Project Roadmap
Relevant source files
The following files were used as context for generating this wiki page:
The Caro project roadmap outlines the strategic plan for developing a safe, AI-powered shell command generation tool. The development is structured into three major milestones, each building upon the last: establishing a robust core, launching publicly with comprehensive documentation, and innovating with advanced AI and distributed features. The overarching goal is to create a local-first, safety-first, and open-source tool that democratizes terminal access safely.
The roadmap follows a phased approach, starting with the production-ready v1.0.1 release and progressing through v1.1.0, v1.2.0, and culminating in the v2.0.0 release, which introduces significant new capabilities.
Sources: presentation/roadmap-slides.md:1-17, presentation/slides.md:1-10, CONTRIBUTING.md:21-34
Timeline Overview
The development plan for 2026 is divided into three distinct phases, each corresponding to a major version release and focusing on a specific strategic theme.
This Gantt chart illustrates the planned timeline for the three major milestones in the first half of 2026.
Sources: presentation/roadmap-slides.md:41-48
The strategic evolution follows a clear progression:
- Q1 (Foundation & Performance): Solidify the core application, focusing on performance, reliability, and infrastructure.
- Q2 (Public Launch & Community): Establish a public presence with a professional website, documentation, and community-building initiatives.
- Q3 (Innovation & Advanced AI): Introduce cutting-edge features, including distributed intelligence and voice synthesis.
Sources: presentation/roadmap-slides.md:50-58, presentation/roadmap-slides.md:219-227
v1.1.0: Core Improvements
This foundational phase focuses on making the Caro CLI production-ready by enhancing performance, reliability, and core infrastructure.
Sources: presentation/roadmap-slides.md:60-89
Key Deliverables & Success Criteria
| Deliverable | Success Criteria |
|---|---|
| Hugging Face Model Download | Stable, offline-capable model caching and download functionality. |
| Performance Optimization | Startup time under 100ms and first inference under 2 seconds on an M1 Mac. |
| LLM Evaluation Harness | A comprehensive test suite for evaluating LLM performance and accuracy. |
| Benchmark Suite | Established performance benchmarks for startup time and inference speed. |
| Safety Enhancements | Refined risk assessment, enhanced pattern matching, and improved user confirmation flows. |
Sources: presentation/roadmap-slides.md:69-89, specs/001-create-a-comprehensive/tasks.md:79-83
Focus Areas
The work for v1.1.0 is divided into four key areas:
| Area | Tasks |
|---|---|
| Performance | Startup time optimization, lazy loading, memory profiling, and benchmark infrastructure. |
| Safety | Enhanced pattern matching, POSIX compliance checks, and risk assessment refinement. |
| Infrastructure | Hugging Face model integration, automated GitHub releases, and self-healing features. |
| Testing | LLM evaluation harness, property-based tests, and local Chroma DB indexing for advanced tool use patterns. |
Sources: presentation/roadmap-slides.md:95-132, specs/001-create-a-comprehensive/tasks.md:79-83
v1.2.0: Website & Docs Launch
This phase is centered around the public launch of Caro, establishing a professional online presence, and fostering a community.
Sources: presentation/roadmap-slides.md:134-163
Key Deliverables & Success Criteria
| Deliverable | Success Criteria |
|---|---|
| Interactive Website | caro.sh is live with an interactive terminal demo and is SEO optimized. |
| Astro Starlight Docs | Comprehensive documentation including installation guides, API docs, and usage examples. |
| Marketing Materials | A multi-channel announcement plan and a pipeline for user-generated content (UGC). |
| Community Engagement | Active GitHub discussions, an attribution page, and initial i18n support (Simplified Chinese). |
| Strategic Roadmap | A public 12-month roadmap and a clear articulation of the "batteries included" philosophy. |
Sources: presentation/roadmap-slides.md:143-163
v2.0.0: Advanced Features
The innovation phase introduces cutting-edge capabilities that expand Caro beyond simple command generation into a distributed intelligence system.
Sources: presentation/roadmap-slides.md:165-194, specs/005-karo-distributed-mesh/spec.md
Key Deliverables & Success Criteria
| Deliverable | Success Criteria |
|---|---|
| Karo Distributed Intelligence | A functional prototype of the "Karo System" for multi-terminal coordination and cross-device sync ("Jazz"). |
| Voice Synthesis | Integration of a natural "Caro character" voice for audio feedback and a voice-first interface. |
| Dogma Rule Engine | An advanced, policy-based rule engine for security hardening and organizational policy enforcement. |
| Exo Cluster Support | Support for distributed inference across "Exo clusters". |
| Self-Healing Features | Automatic error recovery and context-aware smart caching. |
Sources: presentation/roadmap-slides.md:174-194, specs/005-karo-distributed-mesh/spec.md:1-10
Karo Distributed Mesh System
A core component of v2.0.0 is the Caro Distributed Mesh, a system for peer-to-peer node communication within closed or air-gapped networks.
Core Concepts:
- Node Identity: Each Caro installation has a unique cryptographic identity using an Ed25519 keypair.
- Secure Communication: All inter-node communication uses TLS 1.3 with mutual authentication.
- Data Classification: Data is tiered into Raw (L0), Summarized (L1), and Aggregated (L2) levels. Raw data never leaves the local node.
- Privacy-Preserving Aggregation: The system allows for querying aggregate usage patterns across the mesh without exposing individual command history.
- Access Control: A trust model with levels (e.g., Peer, Supervisor) and per-node policies governs data sharing.
Sources: specs/005-karo-distributed-mesh/spec.md:1-10, specs/005-karo-distributed-mesh/spec.md:73-145
Future Ideas
Beyond the defined roadmap, several innovative concepts are under consideration for future development.
Community Governance & Safety Council
A democratic process for managing safety rules, allowing the community to propose, review, and vote on new safety patterns. This ensures that safety decisions are transparent and not controlled by a single entity.
This diagram shows the proposed workflow for community-governed safety pattern updates.
Sources: presentation/slides.md:278-294
Static & Hybrid Generation
To improve performance for common commands, a hybrid approach is planned. Frequently used commands will be pre-compiled into a static database for instant retrieval, while novel requests are sent for full model inference. The system will learn over time to optimize this balance.
This flowchart illustrates the hybrid model, combining instant static responses with AI-powered inference.
Sources: presentation/slides.md:310-337
Summary
The Caro roadmap is a comprehensive plan that balances foundational stability with forward-looking innovation. It begins by hardening the core application's performance and reliability (v1.1.0), moves to public engagement and documentation (v1.2.0), and culminates in advanced, distributed AI features (v2.0.0). The strategic themes of Foundation, Launch, and Innovation guide the project's evolution, with a constant emphasis on safety, performance, and community-driven development.