Hrisa Code
Local AI Coding Assistant with autonomous multi-step task execution
Overview
Hrisa Code is a sophisticated AI-powered CLI coding assistant that brings autonomous development capabilities to your terminal. Built on Ollama for local LLM execution, it offers three powerful modes: Normal (conversational), Agent (autonomous multi-step), and Plan (complex task decomposition with adaptive execution). The project's Plan Mode represents a significant innovation in AI coding assistants, featuring automatic task complexity detection, dynamic execution plan generation, and step context passing that reduces redundant tool calls by 40-50% and parameter errors by 70%. With multi-model orchestration, the system can route different tasks to specialized models (qwen2.5:72b for reasoning, deepseek-coder-v2 for code understanding), ensuring optimal performance for each operation. Safety features include comprehensive approval systems, loop detection, and goal tracking to prevent runaway execution. The assistant excels at progressive context building, intelligently exploring repositories, generating comprehensive documentation (README, API docs, HRISA.md), and executing complex development workflows with minimal human intervention.
Key Highlights
- •Plan Mode with automatic task complexity detection (SIMPLE/MODERATE/COMPLEX)
- •40-50% reduction in redundant tool calls through intelligent step context passing
- •70% reduction in parameter errors via built-in validation checklists
- •Multi-model orchestration routing tasks to specialized LLMs
- •Comprehensive approval system for write operations and destructive commands
- •Loop detection preventing infinite repetition (max 3 identical calls)
- •Goal tracking with automatic task completion detection
- •Autonomous documentation generation (README, API, HRISA.md)
- •Background task management with async command execution
- •Rich visual feedback with animated spinners and persistent mode indicators
- •48 automated tests with >80% coverage
- •Docker deployment ready with docker-compose support
Technologies Used
Technical Challenges
- →Balancing autonomy with safety - preventing runaway executions
- →Optimizing context passing to reduce redundant operations
- →Handling diverse task complexities with appropriate strategies
- →Multi-model coordination and selection logic
- →Comprehensive testing of autonomous agent behavior
Impact
Demonstrated 40-50% efficiency gains in development workflows through software improvements alone, enabling developers to leverage local LLMs for autonomous coding assistance without cloud dependencies or privacy concerns.