vayu

Vayu — AI-powered coding assistant for the terminal. Built in Rust.

Features

• Multi-provider: Anthropic (Claude), OpenAI, Azure, Ollama, MLX (Apple Silicon)
• Interactive TUI: Rich terminal UI with markdown rendering, tables, code blocks
• Tool calling: bash, read_file, write_file, edit_file, glob, grep
• Inline diffs: File edits show colored diff snippets with context
• Conversation history: JSONL storage with SQLite index, resume/continue sessions
• Image support: /image <path>, Ctrl+V clipboard paste, drag-and-drop
• Streaming: Live token streaming in both TUI and exec modes
• Mouse scroll + copy: Scroll wheel works alongside native text selection
• Tool approval: Per-call approval, approve-all ([a]), or --yolo mode
• Context tracking: Live context size and peak usage in status bar

Install

Prerequisites

• macOS (primary target)
• Rust (stable, edition 2024)

# Install Rust if needed
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source "$HOME/.cargo/env"

Download prebuilt binary (no Rust needed)

Download the latest release for your platform from GitHub Releases:

# macOS Apple Silicon (M1/M2/M3/M4)
curl -L https://github.com/Sravani2201/Vayu/releases/latest/download/vayu-aarch64-apple-darwin.tar.gz | tar xz
sudo mv vayu /usr/local/bin/

# macOS Intel
curl -L https://github.com/Sravani2201/Vayu/releases/latest/download/vayu-x86_64-apple-darwin.tar.gz | tar xz
sudo mv vayu /usr/local/bin/

# Linux x86_64
curl -L https://github.com/Sravani2201/Vayu/releases/latest/download/vayu-x86_64-unknown-linux-gnu.tar.gz | tar xz
sudo mv vayu /usr/local/bin/

Install from source (requires Rust)

git clone https://github.com/Sravani2201/Vayu.git
cd Vayu
cargo install --path crates/vayu-cli

Install from GitHub directly

cargo install --git https://github.com/Sravani2201/Vayu.git vayu-cli

Build without installing

cargo build --release
./target/release/vayu --help

Quick Start

# Interactive TUI (current directory as workspace)
vayu

# One-shot prompt
vayu "explain this code"

# Auto-approve all tools
vayu --yolo

# Continue last conversation
vayu --continue    # or vayu -c

# Pick a past conversation
vayu --resume      # or vayu -r

# Verbose logging (writes to ~/.vayu/logs/)
vayu -v

Usage

CLI Flags

Flag	Short	Description
--continue	-c	Resume the most recent conversation
--resume	-r	Pick a past conversation interactively
--session <ID>		Resume a specific session by UUID
--yolo		Auto-approve all tool calls
--model <M>	-m	Override model (e.g. ollama:llama3.2:8b)
--autonomy <A>	-a	Override autonomy: ask, auto-read, auto-all
--verbose	-v	Debug logging to ~/.vayu/logs/
--format <F>		Output format: human (default), json

Subcommands

vayu exec "prompt"          # One-shot execution
vayu history [--limit N]    # List past conversations
vayu config                 # Print resolved configuration
vayu version                # Show version info

TUI Keyboard Shortcuts

Key	Action
Enter	Send message
Shift+Enter	Newline
Ctrl+C	Quit / Cancel
Ctrl+L	Clear screen
Ctrl+U / Ctrl+K	Kill line
PgUp / PgDn	Scroll chat
Up / Down	Scroll (empty input) or history (with text)
Ctrl+V	Paste image from clipboard
Ctrl+O	Expand/collapse tool output

TUI Slash Commands

Command	Description
/help	Show help
/clear	Clear conversation
/cost	Show token usage and context size
/history	List past conversations
/image <path>	Attach an image
/model <name>	Switch model
/quit	Exit

Tool Approval

When a tool needs approval, you'll see:

• [y] — approve this call
• [a] — approve all for the rest of the session
• [n] — deny

Or start with vayu --yolo to skip all prompts.

Configuration

Config files are loaded in order (later overrides earlier):

1. Built-in defaults
2. ~/.vayu/config.toml (user-level)
3. .vayu/config.toml (project-level)
4. Environment variables (VAYU_MODEL, VAYU_AUTONOMY)
5. CLI flags

Example: Anthropic

[model]
provider = "anthropic"
model = "claude-sonnet-4-20250514"
max_tokens = 16384

[providers.anthropic]
api_key = "sk-..."  # or set ANTHROPIC_API_KEY

Example: Azure OpenAI

[model]
provider = "azure"
model = "gpt-4o"
max_tokens = 32768

[providers.azure]
chat_url = "https://your-resource.openai.azure.com/openai/v1/chat/completions"
deployment = "gpt-4o"
api_key = "..."  # or set AZURE_OPENAI_API_KEY

Example: MLX (Apple Silicon local)

[model]
provider = "mlx"
model = "mlx-community/Qwen3.5-9B-MLX-4bit"
max_tokens = 32768

[providers.mlx]
auto_start = true
default_model = "mlx-community/Qwen3.5-9B-MLX-4bit"
server_command = "mlx_vlm.server"
python_path = "/path/to/venv/bin/python3"
port = 8080

Example: Ollama

[model]
provider = "ollama"
model = "llama3.2:8b"

[providers.ollama]
base_url = "http://localhost:11434"
default_model = "llama3.2:8b"
auto_pull = false

Tool Configuration

[tools]
autonomy = "ask"  # ask | auto-read | auto-all
shell_auto_approve = ["git status", "git diff", "ls"]
file_edit_auto_approve = false

Data Storage

Path	Purpose
~/.vayu/config.toml	User configuration
~/.vayu/sessions/*.jsonl	Conversation data (JSONL, source of truth)
~/.vayu/index.db	SQLite index for listing/search
~/.vayu/logs/vayu-*.log	Verbose debug logs (one per session)

Architecture

Vayu is a Rust workspace with these crates:

Crate	Purpose
vayu-cli	CLI entry point, argument parsing
vayu-tui	Terminal UI (ratatui), markdown rendering
vayu-exec	Headless/non-interactive runner
vayu-core	Agent loop, session management, events
vayu-provider	LLM providers (Anthropic, OpenAI, Azure, MLX, Ollama)
vayu-tools	Built-in tools (bash, file ops, glob, grep)
vayu-context	Conversation history, JSONL+SQLite store
vayu-config	Configuration loading and merging
vayu-sandbox	Tool approval policies
vayu-mcp	Model Context Protocol support

Development

cargo test          # Run all tests
cargo build         # Debug build
cargo build --release  # Optimized build

License

MIT