AI-powered terminal assistant that provides command suggestions and chat capabilities.

• 🤖 Natural Language to Commands: Convert natural language requests to bash commands
• 💬 Chat Mode: Direct conversation with AI
• ⚡ Multi-Session Support: Run multiple terminal sessions simultaneously
• 🛡️ Safety Indicators: Visual warnings for dangerous commands
• 🔄 Mode Switching: Switch between execution and chat modes
• 📝 Session Management: Join, detach, and switch between sessions

git clone <repository-url>
cd halpdesk
pip install -e .

Option A: Local install from built .deb (for testing):

1. Download the latest .deb from GitHub Actions artifacts or build locally:

   bash scripts/build_deb.sh 0.1.0
   sudo dpkg -i dist/halpdesk_0.1.0_amd64.deb

2. Enable the daemon as a user service (run as the target user):

   systemctl --user daemon-reload
   systemctl --user enable --now halpdesk.service

Option B: Build from a GitHub release tarball (stable):

# Build .deb from a tag (uses public repo tarball)
gem install --no-document fpm
bash scripts/build_deb_from_release.sh v0.1.0 asharalam11/HALpdesk
sudo dpkg -i dist/halpdesk_0.1.0_amd64.deb
systemctl --user daemon-reload && systemctl --user enable --now halpdesk.service

Option C: APT repository (recommended for users):

• Host an APT repo (Cloudsmith, PackageCloud, or Aptly).
• Publish the .deb built by CI on tags (workflow uploads to the GitHub Release; mirror it to your repo).
• Users add your signed repo and run apt install halpdesk.

Option A: Local formula tap (for testing):

1. Edit packaging/homebrew/halpdesk.rb to point to a released tarball URL and SHA256.
2. Install:

   brew install --formula packaging/homebrew/halpdesk.rb

3. Run the daemon on login:

   brew services start halpdesk

Option B: Dedicated tap (recommended):

• Create yourorg/homebrew-halpdesk and publish halpdesk.rb there.
• Users install with:

  brew tap yourorg/halpdesk
  brew install halpdesk

• Python 3.8+
• Either Ollama running locally OR OpenAI API key

1. Start the daemon:

   halpdesk-daemon

2. Start HAL in a terminal:

   halp

3. Try some commands:

   HAL> list files
   Suggested: ls -la 🟢
   Press Enter to execute, or type your own command
   
   HAL> /chat
   HAL (chat)> What does ls do?
   

• Check daemon health: curl -s http://127.0.0.1:8080/health
• Inspect provider and connectivity: curl -s http://127.0.0.1:8080/diagnostics | jq (or omit | jq)

• halp - Start new session or join existing
• halp list - List all active sessions
• halp join <session_id> - Join existing session
• halp new - Force create new session

• /sessions - List all sessions
• /switch <id> - Switch to different session
• /detach - Detach from session (keeps running)
• /chat - Switch to chat mode
• /exec - Switch to execution mode
• /help - Show help
• exit - Exit session

You can configure the daemon endpoint and the CLI via a config file or environment variables.

Config file (optional): ~/.config/halpdesk/config.toml

Example:

[server]
# Preferred: set a unified endpoint
endpoint = "http://127.0.0.1:8080"
# Or set host/port individually
# host = "127.0.0.1"
# port = 8080

[client]
# Where the `halp` CLI reaches the daemon
daemon_url = "http://127.0.0.1:8080"

Environment overrides:

export HALPDESK_DAEMON_ENDPOINT="http://127.0.0.1:8080"  # or use HALPDESK_DAEMON_HOST/PORT
export HALPDESK_DAEMON_URL="http://127.0.0.1:8080"       # used by halp CLI
export OPENAI_API_KEY="your-key-here"                     # if using OpenAI
export ANTHROPIC_API_KEY="your-key-here"                  # if using Claude
export OLLAMA_HOST="http://localhost:11434"               # if using Ollama
export HALPDESK_OLLAMA_BIN="/usr/local/bin/ollama"         # optional: binary path for auto-start
export HALPDESK_OLLAMA_AUTOSTART=1                         # set 0/false to disable

Provider configuration (optional) in `~/.config/halpdesk/config.toml`:

```toml
[providers]
default = "ollama"  # or "openai" / "claude"

  [providers.openai]
  base_url = "https://api.openai.com/v1"
  model = "gpt-3.5-turbo"

  [providers.claude]
  base_url = "https://api.anthropic.com"
  model = "claude-3-haiku-20240307"

  [providers.ollama]
  base_url = "http://localhost:11434"
  model = "llama2"
  # optional: where to find the 'ollama' binary; if omitted, PATH is used
  # binary = "/usr/local/bin/ollama"

Ollama auto-start: When the default or selected provider is Ollama and the server
is not reachable at `providers.ollama.base_url`, the daemon attempts to launch
`ollama serve` in the background (respecting `OLLAMA_HOST` derived from the URL).

## Architecture

Terminal 1 ─┐ Terminal 2 ─┼─→ HALpdesk Daemon ─→ Ollama/OpenAI Terminal 3 ─┘

Each terminal maintains its own session with independent context and history.

## Safety Features

Commands are analyzed for safety:
- 🟢 Safe commands  
- 🟡 Caution required
- 🔴 Potentially dangerous

## Development

```bash
pip install -e ".[dev]"
pytest