Skip to content
| Marketplace
Sign in
Visual Studio Code>Machine Learning>CEVIZNew to Visual Studio Code? Get it now.
CEVIZ

CEVIZ

CEVIZ AI

| (0) | Free
Local Personal Hybrid AI — Local/Cloud/Hybrid AI chat in VS Code sidebar
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

CEVIZ

VS Code Marketplace Version License Platform

A VS Code extension that routes prompts between a local Ollama backend and cloud AI APIs — automatically, based on complexity.

No subscription. One-time purchase. Your data stays local.


Screenshot

Screenshot coming soon. Install and run the Setup Wizard to get started.


Why CEVIZ?

Cloud-only AI tools send every keystroke off-device and bill you monthly forever. Local-only tools are slow on complex tasks. CEVIZ does both: fast local inference for routine work, cloud escalation for hard problems — with your explicit approval before anything leaves your machine.

CEVIZ Personal ChatGPT Plus GitHub Copilot
Price $49 once $20/month $10/month
Cost at 12 months $49 $240 $120
Data stays local ✅ ❌ ❌
Offline support ✅ ❌ ❌
Obsidian vault RAG ✅ ❌ ❌
RAG self-development ✅ ❌ ❌
Bring your own API key ✅ ❌ ❌
Game-dev context ✅ General General

Key Features

🔀 Hybrid AI Routing
Scores each prompt on a 0–100 complexity scale (length, code blocks, domain keywords, paragraph count). Scores under 45 go local; 45–69 show an escalation banner; 70+ recommend cloud. You stay in control.

🔒 Privacy-First Architecture
Session data, project context, and vault notes never leave your machine. Cloud API keys are stored in VS Code SecretStorage. No telemetry, no analytics, no external logging.

📚 Obsidian Vault Integration
Indexes your vault with ripgrep (5-minute TTL cache, up to 5,000 notes). Relevant notes are injected into prompt context automatically. AI responses include [[wiki-link]] badges that open the source note with one click.

🧬 RAG Self-Development System (market-unique)
Ingests your whitepapers and RSS articles into a ChromaDB vector store. Proposes updates to its own system prompt based on absorbed knowledge. Every change requires your explicit approval — with diff preview and one-click rollback.

💬 Session & Project Management
Per-workspace session isolation with inline rename, pin, star, and sort. Generates CONTEXT.md per project directory and auto-detects completion keywords to keep project state current across sessions.

🌐 Multi-OS · 6 Languages
One-command backend installers for Linux (systemd), macOS (launchd), and Windows (WSL2 / native). UI localized in Korean, English, Turkish, Arabic, Persian, and Russian.


Quick Start

1 — Install the Extension

From the Marketplace:
Search CEVIZ in the VS Code Extensions panel, or:

code --install-extension ceviz-ai.ceviz

From a release .vsix:

# download ceviz-0.2.1.vsix from https://github.com/eonyakoh-tech/ceviz/releases
code --install-extension ceviz-0.2.1.vsix

2 — Install the Backend

The CEVIZ backend runs an Ollama inference engine alongside a FastAPI proxy. Download the installer from Releases.

Linux (Ubuntu / Debian)

chmod 700 install-linux.sh && bash install-linux.sh
# Preview without applying: bash install-linux.sh --dry-run
# English output:            bash install-linux.sh --lang=en

macOS (Intel / Apple Silicon)

chmod 700 install-macos.sh && bash install-macos.sh

Windows (WSL2 recommended)

Set-ExecutionPolicy RemoteSigned -Scope CurrentUser
.\install-windows.ps1
# Force native install (no WSL2): .\install-windows.ps1 -ForceNative

All installers are idempotent. Run the dependency checker first if you want to verify prerequisites:

bash check-dependencies.sh      # Linux / macOS
.\check-dependencies.ps1        # Windows

3 — Configure & Chat

  1. Click the CEVIZ brain icon in the Activity Bar to open the sidebar panel.
  2. Click ⚙️ Setup Wizard → pull gemma3:4b + nomic-embed-text (RAG required).
  3. Set ceviz.serverIp in VS Code Settings if your backend is on a remote host (default: localhost:8000).
  4. Type your first prompt. The routing indicator shows 🟢 Local / 🟡 Hybrid / 🔴 Cloud.

