Skip to content

waveai.mdx

Latest commit

 

History

History
110 lines (77 loc) · 4.24 KB

waveai.mdx

File metadata and controls

110 lines (77 loc) · 4.24 KB
sidebar_position 1.5
id waveai
title Wave AI

import { Kbd } from "@site/src/components/kbd"; import { PlatformProvider, PlatformSelectorButton } from "@site/src/components/platformcontext";



Context-aware terminal assistant with access to terminal output, widgets, and filesystem.

Keyboard Shortcuts

Shortcut Action
Toggle AI panel
Focus AI input
Clear chat / start new
Send message
New line

Widget Context Toggle

Controls AI's access to your workspace:

ON: AI can read terminal output, capture widget screenshots, access files/directories (with approval), navigate web widgets, and use custom widget tools. Use for debugging, code analysis, and workspace tasks.

OFF: AI only sees your messages and attached files. Standard chat mode for general questions.

File Attachments

Drag files onto the AI panel to attach (not supported with all models):

Type Formats Size Limit Notes
Images JPEG, PNG, GIF, WebP, SVG 10 MB Auto-resized to 4096px max, converted to WebP
PDFs .pdf 5 MB Text extraction for analysis
Text/Code .js, .ts, .py, .go, .md, .json, .yaml, etc. 200 KB All common languages and configs

CLI Integration

Use wsh ai to send files and prompts from the command line:

git diff | wsh ai -                          # Pipe to AI
wsh ai main.go -m "find bugs"                # Attach files with message
wsh ai $(tail -n 500 my.log) -m "review" -s  # Auto-submit with output

Supports text files, images, PDFs, and directories. Use -n for new chat, -s to auto-submit.

AI Tools (Widget Context Enabled)

Terminal

  • Read Terminal Output: Fetches scrollback from terminal widgets, supports line ranges

File System

  • Read Files: Reads text files with line range support (requires approval)
  • List Directories: Returns file info, sizes, permissions, timestamps (requires approval)
  • Write Text Files: Create or modify files with diff preview and approval (requires approval)

Web

  • Navigate Web: Changes URLs in web browser widgets

All Widgets

  • Capture Screenshots: Takes screenshots of any widget for visual analysis (not supported on all models)

:::warning Security File system operations require explicit approval. You control all file access. :::

Local Models & BYOK

Wave AI supports using your own AI models and API keys:

  • Local Models: Run AI models locally with Ollama, LM Studio, vLLM, and other OpenAI-compatible servers
  • BYOK (Bring Your Own Key): Use your own API keys with OpenAI, OpenRouter, Google AI (Gemini), Azure OpenAI, and other cloud providers
  • Multiple Modes: Configure and switch between multiple AI providers and models
  • Privacy: Keep your data local or use your preferred cloud provider

See the Local Models & BYOK guide for complete configuration instructions, examples, and troubleshooting.

Privacy

Default Wave AI Service:

  • Messages are proxied through the Wave Cloud AI service (powered by OpenAI's APIs). Please refer to OpenAI's privacy policy for details on how they handle your data.
  • Wave does not store your chats, attachments, or use them for training
  • Usage counters included in anonymous telemetry
  • File access requires explicit approval

Local Models & BYOK:

  • When using local models, your chat data never leaves your machine
  • When using BYOK with cloud providers, requests are sent directly to your chosen provider
  • Refer to your provider's privacy policy for details on how they handle your data

:::info Under Active Development Wave AI is in active beta with included AI credits while we refine the experience. Share feedback in our Discord.

Coming Soon:

  • Remote File Access: Read files on SSH-connected systems
  • Command Execution: Run terminal commands with approval
  • Web Content: Extract text from web pages (currently screenshots only) :::