Your local AI coding assistant. No subscription required.
Billy.sh is a Copilot CLI alternative with local Ollama by default and paid support for custom OpenAI-compatible endpoints. It runs entirely on your machine in local mode, works offline, and has no recurring cost for the core local experience. Built with Go and the Charm terminal toolkit for a polished terminal experience.
| Library | Role |
|---|---|
| Bubble Tea | TUI framework (Elm-inspired state machine) |
| Lipgloss | Terminal styling & layout |
| Glamour | Markdown + syntax highlighting in the terminal |
| Bubbles | UI components (viewport, textarea, spinner, list) |
| Ollama | Local LLM inference server |
- 💬 Interactive TUI — full-screen chat with scrollable history, built with Bubble Tea and Lipgloss
- 🤖 Local-first AI — Ollama by default, with paid support for custom OpenAI-compatible endpoints
- 🤖 Agentic mode (default) — Billy detects shell commands in responses and asks permission to run them, Copilot-style
- 🧠 Memory system — Billy learns about you over time; just say "remember that..."
- 📜 Conversation history — resume past sessions with an interactive
/historypicker - 🗜 Context compaction —
/compactsummarizes long conversations to stay within model context - 💾 Session checkpoints —
/sessionsaves named AI-generated checkpoints you can reload - 🎨 Syntax highlighting — code blocks rendered with full markdown support
- 🔄 Model management — list and switch models on any backend; pull new ones directly from Ollama
- 🔑 License tiers — Free / Pro / Premium / Team / Enterprise; activate with
/activate, free a seat with/deactivate - 🖥 One-shot mode — run
billy "explain this"directly from the terminal without launching the TUI - ⚙️ Simple config — single TOML file at
~/.localai/config.toml - 💾 SQLite backend — all history, memories, and settings stored locally at
~/.localai/history.db
- Go 1.24+ (build from source only)
- Ollama running locally on
localhost:11434for the default local mode - Or a paid custom endpoint configured via
backend.type = "custom"
brew install jd4rider/billy/billycurl -fsSL https://raw.githubusercontent.com/jd4rider/billy-app/main/scripts/install.sh | bashThen install Ollama if you haven't already.
curl -fsSL https://raw.githubusercontent.com/jd4rider/billy-app/main/scripts/install.sh | bash -s -- --fullThe full variant (~80 MB) bundles a headless Ollama binary that Billy starts automatically — no separate install needed.
scoop bucket add billy https://github.com/jd4rider/scoop-billy
scoop install billygit clone https://github.com/jd4rider/billy-app.git
cd billy-app
go build -o billy ./cmd/billy
./billy| Variant | Ollama | Binary size |
|---|---|---|
| Slim | Detects/prompts — you install Ollama | ~10 MB |
| Full | Bundled + auto-starts headless | ~80 MB |
Alpha — stable enough to use, still moving fast. Star the repo to follow along.
# Make sure Ollama is running with a model pulled
ollama pull qwen2.5-coder:14b # Billy's default model
# Launch the TUI
billy
# Or use one-shot mode directly
billy "explain what this codebase does"
billy read ./src
billy fix broken_code.go
cat error.log | billy "what's causing this?"Run Billy directly from your terminal without launching the TUI:
# Ask anything
billy "what does a Dockerfile ENTRYPOINT do?"
# Analyze a file or entire directory
billy read ./src
billy read main.go
# Explain what code does
billy explain internal/tui/chat.go
# Review and suggest fixes
billy fix broken_code.go
# Run a file — Billy diagnoses errors automatically
billy run script.py
# Pipe input
cat error.log | billy "what's causing this?"
git diff | billy "summarize these changes"Type / in the chat to open the command picker — scroll through all commands with arrow keys and press Enter to select.
| Command | Description |
|---|---|
/history |
Browse past conversations (arrow keys + Enter to resume) |
/resume <id> |
Jump directly to a conversation by ID |
/save |
Save the current conversation |
/clear |
Clear the current chat |
/compact |
Summarize & compress conversation to free up context |
/session |
Save a named session checkpoint (AI-generated summary) |
/session list |
List all saved checkpoints |
/session load <name> |
Restore a session from a saved checkpoint |
| Command | Description |
|---|---|
/backend |
Show the active backend and config path |
/backend reload |
Reload backend settings from ~/.localai/config.toml |
/model |
List models exposed by the current backend |
/model <name> |
Switch to a different model |
/pull <name> |
Download a model from the Ollama library (local backend only) |
/mode agent |
Enable agentic mode (default) — auto-detects and runs shell commands |
/mode chat |
Disable command detection — pure conversation mode |
/mode teach |
Teaching mode — Socratic guidance, step-by-step (coming soon) |
| Command | Description |
|---|---|
/memory |
List everything Billy remembers about you |
/memory forget <id> |
Remove a specific memory |
/memory clear |
Wipe all memories |
| Command | Description |
|---|---|
/pwd |
Print current working directory |
/cd <path> |
Change directory — type /cd to open live directory picker |
/ls [path] |
List files and directories (with counts) |
/git |
Show git branch, status, and recent commits |
/suggest <task> |
Suggest the best shell command for a natural-language task |
/explain <cmd> |
Explain what a shell command does, flag by flag |
Tip: Type
/cd(with a space) and the picker pops up with all subdirectories. Use ↑↓ to navigate,..to go up, or keep typing to filter by name.
| Command | Description |
|---|---|
/run <cmd> |
Run a shell command with permission prompt |
When Billy suggests a shell command in a bash block, it will prompt:
┌─ Run command? ──────────────────┐
│ git commit -m "fix typo" │
│ [Y]es [A]lways [N]o │
└─────────────────────────────────┘
- Y / Enter — run once
- A — always run this command type for this session
- N / S — skip (cancels pending queue)
| Command | Description |
|---|---|
/activate |
Prompt for your Billy license key and activate this machine |
/deactivate |
Release this machine's seat back to your license |
/license |
Show current license tier and status |
| Command | Description |
|---|---|
/help |
Show all available commands |
/quit |
Exit Billy |
--version |
Print version, commit, date, and build variant |
You don't need a command — just tell Billy:
remember that I prefer TypeScript over JavaScript
save that my project uses PostgreSQL
don't forget I'm building a SaaS product
Billy detects the intent and stores it automatically, then injects relevant memories into future conversations.