Tip: Press Ctrl+Alt+C in the editor to attach selected code to the next chat message.


Requirements

Component Minimum
VS Code ≥ 1.80
Backend OS Linux x86-64 / arm64 · macOS Intel / Apple Silicon · Windows 10+
RAM for local models 8 GB (gemma3:4b) · 16 GB (llama3.1:8b)
Disk ≥ 10 GB free for model weights
Network Required only for cloud API calls and initial model download
API keys (optional) Anthropic · Google Gemini — stored locally in vscode.SecretStorage

Pricing

Plan Price Devices Cloud quota
Personal $49 2 BYOK only
Pro $99 5 BYOK + 100 req/day hosted
Founder $149 10 BYOK + unlimited hosted
  • Free trial: 14 days, no credit card required.
  • BYOK: Bring your own Anthropic or Gemini API key — no markup, no proxy.
  • Offline grace: License re-validates every 7 days; works fully offline for 14 days after the last successful check.

Licenses available at LemonSqueezy (store coming soon).


Architecture

VS Code Extension (Webview)
        │
        │  prompt + complexity score (0–100)
        │
   ┌────▼──────────────────────────────────┐
   │         Routing Decision              │
   │  score < 45 → Local                  │
   │  45 ≤ score < 70 → Hybrid (ask user) │
   │  score ≥ 70 → Cloud (recommend)      │
   └────┬──────────────────┬──────────────┘
        │                  │
        ▼                  ▼
┌───────────────┐   ┌──────────────────────┐
│ Local Backend │   │   Cloud API (direct) │
│  ─────────── │   │  ──────────────────  │
│  Ollama       │   │  Anthropic Claude    │
│  FastAPI      │   │  Google Gemini       │
│  ChromaDB RAG │   │                      │
│  ripgrep      │   │  3-tier fallback:    │
│               │◄──│  Cloud A → Cloud B   │
│  (fallback)   │   │        → Local       │
└───────────────┘   └──────────────────────┘
        │
   Obsidian Vault (read-only, local)
   Project CONTEXT.md (auto-managed)
   Session store (vscode.globalState)

FAQ

Does CEVIZ send my code to the cloud?
Only if you explicitly choose Cloud mode or confirm an escalation prompt. In Local mode, all processing happens on your machine. The extension contains no telemetry.

Which local models are recommended?
The Setup Wizard recommends the best fit for your hardware. A reasonable starting point: gemma3:4b for general chat, qwen2.5-coder:1.5b (PN40-class hardware) or qwen2.5-coder:3b (16 GB RAM) for code tasks. nomic-embed-text is required for RAG.

Can I use my own API key?
Yes — all plans are BYOK. Keys are stored in VS Code's SecretStorage, encrypted at rest, and never transmitted to CEVIZ servers.

What happens if the backend goes offline?
CEVIZ caches up to 20 recent responses and serves keyword-matched answers from cache. An offline banner appears in the UI and clears automatically when the connection is restored.

Is my Obsidian vault modified?
Read-only. CEVIZ never modifies existing notes. It can create new notes in 00_Inbox/ when you explicitly save a chat response via the save dialog.


Contributing

Bug reports and feature requests are welcome via GitHub Issues.

For pull requests:

git clone https://github.com/eonyakoh-tech/ceviz.git
cd ceviz
npm install
npm run compile      # development build (webpack)
npm run watch        # watch mode
npx vsce package     # produce .vsix for local testing

Code style is enforced by eslint.config.mjs (camelCase/PascalCase, strict equality, semicolons). Run npx eslint src/ before opening a PR.


License

Proprietary. © 2026 EONYAK OH. All rights reserved.

This software is not open-source. See LICENSE for full terms. A 14-day evaluation period is granted to any user upon first install.

For licensing inquiries: eonyakoh@gmail.com

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft