Core CLI
Comprehensive argument parsing, colored output, and interactive confirmation flows.
This page shows the current development status of caro, what’s working, what’s in progress, and what’s planned for the future.
These features are fully implemented and working:
Core CLI
Comprehensive argument parsing, colored output, and interactive confirmation flows.
Embedded LLM
Local inference with Qwen 2.5 Coder 1.5B, no external dependencies required.
Safety Validation
52+ dangerous command patterns detected and blocked before execution.
Multi-Platform
Pre-built binaries for Linux (x64/ARM64), macOS (Intel/Silicon), and Windows.
| Feature | Status |
|---|---|
| CLI argument parsing | Done |
| Module architecture | Done |
| Backend trait system | Done |
| Configuration management (TOML) | Done |
| Multiple output formats (JSON, YAML, Plain) | Done |
| Feature | Status |
|---|---|
| Dangerous pattern detection (52+ patterns) | Done |
| POSIX compliance checking | Done |
| User confirmation workflows | Done |
| Risk assessment with color coding | Done |
| Critical path protection | Done |
| Feature | Status |
|---|---|
| Embedded MLX backend (Apple Silicon) | Done |
| Embedded CPU backend (cross-platform) | Done |
| Ollama local backend | Done |
| vLLM HTTP API support | Done |
| Response parsing with fallback | Done |
| Automatic backend selection | Done |
| Feature | Status |
|---|---|
| OS/architecture detection | Done |
| Shell type detection | Done |
| Available command detection | Done |
| Platform-specific rules (BSD vs GNU) | Done |
| 2-iteration agentic refinement loop | Done |
| Command execution engine | Done |
| Feature | Status |
|---|---|
| Published to crates.io | Done |
| Install script with binary fallback | Done |
| Multi-platform CI/CD | Done |
| SHA256 checksum verification | Done |
| Official website (caro.sh) | Done |
| Documentation site | Done |
Currently under active development:
| Feature | Status |
|---|---|
| Model downloading optimization | In Progress |
| Performance benchmarking suite | In Progress |
| Extended test coverage | In Progress |
| Command history tracking | In Progress |
| Learning from user feedback | Planned |
Features on the roadmap for future releases:
| Feature | Description |
|---|---|
| Multi-step goal completion | Execute complex workflows with dependency resolution |
| Shell script generation | Generate reusable scripts for complex workflows |
| Interactive refinement | Iteratively refine commands with explanations |
| Plugin system | Custom backends, validators, and extensions |
| Command templates | Save and reuse common command patterns |
| Feature | Description |
|---|---|
| Team configurations | Shared safety policies and backends |
| Audit logging | Track all generated commands |
| Custom model support | Bring your own fine-tuned models |
| Air-gapped deployment | Fully offline installation packages |
| SSO integration | Enterprise authentication support |
┌─────────────────────────────────────────────────────────────┐│ caro CLI │├─────────────────────────────────────────────────────────────┤│ Input Layer ││ ├── Argument Parser (clap) ││ ├── Configuration Loader (TOML) ││ └── Platform Detector │├─────────────────────────────────────────────────────────────┤│ Processing Layer ││ ├── Agentic Loop (2-iteration refinement) ││ ├── Context Enrichment ││ └── Backend Router │├─────────────────────────────────────────────────────────────┤│ Backend Layer ││ ├── Embedded (MLX/CPU) ││ ├── Ollama ││ └── vLLM │├─────────────────────────────────────────────────────────────┤│ Safety Layer ││ ├── Pattern Matcher (52+ patterns) ││ ├── Risk Assessor ││ └── Confirmation Flow │├─────────────────────────────────────────────────────────────┤│ Output Layer ││ ├── Formatter (JSON/YAML/Plain) ││ ├── Execution Engine ││ └── Shell Integration │└─────────────────────────────────────────────────────────────┘Want to help build these features? Check out: