canvascoding/canvas-notebook v1.0.0

v1.0.0 — Canvas Notebook

on GitHub

Canvas Notebook v1.0.0

Canvas Notebook is a self-hosted AI-powered workspace combining a code editor, file browser, terminal, spreadsheet viewer, and a capable AI agent — all in one app.

Features

Workspace & File Management

• File browser with tree navigation, drag-and-drop, context menu, and search
• Upload and download files between local machine and workspace
• Resizable split-pane layout (file browser / editor / chat)

Code & Document Editor

• CodeMirror-based editor with syntax highlighting for all common languages
• Markdown editor with live preview
• PDF, image, audio, and video viewers
• DOCX editor and PPTX viewer
• Excel/CSV spreadsheet editor with formulas, formatting, and filtering

AI Agent Chat

• Real-time streaming chat with Server-Sent Events
• Persistent conversation history (SQLite)
• Multiple sessions per user
• Tool calling with live feedback
• Token usage tracking and cost estimation

Multi-Provider AI Support

• Anthropic Claude (direct API)
• Google Gemini (direct API)
• OpenRouter (access to many models)
• Ollama (local models)
• Dynamic provider/model switching at runtime
• Provider health dashboard and diagnostics

Agent Tools

• ls, read, write, mkdir, grep, glob — full workspace file access
• bash — shell command execution in workspace context
• qmd — semantic workspace search (BM25, vector, query modes)
• Image generation, video generation, ad localization tools
• Automation job management tools

Dynamic Skill System

• Load custom skills from /data/skills/
• Context skills (injected into system prompt) and callable tool skills
• Enable/disable individual skills or all at once via Skills Gallery

Browser-based Terminal

• Full PTY sessions over WebSocket
• Up to 3 concurrent terminals per user
• Resize support, 30-minute idle timeout

Image Generation

• Google Gemini image generation
• Batch generation (1–4 images), aspect ratio selection
• Images saved directly to workspace

Video Generation (VEO Studio)

• Text-to-video, frames-to-video, references-to-video, extend-video
• Up to 4K resolution, async generation with progress tracking

Ad Localization

• Localize ad images to target markets via Gemini vision
• Preserves layout and typography, translates text

Workflow Automation

• Schedule AI agent jobs: once, daily, weekly, or on interval
• Timezone support, manual trigger, run history and logs
• Automatic retry on failure

Settings & Integrations

• Manage all provider API keys in one place
• OAuth flow for provider auth
• Onboarding wizard for new users

Authentication

• Email/password login via better-auth
• Role-based access (admin/user)
• Optional signup restriction (ALLOW_SIGNUP=false)
• Bootstrap admin from env vars

Self-Hosted & Docker-ready

• Single Docker Compose setup
• Configurable data path (DATA env)
• Workspace, SQLite DB, and skills all under one data directory

Planned

The following features are in progress or planned for upcoming releases:

• Browser Skill — give the agent the ability to browse the web
• Debugging — better error visibility and step-through for agent tasks
• Sub-Agents — spawn and coordinate multiple agents within one task
• Multiple Sessions Simultaneously — run several agent conversations in parallel
• Automation Optimization — smarter scheduling, triggers, and retry strategies
• HTML Artifacts — render live HTML/JS outputs directly in the chat interface