Roadmap¶
This document outlines the development status and future plans for CLIver.
Completed Features¶
- Builtin Tools (17 total): File I/O, shell, web search/fetch, planning (todo), memory, identity, skill activation, workflow creation, Docker, ask user
- Skills System: LLM-driven skill activation from SKILL.md files (project and global scopes)
- Planner: Complexity-aware planning — Simple (direct), Medium (todo_write), Complex (create_workflow)
- Agent Profile: Instance-scoped resources with multi-agent isolation
- Memory & Identity: Persistent memory (append/rewrite) and identity profiles
- Workflow Simplification: Streamlined workflow engine with LLM-driven workflow generation
- Skill-Workflow Integration: Skills can be activated within workflow LLM steps
- Task Scheduling: CRUD task management with cron scheduling support
- Conversation History: Session persistence with LLM-based compression
- Token Usage Statistics: Per-model and per-session token tracking
- Permissions System: Layered permission system (persistent rules, session grants, workflow-scoped)
- Keyring Security: System keyring integration for API key management
- Multi-provider Support: Ollama, OpenAI-compatible, vLLM providers
Planned Features¶
- Deep Search:
cliver deep-search "questions"— multi-step research workflow with caching and resume - Multi-Model Orchestration: Different models assigned to different capabilities via
model_routingconfig - Remote Workflow Definitions: Support for fetching workflow definitions from remote sources
Thank you for your interest in contributing to CLIver!