RishiAi
is a command-line interface (CLI) AI ChatBot
that uses Google’s Gemini
API to generate natural language responses. You can chat with RishiAi about anything, from images to documents, and have fun and engaging conversations.
Recording.2024-06-18.183545.mp4
Installing RishiAi is simple. For global access, execute one of the commands below:
- Using Node Package Manager (NPM):
npm install -g rishiai
- or if you are using (PNPM):
pnpm install -g rishiai
Before using Rishi Ai, you'll need to log in with your Google Gemini API token. To do this, run the following command:
rishi login [token]
Example:
rishi login Flqw9TUeaSkRfRii3lmgZLid
This command sends a request to the Gemini endpoint for authentication. Upon receiving a successful response, your API key will be verified.
To obtain your free API key, head over to Google Gemini API Key.
If you encounter any issues, try running rishi login <token>
again to refresh your credentials.
To view your current RishiAi configuration, simply type:
rishi config
This will display the contents of the rishi.json
file stored on your machine.
To update the default values of RishiAi, run this command:
rishi config set
By default, when you log in, we have set some default values that are required for the Google Gemini API to work properly, such as:
maxOutputTokens
: 2048topK
: 40topP
: 1temperature
: 0.7kwargs
: 1
To update these default values, simply run rishi config set
. After running this command, you will be prompted to enter your desired values in place of the default values.
To start chatting with the RishiAi chatbot, run this command:
rishi chat
Once you are logged in, you can have a natural language conversation with the chatbot.
You can also chat with an image by running:
rishi vision <image path>
When you provide the path to an image, it will be converted to base64
. When you ask a question about the image, RishiAi
will send a request to the Gemini Vision
endpoint with the image data and your question. The response from Gemini will be used to generate a reply.
Example:
rishi vision public/image.jpg
You can ask questions like "What's happening in the image?", "Who are the people in the image?", or "What is the object on the left?".
You can also chat with a document by running:
rishi read <document path>
You need to provide the path
of your document, and optionally specify its file type
using the -f
flag. Default file type
is set to text
because it allows for easier splitting and chunking of content.
However, we generally recommend
using the text
file type for seamless conversations with any document.
Example:
rishi read <document path> -f pdf
RishiAi supports five file types: pdf
, text
, json
, csv
, and url
.
Note: The url
file type is included for convenience, allowing you to fetch data from websites that allow bots to scrape their web pages.
RishiAi also provides verbose
output, displaying the retrieval process of chunks based on your queries.
Example:
rishi read resume.pdf -f pdf -t
Fun Fact: We can't use -v
for verbose
flag as it is already assigned as version flag.
By default, RishiAi use MemoryVectorStore
to manage splitted chunks and indexes.
If you want to save the chunks and indexes to your local machine, you can use the -s
flag. This will create a directory with a nanoid in /user/.config/configstore/rishi
. You can also use the -n
flag to give a custom name
to the directory.
Example:
rishi read resume.pdf -f pdf -s -n resume-store
If you have already created and saved a vector store
on your machine, you can load it by using the -l
flag. This flag requires the location/path of the vector store directory.
Example:
rishi read resume.pdf -f pdf -l C:\Users\asus\.config\configstore\rishi\resume-store
RishiAi utilizes Google's Gemini
API for its chatbot capabilities. To obtain your free API key, head over to Google Gemini API Key.