI, Anthropic, Groq, Gemini, AWS Bedrock, Cerebras, Ollama — swap without restarting. Powered by LiteLLM under the hood.
🔒
Encrypted at Rest
All API keys are stored with Fernet AES‑128 symmetric encryption. Keys live in your OS user directory, readable only by your account.
🌐
Web Search
Ground answers in live web results via DuckDuckGo (free), Serper.dev or Brave Search. Toggle per chat — sources cited inline.
📄
Document Reader
In‑app rendering for PDF, DOCX, XLSX, CSV and Markdown. Inline Q&A on any open file — no switching to an external viewer.
📊
File Summaries
One‑click brief or detailed summary for any file or an entire knowledge base. Exportable as Markdown.
03Setup

Up and running in four steps.
Or fewer, if you use Ollama.

Installation is identical across platforms — pick your model provider once, drop in your documents, and ask away.

STEP 01

Download & install

Run the installer for your OS. Knovex installs in seconds — backend, frontend and database all bundled. No Python or Node needed.

STEP 02

Add a model

Open Settings · LLM and paste an OpenAI, Anthropic or Groq key — or point to a local Ollama. Keys are encrypted on save.

STEP 03

Build a knowledge base

Drag and drop PDFs, DOCX, XLSX or Markdown into a KB. Knovex chunks, indexes with FTS5, and optionally embeds for semantic search — all locally.

STEP 04

Chat & learn

Ask questions with cited sources, toggle web search, or hit /learn to generate a quiz, flashcard set, story, mindmap or timeline.