The current mode and working directory are shown in the status bar at the bottom of the TUI.
| Mode | Badge | Behaviour |
|---|---|---|
| Agent (default) | [AGENT] cyan |
Detects bash blocks in responses, queues them for permission-gated execution |
| Chat | [CHAT] dim |
No command detection — pure conversation |
| Teach (coming soon) | [TEACH] green |
Socratic guidance; shows commands as "type this yourself" prompts |
| Tier | Price | Limits |
|---|---|---|
| Free | $0 | 20 messages/session, 5 history slots, local Ollama only |
| Pro | $19 one-time | Unlimited messages, all backends, full history, all commands |
| Premium | $49 one-time | Pro + future voice mode, IDE plugins, priority support |
| Team | ~$14/seat | Bulk seats, shared memory, admin controls |
| Enterprise | Custom | Unlimited seats, self-hosted, SLA — call 406-396-7246 |
Upgrade at billy.sh, then run /activate inside Billy.
~/.localai/config.toml:
[backend]
type = "ollama"
url = "http://localhost:11434"
[ollama]
model = "qwen2.5-coder:14b"
temperature = 0.7Custom endpoint example for paid tiers:
[backend]
type = "custom"
url = "https://openrouter.ai/api/v1"
model = "anthropic/claude-3.7-sonnet"
api_key = "sk-..."| Key | Default | Description |
|---|---|---|
backend.type |
ollama |
Backend (ollama or paid custom) |
backend.url |
http://localhost:11434 |
Backend base URL |
backend.model |
unset | Model for paid custom endpoints |
backend.api_key |
unset | API key for paid custom endpoints |
ollama.model |
qwen2.5-coder:14b |
Default model for Ollama |
ollama.temperature |
0.7 |
Sampling temperature (0.0–1.0) |
Environment variable overrides:
BILLY_MODEL=llama3 billy
BILLY_BACKEND_TYPE=custom \
BILLY_BACKEND_URL=https://openrouter.ai/api/v1 \
BILLY_BACKEND_MODEL=anthropic/claude-3.7-sonnet \
BILLY_API_KEY=sk-... \
billy| Status | Feature |
|---|---|
| ✅ | Interactive TUI (Bubble Tea + Lipgloss) |
| ✅ | Local Ollama backend |
| ✅ | Conversation history (SQLite) |
| ✅ | Memory system (natural language detection) |
| ✅ | Interactive history picker |
| ✅ | Model list, switch & pull |
| ✅ | Agentic mode — shell command detection & permission prompts |
| ✅ | /run shell execution |
| ✅ | License system — Free / Pro / Premium / Team / Enterprise |
| ✅ | Lemon Squeezy activation with encrypted local activation storage |
| ✅ | Binary distribution — slim + fat (bundled Ollama) |
| ✅ | Homebrew tap, Scoop bucket, .deb/.rpm/.apk packages |
| ✅ | billy.sh landing page live |
| ✅ | Starlight docs site |
| ✅ | One-shot CLI mode (billy "prompt", billy read/explain/fix/run) |
| ✅ | Paid custom / OpenAI-compatible HTTP backends |
| ✅ | Context compaction (/compact) with token estimate in status bar |
| ✅ | Session checkpoints (/session, /session list, /session load) |
| ✅ | /pwd, /cd with live directory autocomplete picker |
| ✅ | /ls, /git, /suggest, /explain shell tools |
| ✅ | Working directory shown abbreviated in status bar |
| 🔜 | Teaching mode (/mode teach) + admin controls |
| 🔜 | Groq / Billy relay presets |
| 🔜 | Integration tests |
| 🔜 | Voice mode (Whisper + Piper TTS) |
| 🔜 | IDE plugins (VS Code, JetBrains) |
| 🔜 | Standalone chat app |
| 🔜 | iPhone companion app |
billy-app/
├── cmd/
│ └── billy/ # Main entry point (TUI + one-shot dispatch)
├── internal/
│ ├── backend/ # AI backend clients (Ollama + custom OpenAI-compatible endpoints)
│ ├── config/ # TOML config + env var overrides
│ ├── launcher/ # Ollama detection, start, embed (slim/fat)
│ ├── license/ # Lemon Squeezy activation/validation + tier constants
│ ├── memory/ # Memory detection & system prompt builder
│ ├── oneshot/ # Headless one-shot execution (no TUI)
│ ├── store/ # SQLite: history, memories, kv (encrypted), checkpoints
│ └── tui/ # Bubble Tea UI (chat view, history picker)
├── scripts/
│ ├── install.sh # Installer (slim or --full)
│ └── fetch-ollama.sh # Download ollama binary for fat CI builds
├── .goreleaser.yml
├── go.mod
└── README.md
This is an alpha project — feedback, issues, and PRs are very welcome. Open an issue or start a discussion.
MIT — see LICENSE
