Hey there, it’s Adela 👋 Welcome to Star History!
Each month, we gather open-source highlights — and occasionally feature a Starlet project that shines a little brighter. ✨
Follow @StarHistoryHQ for daily GitHub gems!
In this issue, we feature Starlet #30: NPC Shell - Intelligent Bash Environment with AI Agents.
NPC Shell transforms your terminal into an intelligent workspace that extends the traditional bash environment with AI capabilities. Inspired by biological intelligence, it provides a unified command-line interface for seamless interaction with multiple LLM providers and specialized agents -- enabling multi-modal collaboration, code execution, and automation of complex tasks.
npcsh provides a ReAct-style agentic shell with tool capabilities through Jinxs. It processes natural language commands and routes to appropriate tools and NPCs using reasoning and action loops.
corca is an MCP tool-calling agent specialized for software development workflows. It uses Model Context Protocol for extended capabilities and code analysis.
guac is a coding agent that directly executes code in an active Python session. You can inspect outputs and it builds on code you input, creating an interactive development environment.
Multi-Provider Support - OpenAI, Anthropic, Gemini, Deepseek, Ollama, LMStudio, vLLM, MLX integration through LiteLLM.
Data Management - Database support for PostgreSQL, SQLite, and other databases for command history and conversations. Knowledge graph for structured information stored within the same database with knowledge derived with context (path, npc, team, etc) in mind, ChromaDB for vector embeddings and RAG, persistent state between sessions.
Extensible Framework - Jinxs are Jinja execution templates that provide function-like capabilities, allowing users to build reusable tools that reference each other. NPC files (.npc) for agent definitions, Context files (.ctx) for team preferences to make context management easier, Assembly lines for workflow pipelines.
SQL Models with NPC Integration - Create SQL models that call NPCs directly within queries using nql. functions. Execute complex data processing workflows where NPCs analyze, summarize, or transform data as part of SQL operations. Support coming soon for PostgreSQL, SQLite, Snowflake, Databricks, and other databases.
npcsh comes with an extensive set of command macros:
/alicanto - Deep research with multiple perspectives, identifying gold insights and cliff warnings
/corca - Enter MCP-powered agentic shell for development
/guac - Enter interactive coding shell with pomodoro workflow
/plonk - Computer use and GUI automation with vision models
/vixynt - Image generation and editing
/wander - Temperature-based problem solving for thinking outside the box
/yap - Voice chat mode with speech-to-text and text-to-speech
/rag - Semantic search with ChromaDB embeddings and command history
/ots - Take screenshot and analyze with vision models
And many more for search, configuration, knowledge graphs, and workflow automation
npcsh is MIT-licensed and actively maintained. Here are some links to get started:
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。