bhai-cli 0.2.5

BHAI — The Dual-Brain AI Coding Agent with Punjabi Swagger. Indic-native voice + universal LLM tool-calling.

🇮🇳 BHAI-CLI

The Dual-Brain Agentic Coding CLI — 22 Indian Languages. One CLI.

भाई · ভাই · ભાઈ · ਭਾਈ · భాయ్ · ಭಾಯ್ · भाऊ · ଭାଇ

BHAI is a lean, zero-state resilient CLI tool that acts as your 10x Senior SDE from India. It uses a Dual-Brain Manager-Worker Architecture — a VibeLLM for persona, planning, and review, and a CoderAgent for autonomous tool-calling execution — all powered by litellm for universal LLM provider support and Sarvam AI for 22 Indian language understanding.

"You are BHAI, a 10x Senior SDE from India. Sab bhasha samajhte ho. You roast O(N²) loops and unoptimized Dockerfiles."

---

✨ Features

• 🧠 Dual-Brain Agentic Architecture: VibeLLM (Manager/Planner) delegates step-by-step tasks to the CoderAgent (Worker) which autonomously executes tools in a loop.
• 🤖 MCP-Powered File Operations: Full agentic capability to read_file, write_file, edit_file, and list_directory autonomously to build features from scratch.
• 💾 Persistent Session Management: Chat histories are saved locally with LLM-auto-generated titles. Resume past context anytime from the interactive menu.
• 🖥️ Interactive Menu-Driven CLI: Simply run bhai-code for a rich, beautifully colored terminal UI to chat, manage sessions, and configure settings.
• 🔒 Secure Secrets Storage: API keys are strictly separated into a 0600 permission secrets.yaml and masked during screen display.
• 🇮🇳 22 Indian Languages: Speak in Hindi, Punjabi, Bengali, Tamil, Telugu, Gujarati, Marathi, Kannada, Malayalam, and more.
• 🌍 Universal LLM Support: DeepSeek, xAI (Grok), Qwen, OpenAI, Anthropic, Groq, Google Gemini, Sarvam, Mistral, Cerebras, Ollama, Bedrock, and more.
• 🎙️ Indic Voice Interface: Sarvam Saaras v3 (STT) translates any Indian language to English, Bulbul v3 (TTS) reads responses back.
• ⚡ Zero-State Resilience: Boots in <300ms, persistent YAML config, hot-swappable model overrides.

---

🚀 Installation

pip install bhai-cli

With voice support (optional):

pip install "bhai-cli[audio]"
# Also needs: sudo apt install libportaudio2  (Linux)

---

⚙️ Setup

bhai-code setup

The interactive wizard will configure:

1. VibeLLM — Your planner/persona brain (e.g., sarvam/sarvam-105b, groq/llama-3.3-70b-versatile)
2. CoderAgent — Your autonomous worker brain (e.g., openai/gpt-4o, anthropic/claude-3-5-sonnet-20241022)
3. Sarvam AI — Voice interface API key (optional)
4. Ollama Port — Local inference configuration

Model Format

Uses litellm's model naming:

Provider	Format	Example
Anthropic	anthropic/model	anthropic/claude-3-5-sonnet-20241022
OpenAI	openai/model	openai/gpt-4o
Google	gemini/model	gemini/gemini-2.5-pro
DeepSeek	deepseek/model	deepseek/deepseek-coder
xAI	xai/model	xai/grok-beta
Groq	groq/model	groq/llama-3.3-70b-versatile
Sarvam	sarvam/model	sarvam/sarvam-105b
Ollama	ollama/model	ollama/llama3

---

💬 Usage

Interactive Menu (Recommended)

# Opens the rich terminal UI for chats, session management, and settings
bhai-code

Text Mode

# Directly start a new chat session interactively
bhai-code text

# Single prompt execution
bhai-code text "explain the pyproject.toml file"

# Resume a specific session
bhai-code text --session-id <uuid>

# Override models on the fly
bhai-code text "optimize this Dockerfile" --vibe-model groq/llama-3.3-70b-versatile --coder-model openai/gpt-4o

Voice Mode

# Live microphone
bhai-code listen

# From audio file
bhai-code listen --file recording.wav

View Config

bhai-code config

---

🏗️ Architecture

User Input (Any of 22 Indian Languages / English)
         │
         ▼
┌─────────────────────┐
│   VibeLLM (Brain A)  │  ← Persona + Planning + Session Management
│   Sarvam-105B / any  │
└──────────┬──────────┘
           │ {"task": "step-by-step instructions"}
           ▼
┌─────────────────────┐
│  CoderAgent (Brain B)│  ← Autonomous Execution Loop
│  Any litellm model   │
│                      │
│  Tools:              │
│  • read/write/edit   │  ← File System Ops
│  • execute_bash      │  ← [Y/n] Guardrail
│  • web_search        │  ← DuckDuckGo
│  • codebase_context  │  ← AST signatures
└──────────┬──────────┘
           │ Results
           ▼
┌─────────────────────┐
│   VibeLLM (Brain A)  │  ← Review & Final Output
└─────────────────────┘

---

🔧 Tools (MCP-Powered)

Tool	Description
read_file	Read the exact contents of a file before editing
write_file	Create a new file with provided content
edit_file	Target specific line ranges to replace code blocks
list_directory	Explore project structure to navigate codebase
execute_bash	Run shell commands (tests, docker, git) with [Y/n] confirmation
web_search	DuckDuckGo search for real-time debugging
codebase_context	Smart directory scanner with AST-based signature extraction

---

📁 Project Structure

bhai-cli/
├── src/bhai_cli/
│   ├── __init__.py          # Package metadata
│   ├── cli.py               # Rich interactive Typer CLI
│   ├── config_manager.py    # Configuration & secure secrets
│   ├── session_manager.py   # Chat history persistence & auto-titling
│   ├── orchestrator.py      # Dual-Brain Agentic Loop
│   ├── audio_engine.py      # Sarvam STT/TTS
│   └── tools/
│       ├── __init__.py      # Tool registry + JSON schemas
│       ├── file_ops.py      # Read, Write, Edit, List tools
│       ├── bash.py          # execute_bash with guardrails
│       ├── search.py        # web_search via DuckDuckGo
│       └── codebase.py      # codebase_context with AST
├── tests/
├── website/                 # Next.js Landing Page Frontend
├── pyproject.toml
├── bhai_cli_documentation.html # Exhaustive local docs (gitignored)
└── README.md

---

🔑 Environment Variables & Secrets

API keys are securely isolated in ~/.config/bhai/secrets.yaml (0600 permissions). Supported keys include: SARVAM_API_KEY, VIBE_API_KEY, CODER_API_KEY, GROQ_API_KEY, OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, DEEPSEEK_API_KEY, XAI_API_KEY.

---

📄 License

Apache-2.0 — see LICENSE.

Built with ❤️ from India by Madhav Kapila.