navar 0.1.0

A CLI coding agent written in Python

Navar

Navigate your codebase, not your context window.

Navar is an interactive AI coding agent designed to provide context-aware assistance by indexing your local codebase and using RAG (Retrieval-Augmented Generation) to ground its responses.

Prerequisites

• Python 3.11+
• Docker (for running Qdrant)

Setup

1. Install Navar:

   pip install -e .
   

2. Start Qdrant:

   docker-compose up -d
   

3. Initialize & Index:

   navar init
   navar index .
   

Configuration

Create a .env file in your project root with your DeepSeek API key:

DEEPSEEK_API_KEY=your_key_here
DEEPSEEK_MODEL=deepseek-chat
QDRANT_HOST=localhost
QDRANT_PORT=6333
QDRANT_COLLECTION=navar
EMBEDDING_MODEL=all-MiniLM-L6-v2
CHUNK_SIZE=400
CHUNK_OVERLAP=80
TOP_K=5

CLI Subcommands

Command	Description
navar	Launch the interactive TUI
navar init	Initialize a new project and configuration
navar index <path>	Crawl and index a directory
navar clear	Clear the current vector collection
navar remove <file>	Remove a specific file from the index
navar status	Show collection statistics
navar models	List available DeepSeek models

TUI Slash Commands

Inside the Navar TUI, use these commands to manage your session:

Command	Action
/clear	Clear conversation history and screen
/model	Cycle through available DeepSeek models
/index <path>	Index a path from within the TUI
/remove <file>	Remove a file from the index
/status	Show collection information panel
/export [path]	Export current session to a markdown file
/help	Show all available commands
/exit	Quit the application

---

Built with Python, FastAPI, Qdrant, and DeepSeek.