NavamAI supercharges you with AI

Use NavamAI to supercharge your productivity and workflow with personal, fast, and quality AI. Turn your Terminal into a configurable, interactive, and unobtrusive personal AI app. Power of 15 LLMs and 7 providers at your finger tips. Pair with workflows in Markdown, VS Code, Obsidian, and GitHub. Get productive fast with three simple commands.

Command Is All You Need: So, the LLM science fans will get the pun in our tagline. It is a play on the famous paper that introduced the world to Transformer model architecture - Attention is all you need. With NavamAI a simple command via your favorite terminal or shell is all you need to bend an large or small language model to your wishes. NavamAI provides a rich UI right there within your command prompt. No browser tabs to open, no apps to install, no context switching... just pure, simple, fast workflow. Try it with a simple command like ask "create a table of planets" and see your Terminal come to life just like a chat UI with fast streaming responses, markdown formatted tables, and even code blocks with color highlights if your prompt requires code in response! NavamAI has released 14 commands to help customize your personal AI workflow.

Top Models and Providers: You can switch private models or hosted frontier LLMs with ease. NavamAI comes with configurable support for more than 15 leading models (GPT 4o, Sonnet 3.5, Gemini 1.5 Pro, Mistral NeMo, Sonar Llama 3.1...) from 7 providers (Ollama, Anthropic, OpenAI, Groq, Google, Bedrock, and Perplexity).

Markdown Workflows: NavamAI works with markdown content (text files with simple formatting commands). So you can use it with many popular tools like VS Code and Obsidian to quickly and seamlessly design a custom workflow that enhances your craft.

Do More With Less: NavamAI is very simple to use out of the box as you learn its handful of powerful commands. As you get comfortable you can customize NavamAI commands simply by changing one configuration file and align NavamAI to suit your workflow. Everything in NavamAI has sensible defaults to get started quickly.

Make It Your Own: When you are ready, everything is configurable and extensible including commands, models, providers, prompts, model parameters, folders, and document types. Another magical thing happens when the interface to your generative AI is a humble command prompt. You will experience a sense of being in control. In control of your workflow, your privacy, your intents, and your artifacts. You are completely in control of your personal AI workflow with NavamAI.

Quick Start

Go to a folder where you want to initialize NavamAI. This could be your Obsidian vault or a VC Code project folder or even an empty folder.

pip install -U navamai # upgrade or install latest NavamAI
navamai init # copies config file, quick start samples
navamai id # identifies active provider and model
ask "How old is the oldest pyramid?" # start prompting the model

Models Setup

You will need to setup model-provider keys. Edit your environment config ~/.zshrc like so.

export ANTHROPIC_API_KEY= # https://docs.anthropic.com/en/api/getting-started
export OPENAI_API_KEY= # https://openai.com/index/openai-api/
export GROQ_API_KEY= # https://console.groq.com/keys
export GEMINI_API_KEY= # https://ai.google.dev/gemini-api
export PERPLEXITY_KEY= # https://www.perplexity.ai/settings/api

If you do not want to use any of the model then all you need to do is remove the corresponding entries from navamai.yml in the model-mapping and the provider-model-mapping sections. Also ensure that the other model configs only refer to available models.

For local models install Ollama and download the latest models you want to run.

Now you are ready to test all models and providers.

navamai test ask
navamai test vision

Python Environment Setup (optional)

Skip this section if you have your Python environment already working.

First, you should be running the latest Python on your system with Python package manager upgraded to the latest.

python --version
# should return Python 3.12.x or higher as on Sep'24

1. Optionally install latest Python on Mac.
2. Safely install NavamAI. Follow this thread to setup pyenv version manager.

First change to the directory to your VS Code workspace, Obsidian vault, or folder where you want to install NavamAI. Next create the virtual environment.

python -m venv env

Now you can activate your virtual environment like so. You will notice that directory prefixed with (env) to indicate you are now running in the virtual environment.

. env/bin/activate

To leave the virtual environment using deactivate command. Re-enter using same command as earlier.

Now you are ready to install NavamAI.

Command Reference

*Note that navamai, ask, image and refer are four top level commands available to you when you install NavamAI.

| Command | Example and Description | | ------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | *ask | ask "your prompt"<br>Prompt the LLM for a fast, crisp (default up to 300 words), single turn response<br><br>ask<br>Browses the configured prompts folder, lists prompt templates for user to run. | | audit | navamai audit<br>Analyze your own usage of NavamAI over time with an insightful command line dashboard and markdown report. | | config | navamai config ask save true<br>Edit navamai.yml file config from command line | | gather | navamai gather "webpage url"<br>Cleanly scrape an article from a webpage, without the menus, sidebar, footer. Includes images. Saves as markdown formatted similar to the HTML source.<br> <br>Refer the scraped markdown content using refer gather command.<br>Use vision on scraped images using navamai vision command. | | *refer | refer text-to-extract<br>Browse a set of raw text files and run custom prompts to extract new structured content.<br><br>refer inline-prompts-to-run<br>Run prompts embedded inline within text to expand, change, and convert existing documents.<br><br>refer intents<br>Expand a set of intents and prompts within an intents template<br><br>refer your-command<br>You can configure your own extensions to refer command by simply copying and changing any of the existing refer-existing model configs. | | run | navamai run<br>Browse Code folder for markdown files with code blocks, create, setup, and run the app in one single command!<br><br>The Code folder markdown files with code blocks are created using ask command running Create Vite App.md prompt template or similar. | | id | navamai id<br>Identifies the current provider and model for ask command<br><br>navamai id section-name<br>Identifies the provider and model defined in specific section