YUVA Code gives an AI model real tools — file read/write/edit, shell access, and grep — so it can actually work on your codebase. It supports 14 providers out of the box, including free options.

Quick Start

npm install -g yuva-code
yuva --setup     # pick a provider and model
yuva             # start coding

Providers

| Provider | Cost | Key | |----------|------|-----| | NVIDIA NIM | Free | build.nvidia.com | | Ollama | Free (local) | No key needed | | LM Studio | Free (local) | No key needed | | Jan | Free (local) | No key needed | | Google Gemini | Paid | aistudio.google.com | | OpenAI | Paid | platform.openai.com | | Anthropic Claude | Paid | console.anthropic.com | | DeepSeek | Paid | platform.deepseek.com | | Moonshot Kimi | Paid | platform.moonshot.cn | | Groq | Paid | console.groq.com | | Alibaba Qwen | Paid | dashscope.console.aliyun.com | | xAI Grok | Paid | console.x.ai | | OpenRouter | Paid | openrouter.ai | | Custom API | — | Any OpenAI-compatible endpoint |

Switch providers anytime with /provider.

Usage

yuva                  # start interactive chat
yuva --setup, -s      # run setup wizard
yuva --help, -h       # show help
yuva --version, -v    # show version

Type a request like add a hello function to src/foo.js. The model reads files, edits them, and runs shell commands as needed.

Tools

The model has access to 7 tools:

| Tool | Permission | Description | |------|------------|-------------| | read_file | Auto | Read file contents (capped at 256 KB) | | list_files | Auto | List directory entries (recursive up to 5 levels) | | grep_search | Auto | Search for patterns across the codebase | | write_file | Ask | Write a file (creates parent dirs) | | edit_file | Ask | Replace a unique substring in a file | | shell | Ask | Run a shell command (5 min timeout) | | delete_file | Ask | Delete a file or empty directory |

Destructive tools (write_file, edit_file, shell, delete_file) prompt for approval. Type a to allow all for the session.

Slash Commands

| Command | Description | |---------|-------------| | /help | Show all commands | | /model | Switch AI model (interactive picker) | | /provider | Switch AI provider | | /undo | Revert the last file change | | /diff | Show all files changed this session | | /save | Export conversation to markdown | | /cost | Show token usage stats | | /clear | Clear conversation and reset | | /config | Show config path and masked API key | | /cd <path> | Change working directory | | /exit | Quit |

Shell Escape

Prefix any input with ! to run a shell command directly (no approval prompt):

❯ !git status
❯ !npm test
❯ !ls -la

Features

• Multi-provider — 14 providers including 4 free/local options
• Streaming — responses stream in real-time with markdown rendering
• Tab completion — slash commands auto-complete
• Undo — /undo reverts the last file change
• Session export — /save dumps the full conversation to a markdown file
• Auto-context — project tree and package.json are sent to the model on first message
• Loop detection — stops the model if it calls the same tool with the same args repeatedly
• Rate limit retry — automatically retries on 429/502/503/504 with exponential backoff

Requirements

• Node.js >= 20
• An API key for your chosen provider (free for NVIDIA NIM, Ollama, LM Studio, Jan)

Config

Config is stored at ~/.yuva-ai/config.json. Override the directory with:

set YUVA_CONFIG_DIR=C:\my-config    # Windows
export YUVA_CONFIG_DIR=/my-config   # macOS/Linux

License

MIT